systems and methods for automated detection in magnetic resonance imaging

专利摘要:
The present invention relates to a method for detecting change in the degree of midline deviation in a patient's brain. while the patient remains positioned within the low field magnetic resonance imaging device, acquiring the first magnetic resonance imaging (mr) data and the second mr imaging data from the patient's brain; provide the first and second mr data as input to a statistical classifier trained to obtain the corresponding first and second outputs, identify from the first output at least one initial location of at least one landmark associated with at least one line structure average of the patient's brain; identify from the second exit at least one updated location of at least one landmark; and determining a degree of change in midline offset using at least one starting location of at least one landmark and at least one updated location of at least one landmark.
公开号:BR112019010225A2
申请号:R112019010225
申请日:2017-11-21
公开日:2019-08-20
发明作者:L Charvat Gregory；M Rothberg Jonathan；Sofka Michal；S Ralston Tyler
申请人:Hyperfine Res Inc；
IPC主号:

专利说明:

Descriptive Report of the Invention Patent for SYSTEMS AND METHODS FOR AUTOMATED DETECTION IN MAGNETIC RESONANCE IMAGES ”.
Cross Reference to Related Orders [001] This application claims benefits under 35 USC § 119 (e) of US provisional patent application No. 62 / 425,569, entitled METHODS AND CHANGE DETECTION APPARATUS, filed on November 22, 2016, which is incorporated by reference here in its entirety.
Background [002] Magnetic resonance imaging (MRI) imaging provides an important imaging modality for numerous applications and is widely used in clinical and research facilities for imaging inside the human body. MRI is based on the detection of magnetic resonance (MR) signals, which are electromagnetic waves emitted by atoms in response to changes in state resulting from applied electromagnetic fields. For example, nuclear magnetic resonance (NMR) techniques involve the detection of MR signals emitted from the nuclei of excited atoms (for example, atoms in human body tissue). Detected MR signals can be processed to produce images, which, in the context of medical applications, allow the investigation of internal structures and / or biological processes within the body, for the purposes of diagnosis, therapy and / or research.
[003] MRI provides an attractive imaging modality for the creation of biological images, due to its ability to produce non-invasive images with relatively high resolution and contrast without concerns for the safety of other modalities (for example, without needing expose the subject to ionization radiation, such as X-rays, or without introducing radioactive material into the
Petition 870190046951, of 20/05/2019, p. 35/196
2/110 body). In addition, MRI is particularly well suited for providing soft tissue contrast, which can be exploited to image matter, which other imaging modalities cannot reproduce satisfactorily. Furthermore, MR techniques can capture information about biological structures and / or processes that other modalities are unable to acquire. However, there are several disadvantages to conventional MRI techniques that, for a particular imaging application, may include relatively high equipment cost, limited availability (for example, difficulty and costs in obtaining access to MRI scanners) clinical), the duration of the image acquisition process, etc.
[004] The trend in clinical MRI has been to increase the field strength of MRI scanners to improve one or more of the scanning time, image resolution, and image contrast, which in turn increases costs of MRI imaging. A vast majority of installed MRI scanners operate using at least 1.5 or 3 tesla (T), which refers to the field strength of the scanner's main magnetic field B0. An approximate cost estimate for a clinical MRI scanner is on the order of a million dollars per tesla, which is not even included in the substantial costs of operating, servicing and maintaining the operation of such MRI scanners.
[005] Additionally, conventional high field MRI systems typically require large superconducting magnets and associated electronic parts to generate a uniform and strong static magnetic field (B0), in which a subject (for example, a patient) is represented by images . Superconducting magnets additionally require cryogenic equipment to keep conductors in a superconducting state. The size of such systems is considerable with
Petition 870190046951, of 20/05/2019, p. 36/196
3/110 an MRI installation including multiple rooms for magnetic, electronic components, thermal management system and control console areas, including a specially protected room to isolate the magnetic components of the MRI system. The size and cost of MRI systems generally limit their use to facilities, such as hospitals and academic research centers, that have sufficient space and resources to purchase and maintain them. The high cost and substantial space requirements of high-field MRI systems result in limited availability of MRI scanners. As such, there are often clinical situations in which an MRI scan would be beneficial, but it is impractical or impossible due to the limitations described above and as discussed in greater detail below.
Summary [006] Some modalities are directed at a method of detecting change in the degree of deviation from the midline in a patient's brain positioned within a low-field magnetic resonance imaging (MRI) device, the method comprising: while the patient remains positioned inside the low-field MRI device: acquire the first magnetic resonance imaging (MR) data from the patient's brain; provide the first MR data as input to a trained statistical classifier to obtain the first corresponding output; identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain; acquiring the second MR image data from the patient's brain following the acquisition of the first MR image data; providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output; identify, from the second exit,
Petition 870190046951, of 20/05/2019, p. 37/196
4/110 at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location of at least one landmark and at least one updated location of at least one landmark. [007] Some modalities are directed to a low-field magnetic resonance imaging device, configured to detect the change in degree of midline deviation in a patient's brain positioned within a resonance imaging device low field magnetic (MRI) means the low field MRI device comprising: a plurality of magnetic components, including a magnet B0 configured to produce, at least in part, a magnetic field B0; at least one gradient magnet configured to spatially encode MRI data; and at least one radio frequency coil configured to stimulate an magnetic resonance response and detect the magnetic components configured to, when operated, acquire magnetic resonance image data; and at least one controller configured to operate the plurality of magnet components for, while the patient remains positioned within the low field magnetic resonance device, acquires the first magnetic resonance imaging (MR) data from the patient's brain, and acquires the second MR image data from the patient's brain after the acquisition of the first MR image data, where at least one controller is additionally configured to perform: provide the first and second MR data as recorded in a statistical classifier trained to obtain the first exit and the second corresponding exit; identify, from the first exit, at least one initial location of at least one landmark associated with at least one line structure
Petition 870190046951, of 20/05/2019, p. 38/196
5/110 average of the patient's brain; identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location, of at least one landmark, and at least one updated location of at least one landmark.
[008] Some modalities are directed to at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, causes at least one computer hardware processor perform a method of detecting change in the degree of midline deviation in a patient's brain positioned inside a low field magnetic resonance imaging (MRI) device. The method comprises, while the patient remains positioned inside the low field MRI device, acquiring the first magnetic resonance imaging (MR) data from the patient's brain; provide the first MR data as recorded to a trained statistical classifier to obtain the first corresponding output; identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain; acquiring the second MR imaging data from the patient's brain following the acquisition of the first MR imaging data; provide the second MR imaging data as input to the trained statistical classifier to obtain the corresponding second output; identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain;
Petition 870190046951, of 20/05/2019, p. 39/196
6/110 and determine a degree of change in the midline deviation using at least one initial location of at least one landmark and at least one updated location of at least one landmark.
[009] Some modalities are directed to a system that comprises at least one computer hardware processor and at least one computer-readable, non-transitory storage medium executing instructions executable by processor that, when executed by at least one hardware processor make at least one computer hardware processor perform a method of detecting change in the degree of deviation from the midline in a patient's brain positioned within a magnetic resonance imaging (MRI) device. low field. The method comprises, while the patient remains positioned inside the low field MRI device, acquiring the first magnetic resonance imaging (MR) data from the patient's brain; provide the first MR data as input to a trained statistical classifier to obtain the first corresponding output; identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain; acquire the second MR image data from the patient's brain as recorded in the trained statistical classifier to obtain the corresponding second output; identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location of at least one landmark and at least one updated location of at least one landmark.
[0010] Some modalities are directed to a method of
Petition 870190046951, of 20/05/2019, p. 40/196
7/110 determination of change in the size of an anomaly in a patient's brain positioned within a low-field magnetic resonance imaging (MRI) device, the method comprising, while the patient remains positioned within the MRI device low field, acquire the first magnetic resonance imaging (MR) data from the patient's brain; provide the first MR image data as input to a trained statistical classifier to obtain the first corresponding output; identifying, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality of the patient's brain; acquiring the second MR image data from the patient's brain following the acquisition of the first MR image data; provide the second MR image data as recorded in the trained statistical classifier to obtain the corresponding second output; identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain; determine the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic. [0011] Some modalities are directed to a low field magnetic resonance imaging (MRI) device configured to determine the change in the size of an anomaly in a patient's brain, the low field MRI device comprising a plurality magnetic components including a magnet B0 configured to produce, at least in part, a magnetic field B0; at least one gradient magnet configured to spatially encode MRI data; and at least one radio frequency coil configured to stimulate an MRI response and detect the mag components
Petition 870190046951, of 20/05/2019, p. 41/196
8/110 configured to, when operated, acquire magnetic resonance image data; and at least one controller configured to operate the plurality of magnet components so that, while the patient remains positioned within the low field magnetic resonance device, acquire the first magnetic resonance imaging (MR) data from the patient's brain, and acquire the second MR image data from the patient's brain after the acquisition of the first MR image data, where at least one controller is additionally configured to function, providing the first and second MR image data as recorded in a trained statistical classifier to obtain first and second corresponding outputs; identifying, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain; acquiring the second MR imaging data for the part of the patient's brain subsequent to the acquisition of the first MR imaging data; identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the patient's brain anomaly; determine the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[0012] Some modalities are directed to at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, make at least one computer hardware processor perform the method of determining the change in the size of an anomaly in the brain of a patient positioned inside a magnetic resonance imaging (MRI) device
Petition 870190046951, of 20/05/2019, p. 42/196
9/110 low field, the method comprising, while the patient remains positioned inside the low field MRI device, acquiring the first magnetic resonance (MR) imaging data of the patient's brain; provide the first MR image data as recorded in a trained statistical classifier to obtain the first corresponding output; identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain; acquiring the second MR imaging data from the patient's brain following the acquisition of the first MR imaging data; provide the second MR imaging data as recorded in the trained statistical classifier to obtain the corresponding second output; identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain; determine the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[0013] Some modalities are directed to a system, comprising at least one computer hardware processor; at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform the size change determination method of an anomaly in a patient's brain positioned inside a low-field magnetic resonance imaging (MRI) imaging device. The method comprises, while the patient remains positioned inside the low field MRI device, acquiring the first image generation data
Petition 870190046951, of 20/05/2019, p. 43/196
10/110 gem by magnetic resonance (MR) of the patient's brain; provide the first MR imaging data as recorded in a trained statistical classifier to obtain the first corresponding output; identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain; acquiring second MR imaging data from the patient's brain after acquiring the first MR imaging data; provide the second MR imaging data as recorded in the trained statistical classifier to obtain the corresponding second output; identify, using the second output, at least an updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain; and determining the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[0014] Some modalities are directed to a method of detecting change in the biological matter of a patient positioned within a low-field magnetic resonance imaging (MRI) generating device, the method comprising, while the patient remains positioned within the low field MRI device; acquire the first data from magnetic resonance images of a part of the patient; acquire the second magnetic resonance imaging data from the patient after the acquisition of the first magnetic resonance imaging data; align the first magnetic resonance imaging data and the second magnetic resonance imaging data; and compare the first magnetic resonance imaging data and the second magnetic resonance imaging data aligned to detect at least one
Petition 870190046951, of 20/05/2019, p. 44/196
11/110 change in biological matter on the part of the patient.
[0015] Some modalities are directed to a low field magnetic resonance imaging device configured to detect the change in the biological matter of a patient positioned within the low field magnetic resonance imaging device, comprising a plurality of magnetic components, including a B0 magnet configured to produce, at least in part, a B0 magnetic field, at least one gradient magnet configured to spatially encode MRI data; and at least one radiofrequency spiral configured to stimulate an MRI response and detect the magnetic components configured to, when operated, acquire magnetic resonance imaging data and at least one controller configured to operate the plurality of magnet components for , while the patient remains positioned inside the low field MRI device, acquiring the first MRI data from a part of the patient, and acquiring the second MRI data from the patient, subsequent to the acquisition of the first magnetic resonance imaging data, the at least one controller additionally configured to align the first magnetic resonance imaging data and the second magnetic resonance imaging data and compare the first magnetic resonance datamagnetic resonance imaging and the second magnetic resonance imaging data aligned to detect at least one change in biological matter on the part of the patient.
[0016] Some modalities are directed to at least one non-transitory computer-readable storage medium weapon
Petition 870190046951, of 20/05/2019, p. 45/196
12/110 zen executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one hardware processor to perform a method of detecting change in the biological matter of a patient positioned within a device. low-field magnetic resonance imaging (MRI), the method comprising, while the patient remains positioned inside the low-field MRI device; acquire the first magnetic resonance imaging data of a part of the patient; acquire the second magnetic resonance imaging data from the patient, subsequent to the acquisition of the first magnetic resonance imaging data; align the first magnetic resonance imaging data and the second magnetic resonance imaging data; and comparing the first magnetic resonance imaging data and the second magnetic resonance imaging data aligned to detect at least one change in biological matter on the part of the patient.
[0017] Some modalities are directed to a system, comprising at least one computer hardware processor; and at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform a change detection method on the biological matter of a patient positioned within a low field magnetic resonance imaging (MRI) device, the method comprising, while the patient remains within the low field magnetic resonance imaging (MRI) device, acquire the first data
Petition 870190046951, of 20/05/2019, p. 46/196
13/110 magnetic resonance imaging of part of the patient; acquire second magnetic resonance imaging data from the patient, subsequent to the acquisition of the first magnetic resonance imaging data; align the first magnetic resonance imaging data and second magnetic resonance imaging data; and comparing the first magnetic resonance imaging data and the second magnetic resonance imaging data aligned to detect at least one change in biological matter on the part of the patient.
Brief Description of the Drawings [0018] Various aspects and modalities of the technology described will be referred to by reference to the attached figures. It must be appreciated that the figures are not necessarily to scale.
[0019] Figure 1 is a schematic illustration of a low-field MRI system, according to some types of technology described here;
[0020] Figures 2A and 2B illustrate bi-planar magnet configurations for a Bo magnet, according to some modalities of the technology described here;
[0021] Figures 2C and 2D illustrate a bi-planar electromagnet configuration for a Bo magnet, according to some modalities of the technology described here;
[0022] Figures 2E and 2F illustrate bi-planar permanent magnet configurations for a Bo magnet, according to some modalities of the technology described here;
[0023] Figures 3A and 3B illustrate a transportable low-field MRI system suitable for use with change detection techniques described here, in accordance with some modalities of technology described here;
Petition 870190046951, of 20/05/2019, p. 47/196
14/110 [0024] Figures 3C and 3D illustrate views of a portable MRI system, according to some types of technology described above;
[0025] Figure 3E illustrates a portable MRI system performing a head scan, according to some types of technology described here;
[0026] Figure 3F illustrates a portable MRI system performing a knee scan, according to some modalities of the technology described here;
[0027] Figure 3G illustrates another example of the portable MRI system, according to some modalities of technology described here;
[0028] Figure 4 illustrates a method for carrying out change detection, according to some types of technology described here;
[0029] Figure 5 illustrates a method of modifying acquisition parameters based on change detection information, according to some technology modalities described here;
[0030] Figure 6 illustrates a method of joint recording of MR imaging data, according to some types of technology described here;
[0031] Figure 7A illustrates a midline deviation measurement, according to some modalities of technology described here;
[0032] Figure 7B illustrates another measurement of midline deviation, according to some modalities of technology described here;
[0033] Figure 8 illustrates a method of determining a degree of change in a patient's midline deviation, according to some modalities of technology described here;
[0034] Figures 9A to 9C illustrate convolutional neural network architectures for performing midline deviation measurements, according to some types of technology described here;
Petition 870190046951, of 20/05/2019, p. 48/196
15/110 [0035] Figure 10 illustrates fully convolutional neural network architectures for performing midline deviation measurements, according to some types of technology described here; [0036] Figures 11A to 11E illustrate the measurements that can be used to determine the size of a patient's hemorrhage, according to some modalities of technology described here;
[0037] Figures 12A to C illustrate measurements that can be used to determine a change in the size of a patient's hemorrhage, according to some modalities of technology, described here;
[0038] Figure 13 illustrates a method of determining a degree of change in the size of an anomaly (eg, hemorrhage) in a patient's brain, according to some modalities of technology described here;
[0039] Figure 14 illustrates a fully convolutional neural network architecture to perform measurements that can be used to determine the size of an anomaly (eg, hemorrhage) in a patient's brain, according to some technology modalities described here .
[0040] Figure 15 illustrates a convolutional neural network architecture to perform measurements that can be used to determine the size of an anomaly (for example, hemorrhage) in a patient's brain, according to some modalities of technology described here;
[0041] Figure 16 is a diagram of an illustrative computer system in which the modalities described here can be implemented.
Detailed Description [0042] The MRI scanner market is dominated assusively
Petition 870190046951, of 20/05/2019, p. 49/196
16/110 by high field systems, and particularly for medical or clinical MRI applications. As discussed above, the general trend in medical imaging has been to produce MRI scanners with increasing field strengths, with the vast majority of clinical MRI scanners operating at 1.5T or 3T, with higher field strengths of 7T and 9T used in research facilities. As used here, high field ”generally refers to MRI systems currently in use in a clinical setting and, more particularly, to MRI systems operating with a major magnetic field (i.e., a Bo field) at or above 1, 5T, although clinical systems operating between .5T and 1.5T are often also characterized as high field. Field strengths between approximately .2T and .5T have been used as an intermediate field and, as field strengths in the high field regime continue to increase, field strengths in the range between .5T and 1T have also been characterized as an intermediate field . By contrast, low field generally refers to MRI systems operating with a Bo field or less than or equal to approximately 0.2T, although systems having a Bo field of between .2T and approximately, 3T have sometimes been characterized as low field as a result of increased field intensities at the high end of the high field regime. Within the low field regime, low field MRI systems operating with a Bo field of less than 1T are referred to here as very low field and low field MRI systems operating with a Bo field of less than 10mT are referred to here as ultra-low field.
[0043] As discussed above, conventional MRI systems require specialized facilities. An electromagnetically protected room is necessary for the MRI system to operate and the room space must be structurally reinforced. Additional rooms of
Petition 870190046951, of 20/05/2019, p. 50/196
17/110 will be supplied for the high energy electronic parts and control area for sweeping technicians. Secure access to the site must also be provided. In addition, a dedicated three-phase electrical connection must be installed to supply power to the electronic parts, which in turn are cooled by an ice water supply. Additional HVAC capacity must typically be provided as well. These site requirements are not only costly, but significantly limit the locations in which MRI systems can be employed. Conventional clinical MRI scanners also require substantial expertise to be operated and maintained. These highly trained technicians and engineers increase large and permanent operating costs for operating the MRI system. Conventional MRI, therefore, is often cost-prohibitive and severely limited in terms of accessibility, preventing MRI from being a widely available diagnostic tool capable of providing a wide range of clinical imaging solutions whenever and wherever needed. Typically, the patient should visit one of the various facilities at a location and date scheduled in advance, preventing MRI from being used in various medical applications for which it is uniquely efficient in assisting with diagnostics, surgery, patient monitoring, and the like.
[0044] As discussed above, high field MRI systems require facilities specially adapted to accommodate the size, weight and energy consumption and protection requirements of these systems. For example, a 1.5T MRI system typically weighs 4 to 10 tons and a 3T MRI system typically weighs between 8 and 20 tons. In addition, high-field MRI systems generally require significant amounts of heavy and expensive protection. Many mid-field scanners are even heavier,
Petition 870190046951, of 20/05/2019, p. 51/196
18/110 weighing between 10 and 20 tons, due in part to the use of permanent magnets and / or very large forks. Commercially available low-field MRI systems (for example, operating with a 0.2T Bo magnetic field) are also typically in the range of 10 tons or more due to the large amount of ferromagnetic material used to generate the Bo field, with additional tons of protection. To accommodate this heavy equipment, rooms (which typically have a minimum size of 30 to 50 square meters) need to be built with reinforced flooring (for example, concrete flooring), and must be specially protected to prevent electromagnetic radiation from interfering with the operation of the MRI system. As such, the available clinical MRI systems are immovable and require a significant cost of large dedicated space within a hospital or facility and in addition to the considerable costs of preparing the space for operation, they require additional permanent costs with specialized operation and maintenance of the system.
[0045] Additionally, currently available MRI systems typically consume large amounts of energy. For example, 1.5T and 3T MRI systems typically consume between 20 and 40 kW of energy during their operation, while 0.5T and 0.2T MRI systems typically consume between 5 and 20 kW, each using dedicated and specialized energy sources. Unless otherwise specified, energy consumption is referred to as the average energy consumed during an interval of interest. For example, the 20 to 40 kW referred to above indicate the average energy consumed by conventional MRI systems during the course of image acquisition, which can include relatively short periods of peak energy consumption that significantly exceed average energy consumption ( for example, when coils of grain
Petition 870190046951, of 20/05/2019, p. 52/196
19/110 customer and / or RF coils are pulsed during relatively short pulse sequence periods). Peak (or large) energy consumption ranges are typically resolved through energy storage elements (eg, capacitors) of the MRI system itself. Thus, the average energy consumption is the most relevant number since it generally determines the type of energy connection needed to operate the device. As discussed above, available clinical MRI systems must have dedicated power sources, typically requiring a dedicated three-phase connection to the facility to power the components of the MRI system. Additional electronic parts are then necessary to convert the three-phase energy into single-phase energy used by the MRI system. The many physical development requirements of conventional clinical MRI systems create a significant availability problem and severely restrict the clinical applications for which MRI can be used.
[0046] Accordingly, the many requirements for high-field MRIs make facilities prohibitive in many situations, limiting their use to large institutional hospitals or specialized facilities and generally restricting their use to highly scheduled appointments, requiring the patient to visit dedicated facilities at previously scheduled times. Thus, many restrictions on high-field MRI prevent MRI from being fully used as an imaging modality. Despite the disadvantages of the high field MRI mentioned above, the call for a significant increase in SNR in higher fields continues to steer the industry towards increasing field strengths for use in clinical and medical MRI applications, thereby increasing the cost and complexity of MRI scanners, and further limiting their availability and preventing their use as a
Petition 870190046951, of 20/05/2019, p. 53/196
20/110 general purpose image and / or generally available.
[0047] The inventors have developed techniques to produce low-field, improved quality, portable and / or cheaper MRI systems that can improve the large-scale use of MRI technology in a variety of environments in addition to large MRI facilities in hospitals and research facilities. The inventors appreciated that the accessibility and availability of such low-field MRI systems (for example, due to the relatively low cost, portability, etc.) allow imaging applications not available or not practicable with other imaging modalities. For example, generally transportable low-field MRI systems can be brought to a patient to facilitate monitoring the patient over an extended period of time by acquiring a series of images and detecting changes occurring during the period of time. Such a monitoring procedure is not realistic with high field MRI. In particular, as discussed above, high-field MRI facilities are generally located in special facilities and require advanced programming at a significant cost. Many patients (for example, an unconscious Neurological CTI patient) cannot be taken to an available facility, and even if a high-field MRI facility could be made available, the cost of an extended MRI scan over the course of several hours it would be prohibitively high.
[0048] Additionally, while CT scanners are generally more available and accessible than high field MRI systems, these systems may not yet be available for very long-term monitoring applications to detect or monitor changes in the patient over a period of time extended. In addition, an extended CT examination subjects the patient to a dose
Petition 870190046951, of 20/05/2019, p. 54/196
21/110 of X-ray radiation, which may be unacceptable in many, if not all, circumstances. Finally, CT is limited in its ability to differentiate soft tissue and may be unable to detect the type of change that may be of interest to the physician. The inventors recognize that low-field MRI installations perform monitoring tasks in circumstances where current imaging modalities cannot.
[0049] The inventors recognize that the transport capacity, accessibility and availability of low field MRI systems allow monitoring applications that are not available using the existing imaging modalities. For example, low-field MRI systems can be used to image continuously and / or regulate a part of the anatomy of interest to detect changes that occur in it. For example, in the neurological intensive care unit (NICU), patients are usually under general anesthesia for a significant amount of time, while the patient is being evaluated or during a procedure. Due to the need for a specialized facility, conventional clinical MRI systems are not available for these and many other circumstances. In addition, doctors may have limited access to only one CT scan device for a patient (for example, once a day). In addition, even when such systems are available, it is inconvenient, and sometimes impossible, to generate images of patients who are, for example, unconscious or, otherwise, cannot be transported to the MRI facility. Thus, conventional MRI is not typically used as a monitoring tool.
[0050] The inventors recognize that low field MRI can be used to monitor a patient by acquiring data from
Petition 870190046951, of 20/05/2019, p. 55/196
22/110 magnetic resonance imaging (MR) over a period of time and detect changes that occur. For example, a transportable low-field MRI system can be brought to a patient that can be positioned within the system, while a sequence of images of the patient's brain is obtained. The acquired images can be aligned and differences between the images can be detected to monitor any changes that occur. Image acquisition can be performed substantially continuously (for example, with an acquisition made immediately after another), on a regular basis (for example, with prescribed breaks between acquisitions), or periodically, according to an acquisition schedule determined. As a result, a measurement can obtain temporal information regarding the phlology of interest. For example, the techniques described here can be used to monitor the patient's brain for changes in the degree of deviation from the midline in the brain. As another example, the techniques described here can be used to monitor a patient's brain to detect the change in size of an anomaly (for example, a hemorrhage in the brain).
[0051] Accordingly, the inventors have developed low-field MRI techniques for monitoring the patient's brain for changes related to brain injury, abnormalities, etc. For example, the low-field MRI techniques described here can be used to determine whether there is a change in a degree of midline deviation for a patient. Midline deviation refers to an amount of displacement of the midline of the brain from its normal symmetrical position due to trauma (for example, stroke, hemorrhage or other injury) and is an important indicator for doctors about the seriousness of the brain trauma.
[0052] In some modalities, monitoring techniques
Petition 870190046951, of 20/05/2019, p. 56/196
23/110 low-field MRI can be combined with machine learning techniques to continuously monitor the amount of midline shift in a patient (if any) and detect a change in the degree of midline shift over time. In such modalities, low-field MRI monitoring allows a sequence of images of a patient's brain to be obtained and learning techniques (for example, deep learning techniques, such as convolutional neural networks) can be used to determine, the from the image sequence, a corresponding sequence of midline brain locations and / or a corresponding sequence of midline displacements from their normal position. For example, in some modalities, deep learning techniques can be used to identify locations of the points where the falx cerebri is attached to the internal board of the patient's skull and a location of a measurement point in the septum pellucidum. These locations can, in turn, be used to obtain a midline deviation measurement.
[0053] It should be appreciated, however, that although in some modalities the central line is detected by the detection locations of the fixation points of the falx cerebri, there are other ways to detect the midline. For example, in some modalities, the midline can be detected by segmenting the left and right parts of the brain and upper and lower parts of the brain (as defined by the measurement plane).
[0054] In some modalities, monitoring of midline deviation involves, while the patient remains positioned inside a low field MRI device: (1) acquiring the first magnetic resonance imaging (MR) data for a part of the patient's brain; (2) providing the first MR data as input to a trained statistical classifier (for example,
Petition 870190046951, of 20/05/2019, p. 57/196
24/110 a convolutional neural network) to obtain the first corresponding output; (3) identify, from the first exit, at least an initial site of at least one landmark associated with at least one midline structure of the patient's brain; (4) acquire the second MR image data for the subsequent patient's brain part (for example, in an hour) to acquire the first MR image data; (5) provide the second MR imaging data as input to the trained statistical classifier to obtain the corresponding second output; (6) identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain and (7) determine a degree of change in the midline deviation using the at least one home location of at least one landmark and at least one updated location of at least one landmark.
[0055] In some embodiments, the at least one landmark associated with at least one midline structure of the patient's brain may include an anterior fixation point of the falx cerebri (to the inner board of the patient's skull), a posterior fixation point of falx cerebri, a point in the septum pellucidum. In other modalities, at least one landmark can indicate segmentation results for the left and right sides of the brain and / or the upper and lower parts of the brain.
[0056] In some modalities, identifying, from the first exit of the trained statistical classifier, at least one initial location of at least one landmark associated with the peto minus a midline structure of the patient's brain includes: (1) identifying a initial location of an anterior fixation point of the falx cerebri; (2) identify an initial location of a posterior fixation point for the falx cerebri; and (3) identify an initial location for a
Petition 870190046951, of 20/05/2019, p. 58/196
25/110 measurement on a septum pellucidum. Identifying, from the second output of the trained statistical classifier, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain, includes: (1) identifying an updated location of the fixation point anterior of the cerebri falx; (2) identify an updated location of the posterior fixation point of the falx cerebri; and (3) identify an updated location of the measurement point in the septum pellucidum. In turn, the degree of change in the midline deviation can be performed using the initial and updated locations identified at the anterior fixation point of the falx cerebri, at the posterior fixation point of the falx cerebri, and at the measurement point in the septum pellucidum.
[0057] In some embodiments, determining the degree of change in midline deviation comprises determining an initial amount of midline deviation using the initial identified locations of the anterior fixation point of the falx cerebri, the posterior fixation point of the falx cerebri, and the measurement point in the septum pellucidum; determine an updated amount of midline deviation using the updated locations identified from the anterior fixation point of the falx cerebri, the posterior fixation point of the falx cerebri, and the measurement point in the septum pellucidum; and determining the degree of change in the midline deviation using the determined initial and updated quantities of the midline deviation.
[0058] In some modalities, the trained statistical classifier can be a neural network of multiple layers. For example, the multilayer neural network can be a convolutional neural network (for example, one having convolutional layers, set layers and a fully connected layer) or a fully convolutional neural network (for example, a convolutional neural network without a layer fully connected). As another example
Petition 870190046951, of 20/05/2019, p. 59/196
26/110 pio, the multilayered neural network may include a convolutional neural network and a recurrent neural network (for example, short term, long memory).
[0059] The inventors have also developed low-field MRI techniques to determine whether there is a change in the size of an anomaly (for example, bleeding, injury, edema, stroke nucleus, stroke penumbra and / or swelling) in the brain of a patient. In some embodiments, low field MRI monitoring techniques can be combined with machine learning techniques to continuously monitor the size of the anomaly and detect a change in its size over time. In such modalities, low-field MRI monitoring allows a sequence of images of a patient's brain to be obtained and machine learning techniques (for example, deep learning techniques, such as convolutional neural networks) can be used to determine , from the image sequence, a corresponding sequence of anomaly sizes. For example, the deep learning techniques developed by the inventors can be used to segment the anomaly in MRI images, identify points that specify main geometric axes of a 2D or 3D boundary region (for example, box), identify the maximum diameter of the anomaly and a maximum orthogonal diameter of the anomaly that is orthogonal to the maximum diameter and / or perform any other processing in addition to identifying the size of the anomaly.
[0060] Accordingly, in some modalities, monitoring the size of the anomaly involves, while a patient is positioned inside a low field MRI device: (1) acquiring the first generation data by magnetic resonance (MR) for a part of the patient's brain; (2) provide the first MR imaging data as input to a statistical classifier
Petition 870190046951, of 20/05/2019, p. 60/196
Trained 27/110 (for example, multilayer neural network, a convolutional neural network, a fully convolutional neural network) to obtain the first corresponding output; (3) identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain; (4) acquire the second MR imaging data for the patient's brain part, subsequent to the acquisition of the first MR imaging data; (5) provide the second MR imaging data as input to the trained statistical classifier to obtain the corresponding second output; (5) identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain; (6) determine the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[0061] In some embodiments, the at least one initial value of at least one characteristic indicative of the size of the anomaly may include multiple values specifying a region surrounding the anomaly (for example, values specifying a boundary region, values specifying the perimeter of the anomaly , etc.). In some embodiments, the at least one starting value of the at least one characteristic may include values that specify one or more diameters of the anomaly (for example, diameter 1102 and diameter 1104 orthogonal to diameter 1102, as shown in figure 11A).
[0062] In some modalities, determining the change in the size of the anomaly involves: (1) determining an initial size of the anomaly using at least one value of at least one characteristic; (2) determine an updated anomaly size using at least one updated value of at least one characteristic; and
Petition 870190046951, of 20/05/2019, p. 61/196
28/110 (3) determine the change in the size of the anomaly using the initial and updated determined sizes of the anomaly.
[0063] Below are more detailed descriptions of the various concepts related to, and modalities of the methods and apparatus to perform the monitoring using low-field magnetic resonance applications including low-field MRI. It should be appreciated that the various aspects described here can be implemented in any one of several ways. Examples of specific implementations are provided here for illustrative purposes only. In addition, the various aspects described in the modalities below can be used alone or in any combination, and are not limited to the combinations explicitly described here.
[0064] Figure 1 is a block diagram of illustrative components of an MRI 100 system. In the illustrative example of Figure 1, the MRI 100 system comprises workstation 104, controller 106, pulse sequence store 108, the energy management system 110, and the magnetic components 120. It should be appreciated that the system 100 is illustrative and that an MRI system may have one or more other components of any suitable type in addition to or in place of the components illustrated in figure 1.
[0065] As illustrated in Figure 1, the magnetic components 120 comprise the Bo 122 magnet, shim coils 124, RF transmission and reception coils 126, and gradient coils 128. The Bo 122 magnet can be used to generate, at least less in part, the main magnetic field Bo. The Bo 122 magnet can be any suitable type of magnet that can generate a main magnetic field (for example, a low field strength of approximately 0.2T or less), and can include one or more Bo coils, correction coils, etc. . Shim coils 124 can be used to contribute fields
Petition 870190046951, of 20/05/2019, p. 62/196
Magnetic 29/110 to optimize the homogeneity of the Bo field generated by magnet 122. Gradient spirals 128 can be arranged to provide gradient fields and, for example, can be arranged to generate gradients in the magnetic field in three substantially orthogonal directions (X , Y, Z) to find where MR signals are induced.
[0066] RF transmission and reception coils 126 may comprise one or more transmission coils that can be used to generate RF pulses to induce a Bi magnetic field. The transmit / receive coils can be configured to generate any suitable type of RF pulses configured to excite an MR response in a subject and detect the resulting MR signals emitted. RF 126 transmit and receive coils may include one or more transmit coils and one or more receive coils. The configuration of the transmit / receive coils varies with implementation and may include a single coil for both transmitting and receiving, separate coils for transmitting and receiving, multiple coils for transmitting and / or receiving, or any combination to achieve channel MRI systems single or parallel. In this way, the magnetic transmitting / receiving component is often referred to as Tx / Rx or Tx / Rx coils to refer, generically, to various configurations for the transmitting and receiving component of an MRI system. Each of the magnetic components 120 can be constructed in any suitable form. For example, in some embodiments, one or more of the magnetic components 120 may be manufactured using the lamination techniques described in the jointly filed orders, incorporated above.
[0067] The power management system 110 includes electronic parts to provide operational power to one or more with
Petition 870190046951, of 20/05/2019, p. 63/196
30/110 components of the low field MRI system 100. For example, the power management system 110 may include one or more power supplies, gradient power amplifiers, transmission coil amplifiers, and / or any electronic parts adequate energies required to provide adequate operational power to power and operate the components of the low field 100 MRI system.
[0068] As illustrated in figure 1, the power management system 110 comprises a power supply 112, amplifiers 114, transmit / receive switch 116, and thermal management components 118. Power supply 112 includes electronic parts to provide operating power for the magnetic components 120 of the low-field MRI system 100. For example, power supply 112 may include electronic parts to provide operating power for one or more Bo coils (for example, Bo 122 magnet) to produce the field main magnetic field for the low field MRI system. In some embodiments, the 112 power supply may be a unipolar continuous wave (CW) power supply, however, any suitable power supply may be used. The transmit / receive switch 116 can be used to select whether the RF transmit coils or the RF receive coils are being operated.
[0069] Amplifiers 114 may include one or more RF reception (Rx) preamplifiers that amplify the MR signals detected by one or more RF receiving coils (e.g., coils 124), one or more RF transmission amplifiers (Tx) configured to supply power to one or more RF transmission coils (for example, coils 126), one or more gradient energy amplifiers configured to supply power to one or more gradient coils (for example, coils of high
Petition 870190046951, of 20/05/2019, p. 64/196
31/110 diente 128), shim amplifiers configured to supply power to one or more shim coils (for example, shim coils 124).
[0070] Thermal management components 118 provide the cooling for low field MRI system components 100 and can be configured to do this by facilitating the transfer of thermal energy generated by one or more components of the field MRI system down 100 away from those components. Thermal management components 118 may include, without limitation, components for performing water or air-based cooling, which may be integrated with or disposed of near heat generating MRI components including, but not limited to Bo coils, air coils gradient, shim coils and / or transmission and reception coils. Thermal management components 118 may include any suitable heat transfer means including, but not limited to, air and water, to transfer heat away from the components of the low field MRI system 100.
[0071] As illustrated in figure 1, the low-field MRI system 100 includes controller 106 (also referred to as a console), having electronic control parts for sending instructions to and receiving information from the power management system 110. Controller 106 can be configured to implement one or more pulse sequences, which are used to determine the instructions sent to the power management system 110 to operate the magnetic components 120 in a desired sequence. For example, controller 106 can be configured to control power management system 110 to operate magnetic components 120, according to a steady-state precession pulse sequence (bSSFP), a gradient echo pulse sequence low field, an echo pulse sequence
Petition 870190046951, of 20/05/2019, p. 65/196
32/110 low-field rotary, a low-field inversion recovery pulse sequence, arterial rotary labeling, diffusion-weighted imaging (DWI), and / or any other suitable pulse sequence. Controller 106 can be implemented as hardware, software or any suitable combination of hardware and software, since aspects of the description provided here are not limited in this regard.
[0072] In some embodiments, controller 106 may be configured to implement a pulse sequence by obtaining information about the pulse sequence from the pulse sequence deposit 108, which stores information for each of one or more sequence of pulses. wrists. The information stored by the pulse sequence deposit 108 for a particular pulse sequence can be any suitable information that allows controller 106 to implement the particular pulse sequence. For example, the information stored in the pulse sequence deposit 108 for a pulse sequence may include one or more parameters for operating the magnetic components 120, according to a pulse sequence (for example, parameters for operating the transmission coils and RF reception 126, parameters to operate the gradient coils 128, etc.), one or more parameters to operate the energy management systems 110 according to the pulse sequence, one or more programs comprising instructions that, when executed at all controller 106, cause controller 106 to control system 100 to operate according to the pulse sequence, and / or any other suitable information. The information stored in the pulse sequence store 108 can be stored in one or more non-transitory storage media.
[0073] As illustrated in figure 1, controller 106 is also integrated
Petition 870190046951, of 20/05/2019, p. 66/196
33/110 rage with the computing device 104 programmed to process the received MR data. For example, computing device 104 can process received MR data to generate one or more MR images using any suitable image reconstruction process. Controller 106 can provide information about one or more pulse sequences to computing device 104 for processing data by the computing device. For example, controller 106 can provide information about one or more pulse sequences to computing device 104 and the computing device can perform an image reconstruction process based, at least in part, on the information provided.
[0074] The computing device 104 can be any electronic device that can process the acquired MR data and generate one or more images of the subject. In some embodiments, computing device 104 can be a fixed electronic device, such as a desktop computer, a server, a rack-mounted computer, or any other suitable fixed electronic device that can be configured to process MR data and generate a or more images of the subject. Alternatively, computing device 104 can be a portable device, such as a smartphone, personal digital assistant, laptop computer, tablet computer, or any other portable device that can be configured to process MR data and generate one or more images of the subject. In some embodiments, the computing device 104 may comprise various computing devices of any suitable type, since aspects are not limited in this regard. A user 102 can interact with workstation 104 to control aspects of the low field MR system 100 (for example, programming system 100 to operate according to a sequence
Petition 870190046951, of 20/05/2019, p. 67/196
34/110 of pulses in particular, adjust the one or more parameters of the system 100, etc.) and / or view images obtained by the low field MR system 100. According to some modalities, the computing device 104 and the controller 106 form a single controller, while in other embodiments, computing device 104 and controller 106 each comprise one or more controllers. It should be appreciated that the functionality realized by the computing device 104 and the controller 106 can be distributed in any way through any combination of one or more controllers, since the aspects are not limited to use with any particular implementation or architecture.
[0075] Figures 2A and 2B illustrate bi-planar magnetic configurations that can be used in a low field MRI system suitable for use with the change detection techniques described here. Figure 2A schematically illustrates a biplanar magnet configured to produce, at least in part, a part of a Bo field suitable for low field MRI. The bi-planar magnet 200 comprises two outer coils 210a and 210b and two inner coils 212a and 212b. When adequate current is applied to the coils, a magnetic field is generated in the direction indicated by the arrow to produce a Bo field, presenting a field of view between the coils that, when designed and properly constructed, can be suitable for low field MRI. The term coil is used here to refer to any conductor or combination of conductors of any geometry having at least one loop that conducts the current to produce a magnetic field, thus forming an electromagnet.
[0076] It should be appreciated that the biplanar geometry illustrated in Figure 2A is generally unsuitable for high field MRI, due to the difficulty in obtaining a Bo field of sufficient homogeneity
Petition 870190046951, of 20/05/2019, p. 68/196
35/110 with high field intensities. High field MRI systems typically use solenoid geometries (and superconducting wires) to achieve high field intensities of sufficient homogeneity for high field MRI. The Bo bi planar magnet illustrated in figure 2A provides a generally open geometry, facilitating its use with patients who suffer from claustrophobia and may refuse to be examined with conventional high field solenoid coil geometries. In addition, the bi-planar design can facilitate use with larger patients due to its open design and, in some cases, a generally larger field of view possible with requirements for homogeneity and low field intensities. In addition, the generally open design facilitates access to the patient being examined and can improve the ability to position a patient within the field of view, for example, an unconscious, sedated or anesthetized patient, as discussed in more detail below. The biplanar geometry in figure 2A is merely illustrative, since more or less coils can be arranged as necessary, since the aspects are not limited in this regard.
[0077] Figure 2B illustrates a hybrid biplanar magnet using lamination techniques to manufacture a Bo magnet or part thereof and / or to manufacture one or more other magnetic components for use in low field MRI. For example, in the illustrative bi-planar magnet 200 'shown in Figure 2B, laminate panels 220a and 220b replace inner coils 212a and 212b to produce a hybrid magnet. Laminated panels 220a and 220b can include any number of laminated layers having one or more Bo coils, gradient coils, correction coils and / or shim coils, etc. manufactured in them. or parts thereof to facilitate the production of magnetic fields used in low-field MRI. Suitable hybrid magnets using lamination techniques are described in U.S. patent application No.
Petition 870190046951, of 20/05/2019, p. 69/196
36/110
14 / 845,652 (order '652), filed on September 4, 2015 and entitled Low Field Magnetic Resonance Imaging Methods and Apparatus, which is incorporated herein by reference in its entirety. In other embodiments, lamination techniques can be used to implement the Bo magnet in its entirety (for example, replacing coils 210a and 210b b).
[0078] Illustrative laminated panels 220a and 220b may, additionally or alternatively, have manufactured in them, one or more gradient coils, or parts thereof, to encode the spatial location of the received MR signals, as a function of frequency or phase . According to some embodiments, a laminated panel comprises at least one standardized conductive layer to form one or more gradient coils, or a part of one or more gradient coils, capable of producing or contributing to the appropriate magnetic fields to provide the encoding of the MR signals detected when operated on a low field MRI system. For example, laminate panel 220a and / or laminate panel 220b may comprise a first gradient coil configured to selectively vary the Bo field in a first direction (X) to perform frequency coding in that direction, a second coil of gradient configured to selectively vary the Bo field in a second direction (Y), substantially orthogonal to the first direction, to perform phase encoding, and / or a third gradient coil configured to selectively vary the Bo field, in a third direction (Z), substantially orthogonal to the first and second directions, to allow slice selection for volumetric imaging applications.
[0079] Illustrative laminated panels 220a and 220b may, additionally or alternatively, include additional magnetic components such as one or more shim coils arranged to generate
Petition 870190046951, of 20/05/2019, p. 70/196
37/110 magnetic fields in support of the system to, for example, increase the resistance and / or improve the homogeneity of the Bo field, react to harmful effects to the field, such as those created by the operation of gradient coils, effects of object loading being examined for, or otherwise supporting, the magnetic parts of the low field MRI system. The bi-planar magnet illustrated in figures 2A and 2B, can be produced using conventional coils, lamination techniques or a combination of both, and can be used to supply magnetic components for a low-field MRI system, adapted to perform the change detection techniques, as discussed in more detail below.
[0080] The inventors recognized that the low field context allows for Bo magnet designs that would not be viable in the high field regime. For example, due, at least in part, to lower field resistances, superconducting material and corresponding cryogenic cooling systems can be eliminated. Due in part to low field resistances, Bo electromagnets, constructed using non-superconducting material (for example, copper) can be used in the low field regime. However, such electromagnets can still consume relatively large amounts of energy during operation. For example, operating an electromagnet using a copper conductor to generate a magnetic field of .2T or more requires a dedicated or specialized power connection (for example, a dedicated three-phase power connection). The inventors have developed MRI systems that can be operated using power from the supply center (ie, standard plug power), allowing the MRI system to be powered at any location that has a common power connection, such as an electrical outlet. standard wall (for example, 120V / 20A connection in the United States) or plugs for it
Petition 870190046951, of 20/05/2019, p. 71/196
38/110 large common appliances (eg 220-240V / 30A). In this way, a low-energy MRI system facilitates portability and availability, allowing an MRI system to be operated in places where it is needed (for example, the MRI system can be brought to the patient, instead of the other way around) ), examples of which are discussed in more detail below. In addition, operation from standard electrical power eliminates the electronic parts conventionally required to convert three-phase energy into single-phase energy and to soften the energy supplied directly from the installation board. In fact, the outlet power can be converted directly to DC and distributed to energize the components of the MRI system.
[0081] Figures 2C and 2D illustrate a Bo magnet formed using an electromagnet and a ferromagnetic fork. In particular, the Bo 2000 magnet is partly formed by a 2010 electromagnet arranged in bi-planar geometry, comprising electromagnetic coils 2012a and 2012b on an upper side and electromagnetic coils 2014a and 2014b on an inner side of the Bo 2000 magnet. according to some modalities, the coils that form the electromagnet 2010 can be formed from several turns of a copper wire or copper tape, or any other conductive material suitable to produce a magnetic field when operated (for example, when the current is driven by conductor windings). While the illustrative electromagnet, illustrated in figures 2C and 2D, comprises two pairs of coils, an electromagnet can be formed using any number of coils in any configuration, since the aspects are not limited in this regard. The electromagnetic coils that make up the 2010 electromagnet can be formed, for example, by wrapping a conductor 2013 (for example, copper tape, wire, paint, etc.) around a 2017 fiberglass ring. For example, the con
Petition 870190046951, of 20/05/2019, p. 72/196
39/110 dutor 2013 can be a suitable insulated copper wire, or alternatively the conductor 2013 can be a copper tape wrapped together with an insulating layer (eg a Mylar layer) to electrically insulate multiple coil windings . A 2019 connector can be provided to allow a power connection to supply current to operate the 2014a and 2014b coils in series. A similar connector on the upper side of the electromagnet (not visible in figures 2C and 2D) can be provided to operate the coils 2012a and 2012b.
[0082] It should be appreciated that electromagnetic coils can be formed from any suitable material and dimensioned in any suitable way in order to produce or contribute to a desired Bo magnetic field, since aspects are not limited to use with any particular type of electromagnet. As a non-limiting example that may be suitable to partly form an electromagnet (eg electromagnet 2010), an electromagnet coil can be constructed using the copper tape to the mylar insulator having 155 turns to form a diameter internal diameter of approximately 23 to 27 inches (for example, approximately 25 inches), an external diameter of approximately 30 to 35 inches (for example, 32 inches). However, different materials and / or dimensions can be used to build an electromagnetic coil presenting the desired characteristics, since the aspects are not limited in this regard. The upper and lower coils can be positioned to provide a distance of approximately 10 to 15 inches (for example, approximately 12.5 inches) between the lower coil on the upper side and the upper coil on the lower side. It should be appreciated that the dimensions will differ depending on the desired characteristics including, for example, field strength, field of view, etc.
Petition 870190046951, of 20/05/2019, p. 73/196
40/110 [0083] In the illustrative Bo magnet, illustrated in figures 2C and 2D, each pair of coils 2012 and 2014 is separated by thermal management components 2030a and 2030b, respectively, to transfer heat produced by the electromagnetic coils and gradient coils ( not shown in figures 2C and 2D) away from the magnets to provide thermal management for the MRI device. In particular, the thermal management components 2030a and 2030b may comprise a cooling plate having ducts that allow refrigerant to be circulated through the cooling plate to transfer heat away from the magnets. The cooling plate 2030a, 2030b can be constructed to reduce or eliminate swirling currents induced by the operation of gradient coils that can produce electromagnetic fields that interrupt the Bo magnetic field produced by the Bo 2000 magnet. For example, the thermal management components 2030a and 2030b may be the same or similar to any thermal management components described in US patent application No. 14 / 846,042, entitled Thermal Management Methods and Apparatus, filed on September 4, 2015, which is incorporated herein by reference in its entirety. According to some modalities, thermal management components can be eliminated, as discussed in more detail below.
[0084] The Bo 2000 magnet additionally comprises a fork 2020 that is magnetically coupled to the electromagnet to capture the magnetic flux that, in the absence of the fork 2020, would be lost and would not contribute to the flow density in the region of interest between the upper electromagnetic coils and lower. In particular, the fork 2020 forms a magnetic circuit connecting the coils on the upper and lower side of the electromagnet in order to increase the flow density in the region between the coils, thus increasing the intensity of the field within the imaging region ( also referred to as cam
Petition 870190046951, of 20/05/2019, p. 74/196
41/110 po of view) of the Bo magnet. The imaging region or field of view defines the volume at which the Bo magnetic field, produced by a given Bo magnet, is suitable for imaging. More particularly, the imaging region or field of view corresponds to the region for which the magnetic field Bo is sufficiently homogeneous at a desired field strength, so that detectable MR signals are emitted by an object positioned there in response the application of radio frequency excitation (for example, an appropriate radio frequency pulse sequence). Fork 2020 comprises fork 2022 and plates 2024a, 2024b, which can be formed using any suitable ferromagnetic material (e.g., iron, steel, etc.). The 2024a, 2024b plates collect the magnetic flux generated by the 2010 electromagnet coil pairs and direct it to the 2022 frame, which in turn returns the flow back to the opposite coil pair, thus increasing by up to one factor of two, the density of the magnetic flux in the imaging region between the coil pairs (for example, the coil pair 2012a, 2012b and the coil pair 2014a and 2014b) in the same proportion as the operating current supplied to the coils . In this way, the 2020 fork can be used to produce a larger Bo field (resulting in a larger SNR) without a corresponding increase in power requirements, or the 2020 fork can be used to reduce the energy requirements of the Bo 2000 magnet for a field Bo determined.
[0085] According to some modalities, the material used for fork parts 2020 (that is, frame 2022 and / or plates 2024a, 2024b) is steel, for example, a low-carbon steel, silicon steel, cobalt steel, etc. According to some modalities, the gradient coils (not shown in figures 2C, 2D) of the MRI system are arranged in proximity to the plates 2024a, 2024b induced
Petition 870190046951, of 20/05/2019, p. 75/196
42/110 using swirling chains for the plates. To mitigate, plates 2024a, 2024b and / or frame 2022 can be constructed of silicon steel, which is generally more resistant to the production of turbid chains than, for example, low carbon steel. It should be appreciated that the fork 2020 can be constructed using any ferromagnetic material with sufficient magnetic permeability and individual parts (for example, frame 2022 and plates 2024a, 2024b) can be constructed from the same ferromagnetic material or another, as seen that flow density increase techniques are not limited to use with any particular type of material or combination of materials. Additionally, it must be appreciated that the fork 2020 can be formed using different geometries and arrangements.
[0086] It should be appreciated that the 2020 fork can be made of any suitable material and can be sized to provide the desired magnetic flux capture while meeting other design constraints such as weight, cost, magnetic properties, etc. As an example, the fork frame (for example, frame 2022) can be formed from a low carbon steel of less than 0.2% carbon or silicon steel, with long beams having a length of approximately 38 inches, a width of approximately 8 inches, and a thickness (depth) of approximately 2 inches, and the short bundles having a length of approximately 19 inches, a width of approximately 8 inches and a thickness (depth of approximately 2 inches). The plates (for example, plates 2024a and 2024b) can be formed from a low carbon steel of less than 0.2% carbon or silicon steel and have a diameter of approximately 30 to 35 inches (for example , approximately 32 inches). However, the dimensions and materials
Petition 870190046951, of 20/05/2019, p. 76/196
43/110 provided above are merely illustrative of a suitable modality of a fork that can be used to capture magnetic flux generated by an electromagnet.
[0087] As an example of the improvement achieved through the use of the 2020 fork, operating the electromagnet 2010 to produce a Bo magnetic field of approximately 20mT without the 2020 fork consumes about 5 kW, while producing the same Bo magnetic field of 20mT with the 2020 fork consumes about 750W of energy. Operating the electromagnet 2010 with the 2020 fork, a Bo magnetic field of approximately 40mT can be produced using 2kW of energy and a Bo magnetic field of approximately 50mT can be produced using approximately 3kW of energy. In this way, energy requirements can be significantly reduced by using fork 220 which allows the operation of a Bo magnet without a dedicated three-phase power connection. For example, electric power from the supply center in the United States and most of North America, is supplied at 120V and 60Hz and is rated at 15 or 20 amps, allowing the use of devices operating below 1800 and 2400 W , respectively. Many installations also have 220-240 VAC outlets with 30 amp ratings, allowing devices operating up to 7200 W to be powered from such outlets. According to some modalities, a low-field MRI system using a Bo magnet comprising an electromagnet and a fork (for example, Bo 2000 magnet) is configured to be powered through a standard wall outlet, as discussed in more detail below. According to some modalities, a low-field MRI system using a Bo magnet comprising an electromagnet and a fork (for example, Bo 2000 magnet) is configured to be powered through a 220-240 VAC outlet, as also discussed in greater detail
Petition 870190046951, of 20/05/2019, p. 77/196
44/110 below.
[0088] Referring again to figures 2C and 2D, the illustrative Bo magnet 2010 additionally comprises shim rings 2040a, 2040b and shim discs 2042a, 2042b configured to increase the magnetic field Bo generated to improve homogeneity in the field of view (for example example, in the region between the upper and lower coils of the electromagnet, where the Bo field is adequate for sufficient production of MR signal), as best seen in figure 2D, where the lower coils were removed. In particular, the shim rings 2040 and the shim disk 2042 are sized and arranged to increase the uniformity of the magnetic field generated by the electromagnet at least within the field of view of the Bo magnet. In particular, the height, thickness and material of the shim rings 2040a, 2040b and the diameter, thickness and material of the shim discs 2042a, 2042b can be chosen in order to achieve a Bo field of adequate homogeneity. For example, the shim disc can be supplied with a diameter of approximately 5 to 6 inches and a width of approximately .3 to .4 inches. A shim ring can be formed from a plurality of circular arc segments (for example, 8 circular arc segments), each having a height of approximately 20 to 22 inches, and a width of approximately 2 inches to form a ring having an internal diameter of approximately about 21 to 22 inches and approximately between 23 and 24 inches.
[0089] The weight of the Bo magnet is a significant part of the overall weight of the MRI system, which in turn impacts the portability of the MRI system. In modalities that basically use low carbon steel and / or silicon steel for the fork and wedge components, an illustrative Bo magnet 2000, sized similarly to the one described above, can weigh approximately 550 kilograms. According
Petition 870190046951, of 20/05/2019, p. 78/196
45/110 some modalities, cobalt steel (CoFe) can be used as the raw material for the fork (and possibly the wedge components), potentially reducing the weight of the Bo 2000 magnet to approximately 450 kilograms. However, CoFe is generally more expensive than, for example, low carbon steel, increasing system costs. Accordingly, in some modalities, the selection components can be formed using CoFe to balance the exchange between cost and weight that arises from its use. Using such illustrative Bo magnets, a portable or otherwise transportable MRI system can be constructed, for example, by integrating the Bo magnet within a housing, frame or other body to which casters, wheels or other means of transportation can be built. be fixed to allow the MRI system to be transported to desired locations (for example, by manually pushing the MRI system and / or including motorized assistance). As a result, an MRI system can be brought to the place where it is needed, increasing its availability and use as a clinical tool and providing MRI applications that were previously not possible.
[0090] The main contributor to the overall energy consumption of a low field MRI system, using a Bo 2000 magnet, is the electromagnet (for example, the electromagnet 2010). For example, in some embodiments, the electromagnet can consume 80% or more of energy from the MRI system as a whole. In order to significantly reduce the energy requirements of the MRI system, the inventors developed Bo magnets that use permanent magnets to produce and / or contribute to the Bo electromagnetic field. According to some modalities, Bo electromagnets are replaced by permanent magnets as the main source of Bo electromagnetic field. A permanent magnet refers to any object or material that maintains its own
Petition 870190046951, of 20/05/2019, p. 79/196
46/110 persistent magnetic field once magnetized. Materials that can be magnetized to produce a permanent magnet are referred to here as ferromagnetic and include, as non-limiting examples, alloys of iron, nickel, cobalt, neodymium (NdFeB), samarium cobalt alloys (SmCo), alnico alloys (AINiCo ), strontium ferrite, barium ferrite, etc. The permanent magnet material (for example, magnetizable material that has been brought to saturation by a magnetizing field) retains its magnetic field when the drive field is removed. The amount of magnetization retained by a particular material is referred to as material remnant. Thus, once magnetized, a permanent magnet generates a magnetic field corresponding to its remnant, eliminating the need to have an energy source to produce the magnetic field. [0091] Figure 2E illustrates a permanent Bo magnet, according to some modalities. In particular, the Bo 2100 magnet is formed by permanent magnets 2110a and 2110b arranged in bi-planar geometry and a fork 2120 that captures the electromagnetic flux produced by the permanent magnets and transfers the flow to the opposite permanent magnet to increase the flow density between permanent magnets 2110a and 2110b. Each of the permanent magnets 2110a and 2110b is formed from a plurality of concentric permanent magnets. In particular, as visible in figure 2E, permanent magnet 2110b comprises an outer ring of permanent magnets 2114a, an intermediate ring of permanent magnets 2114b, an inner ring of permanent magnets 2114c, and a permanent magnet disk 2114d in the center. The permanent magnet 2110a can comprise the same set of permanent magnetic elements as the permanent magnet 2110b.
[0092] The permanent magnet material used can be selected depending on the system's design requirements. For example, according to some modalities, permanent magnets (or some
Petition 870190046951, of 20/05/2019, p. 80/196
47/110 m and part of them) can be made of NdFeB, which produces a magnetic field with a relatively high magnetic field per unit volume of matter once magnetized. According to some modalities, SmCo material is used to form permanent magnets, or some part of them. While NdFeB produces higher field strengths (and generally more economically than SmCo), SmCo exhibits less thermal derivation and thus provides a more stable magnetic field in view of temperature fluctuations. Other types of permanent magnet materials can be used as well, since aspects are not limited in this regard. In general, the type or types of permanent magnet material used will depend, at least in part, on requirements for field strength, temperature stability, weight, cost and / or ease of use of a given Bo magnet implementation.
[0093] The permanent magnet rings are sized and arranged to produce a homogeneous field of a desired intensity in the central region (field of view) between the permanent magnets 2110a and 2110b. In the illustrative embodiment illustrated in figure 2E, each permanent magnet ring comprises a plurality of segments, each segment formed using a plurality of blocks that are stacked in the radial direction and positioned adjacent to each other around the periphery to form the respective ring. The inventors appreciate that by varying the width (in a direction tangent to the ring) of each permanent magnet, less waste of usable space can be achieved while using less material. For example, the space between the stacks that does not produce useful magnetic fields can be reduced by varying the width of the blocks, for example, as a function of the radial position of the block, allowing for a tighter fit to reduce the wasted space and maximize the amount magnetic field that can be generated in a given space.
Petition 870190046951, of 20/05/2019, p. 81/196
48/110
The dimensions of the blocks can also vary in any desired way to facilitate the production of a magnetic field of desired intensity and homogeneity, as discussed in greater detail below.
[0094] The Bo 2100 magnet additionally comprises the fork 2120 configured and arranged to capture the magnetic flux generated by permanent magnets 2110a and 2110b and direct it to the opposite side of the Bo magnet to increase the flow density between the permanent magnets 2110a and 2110b, increasing the field strength within the field of view of the Bo magnet. By capturing the magnetic flux and directing it to the region between the permanent magnets 2110a and 2110b, less permanent magnet material can be used to achieve a desired field strength, thereby reducing the size, weight and cost of the magnet Bo. Alternatively, for certain permanent magnets, the field strength can be increased, thereby improving the system's SNR without having to use increased amounts of permanent magnet material. For the illustrative Bo magnet 2100, the fork 2120 comprises a frame 2122 and plates 2124a and 2124b. In a manner similar to that described above with respect to fork 2020, plates 2124a and 2124b capture the magnetic flux generated by permanent magnets 2110a and 211b and direct it to frame 2122 to be circulated through the magnetic return path of the fork to increase the flux density in the field of view of the Bo magnet. Fork 2120 can be constructed from any desired ferromagnetic material, for example, low carbon steel. CoFe and / or silicon steel, etc. to provide the desired magnetic properties for the fork. According to some modalities, plates 2124a and 2124b (and / or frame 2122 or parts thereof) can be constructed from silicon steel or similar in areas where gradient coils can induce, predetermining
Petition 870190046951, of 20/05/2019, p. 82/196
49/110 nly, swirling chains.
[0095] The illustrative table 2122 comprises arms 2123a and 2123b that fix the plates 2124a and 2124b, respectively, and supports 2125a and 2125b providing the magnetic return path for the flow generated by the permanent magnets. The arms are generally designed to reduce the amount of material needed to support the permanent magnets while providing sufficient cross section for the return path to the magnetic flux generated by the permanent magnets. The 2123a arms have two supports within a magnetic return path for the Bo field produced by the Bo magnet. The supports 2125a and 2125b are produced with a space 2127 formed between them, providing a measure of stability for the frame and / or lightness to the structure, while providing a sufficient cross-section for the magnetic flux generated by the permanent magnets. For example, the cross section required for the return path of the magnetic flux can be divided between the two support structures, thus providing a sufficient return path while increasing the structural integrity of the frame. It should be appreciated that additional supports can be included in the structure, as the technique is not limited to use with just two supports and any particular number of multiple support structures.
[0096] As discussed above, illustrative permanent magnets 2110a and 2110b comprise a plurality of rings of permanent magnetic material, arranged concentric with a permanent magnet disk in the center. Each ring may comprise a plurality of stacks of ferromagnetic material to form the respective ring, and each stack may include one or more blocks, which may have any number (including a single block in some embodiments and / or in some of the rings). The blocks that form each ring can be sized and arranged to produce a magnetic field
Petition 870190046951, of 20/05/2019, p. 83/196
50/110 desired. The inventors recognized that the blocks can be sized in several ways to reduce cost, weight and / or improve the homogeneity of the magnetic field produced, as discussed in greater detail with respect to the illustrative rings that together form the permanent magnets of a magnet. Bo, according to some modalities.
[0097] Figure 2F illustrates a Bo 2200 magnet, according to some modalities. The Bo 2200 magnet can share the design components with the Bo 2100 magnet illustrated in figure 2E. In particular, the Bo 2200 magnet is formed by permanent magnets 2210a and 2210b arranged in bi-planar geometry with a fork 2220 coupled to it to capture the electromagnetic flow produced by permanent magnets and transfer the opposite permanent magnet flow to increase the density of flow between permanent magnets 2210a and 2210b. Each of the permanent magnets 2210a and 2210b is formed from a plurality of concentric permanent magnets, as illustrated by the permanent magnet 2210b comprising an outer ring of permanent magnets 2214a, an intermediate ring of permanent magnets 2214b, an inner ring of permanent magnets 2214c, and a permanent magnet disk 2214d in the center. The permanent magnet 2210a can comprise the same set of permanent magnet elements as the permanent magnet 2210b. The permanent magnet material used can be selected depending on the requirements of the system design (for example, NdFeB, SmCo, etc., depending on the desired properties).
[0098] The permanent magnet rings are dimensioned and arranged to produce a homogeneous field of a desired intensity in the central region (field of view) between the permanent magnets 2210a and 2210b. In particular, in the illustrative embodiment illustrated in figure 2F, each permanent magnet ring comprises a plurality of
Petition 870190046951, of 20/05/2019, p. 84/196
51/110 circular arc segments sized and positioned to produce a desired Bo magnetic field, as discussed in greater detail below. Similar to fork 2120 illustrated in figure 2E, fork 2220 is configured and arranged to capture the magnetic flux generated by permanent magnets 2210a and 2210b and direct them to the opposite side of magnet Bo to increase the flow density between the permanent magnets 2210a and 2210b. The 2220 fork thus increases the field strength within the field of view of the Bo magnet with less permanent magnet material, reducing the size, weight and cost of the Bo magnet. Fork 2220 also comprises a frame 2222 and plates 2224a and 2224b which, in a similar manner to that described above with respect to fork 2220, capture and circulate the magnetic flux generated by permanent magnets 2210a and through the magnetic return path of the fork to increase the flux density in the field of view of the Bo magnet. The structure of the 2220 fork can be similar to that described above to provide enough material to accommodate the magnetic flux generated by the permanent magnets and provide sufficient stability, while minimizing the amount of material used to, for example, reduce the cost and weight of the Bo magnet. .
[0099] Since a permanent Bo magnet, once magnetized, will produce its own persistent magnetic field, energy is not required to operate the permanent Bo magnet to generate its magnetic field. As a result, a significant (often dominant) contributor to the overall energy consumption of an MRI system can be eliminated, thereby facilitating the development of an MRI system that can be energized using the electricity from the power plant. (for example, through a standard wall outlet or outlets for large common appliances). As discussed above, the inventors developed low-energy, portable, low-power MRI systems
Petition 870190046951, of 20/05/2019, p. 85/196
52/110 that can be developed in virtually any environment and that can be brought to the patient who will undergo an imaging procedure. In this way, patients in emergency rooms, intensive care units, operating theaters and a host to other locations can benefit from MRI in circumstances where MRI is conventionally unavailable. [00100] Figures 3A and 3B illustrate a portable low-field MRI system 300 suitable for use in carrying out change detection techniques described here, according to some modalities. System 300 can include magnetic and power components, and potentially other components (eg, thermal management, console, etc.), arranged together in a single, generally transportable and transformable structure. The system 300 can be designed to have at least two configurations: a configuration adapted to transport and store, and a configuration adapted to operate. Figure 3A illustrates system 300 when attached for transportation and / or storage and figure 3B illustrates system 300 when transformed for operation. The system 300 comprises a part 390A that can be slid in and retracted from a part 390B when transforming the system from its transport configuration to its operating configuration, as indicated by the arrows illustrated in figure 3B. The 390A part can house the electronic parts of power, console (which can comprise an interface device, such as a touch panel monitor) and thermal management. The 390A part can also include other components used to operate the system 300, as needed.
[00101] Part 390B comprises the magnetic components of the low-field MRI system 300, including laminated panels in which the magnetic components are integrated in any of the combinations discussed here. When transformed for trust
Petition 870190046951, of 20/05/2019, p. 86/196
53/110 guration adapted to operate the system to perform the MRI (as illustrated in figure 3B), support surfaces of parts 390A and 390B provide a surface on which the patient can lie down. A sliding bed or surface 384 can be provided to facilitate the patient's slide into position, so that a portion of the patient to be examined is within the field of view of the magnetic components of the low field MRI. System 300 provides a compact, portable configuration of a low-field MRI system that facilitates access to MRI imaging in circumstances where it is conventionally unavailable (for example, in NICU).
[00102] Figures 3A and 3B illustrate an example of a convertible low-field MRI system that uses a bi-planar magnet forming and representing the region between housings 386A and 386B. Housings 386A and 386B house magnetic components for the 300 convertible system. According to some modalities, magnetic components can be produced, manufactured and disposed using exclusively lamination techniques, exclusively traditional techniques or using a combination of both (for example, using hybrid techniques). The convertible low-field MRI system 300 allows the system to be brought to the patient to facilitate monitoring of the patient's target anatomy. For example, the convertible low field MRI system 300 can be brought to a patient at NICU and the unconscious patient can be placed on the sliding bed and positioned within the system's field of view. The patient can then be monitored by obtaining continuous, periodic and / or regular MRI images over an extended period of time (for example, over the course of one or more hours) to assess the changes that occur using any one of the several change detection techniques described
Petition 870190046951, of 20/05/2019, p. 87/196
54/110 here.
[00103] Figures 3C and 3D illustrate views of another 3800 portable MRI system, which can be used to implement the various change detection techniques, according to some modalities of the technology described here. The 3800 portable MRI system comprises a Bo 3810 magnet formed, in part, by an upper magnet 3810a and a lower magnet 3810b having a fork 3820 coupled thereto to increase the flow density within the imaging region. The Bo 3810 magnet can be housed in the 3812 magnet housing along with the 3815 gradient coils (for example, any of the gradient coils described in US order No. 14 / 845,652, entitled Low Field Magnetic Resonance Imaging Methods and Apparatus and deposited on September 4, 2015, which is incorporated herein by reference in its entirety). According to some modalities, the Bo 3810 magnet comprises an electromagnet. According to some modalities, the Bo 3810 magnet comprises a permanent magnet (for example, any permanent magnet described in US order No. 15 / 640.369, entitled LOW-FIELD MAGNETIC RESONANCE IMAGING METHODS AND APPARATUS, deposited on June 30, 2017, which is incorporated by reference here in its entirety).
[00104] The 3800 portable MRI system additionally comprises a 3850 base housing the electronic parts necessary to operate the MRI system. For example, the 3850 base can accommodate electronic parts including, but not limited to, one or more gradient power amplifiers, a computer in the system, a power distribution unit (PDU), one or more power supplies, and / or any other power components configured to operate the MRI system using electricity from the supply center (for example, through a connection to a wall outlet
Petition 870190046951, of 20/05/2019, p. 88/196
55/110 standard and / or a socket for large appliances). For example, the 3870 base can house low-energy components, such as those described here, allowing, at least in part, the portable MRI system to be powered from readily available wall outlets. Accordingly, the 3800 portable MRI system can be brought to the patient and plugged into a nearby wall outlet.
[00105] The 3800 portable MRI system additionally comprises 3860 movable sliding elements that can be opened and closed and positioned in a variety of configurations. The 3860 sliding elements include 3865 electromagnetic shielding, which can be made from any suitable conductive or magnetic material, to form a mobile shield to attenuate electromagnetic noise in the operating environment of the portable MRI system to protect the imaging region against at least part of the electromagnetic noise. As used here, the term electromagnetic protection refers to conductive or magnetic material configured to attenuate the electromagnetic field in a spectrum of interest and positioned or arranged to protect a space, object and / or component of interest. In the context of an MRI system, electromagnetic protection can be used to protect electronic components (for example, power components, cables, etc.) of the MRI system, to protect the imaging region (for example, view) of the MRI system, or both.
[00106] The degree of attenuation achieved for electromagnetic protection depends on several factors including the material used, the thickness of the material, the frequency spectrum for which electromagnetic protection is desired or necessary, the size and shape of the openings in the electromagnetic protection ( eg the size of the spaces in a conductive interlacing, the size of the unprotected parts or spaces in the shield, etc.) and / or the orientation of the openings
Petition 870190046951, of 20/05/2019, p. 89/196
56/110 with respect to an incident electromagnetic field. Thus, electromagnetic protection generally refers to any conductive or magnetic protection that acts to attenuate at least part of the electromagnetic radiation and which is positioned to protect, at least partially, a certain space, object or component by attenuating at least part of electromagnetic radiation. [00107] It must be appreciated that the frequency spectrum for which protection (attenuation of an electromagnetic field) is desired may differ depending on what is being protected. For example, electromagnetic protection for certain electronic components can be configured to attenuate different frequencies in addition to electromagnetic protection for the imaging region of the MRI system. With respect to the imaging region, the spectrum of interest includes frequencies that influence, impact and / or degrade the capacity of the MRI system to excite and detect an MR response. In general, the spectrum of interest for the imaging region of an MRI system corresponds to the frequencies around the nominal operating frequency (that is, the Larmor frequency) at a determined magnetic field strength Bo, for which the system receiver is configured for or able to detect. This spectrum is referred to here as the operational spectrum for the MRI system. Thus, the electromagnetic protection that provides protection for the operational spectrum refers to the conductive or magnetic material arranged or positioned to attenuate the frequencies, at least within the operational spectrum for at least part of an imaging region of the MRI system. .
[00108] In the 3800 portable MRI system illustrated in figures 3C and 3D, mobile guards are thus configurable to provide protection in different arrangements, which can be adjusted as necessary to accommodate a patient, provide access to a pa
Petition 870190046951, of 20/05/2019, p. 90/196
57/110 aware and / or according to a determined imaging protocol. For example, for the imaging procedure illustrated in figure 3E (for example, a brain scan), once the patient has been positioned, 3860 sliding elements can be closed, for example, using the 3862 handle to provide protection electromagnetic 3965 around the imaging region except for the opening that accommodates the patient's upper back. In the imaging procedure illustrated in Figure 3F (e.g., a knee scan), the sliding elements 3960 can be arranged to have openings on both sides to accommodate the patient's legs. Accordingly, the mobile guards allow the guards to be configured in the appropriate arrangements for the imaging procedure and to facilitate the positioning of the patient properly within the imaging region.
[00109] In some embodiments, a noise reduction system comprising one or more noise reduction and / or compensation techniques can be performed to suppress at least part of the electromagnetic noise that is not blocked or sufficiently attenuated by protection 3865. In particular , the inventors developed noise reduction systems configured to suppress, prevent and / or reject electromagnetic noise in the operating environment in which the MRI system is located. According to some of the modalities, these noise suppression techniques work in conjunction with mobile protections to facilitate operation in the various protection configurations in which the sliding elements can be arranged. For example, when the 3960 sliding elements are arranged as shown in figure 3F, the increased levels of electromagnetic noise are likely to enter the imaging region through the openings. As a result, the su component
Petition 870190046951, of 20/05/2019, p. 91/196
58/110 noise pressure will detect increased electromagnetic noise levels and adapt the noise suppression and / or noise response accordingly. Due to the dynamic nature of the noise suppression and / or prevention techniques described here, the noise reduction system is configured to respond to changing noise conditions, including those resulting from different arrangements of the mobile guards. In this way, a noise reduction system, according to some modalities, can be configured to operate in conjunction with mobile protections to suppress electromagnetic noise in the operating environment of the MRI system in any of the protection configurations that can be used. , including configurations that are substantially unprotected (for example, configurations without movable guards).
[00110] To ensure that movable guards provide protection regardless of the arrangements in which the sliding elements are located, electric gaskets can be arranged to provide continuous protection along the periphery of the movable guard. For example, as illustrated in figure 3D, electric gaskets 3867a and 3867b can be provided at the interface between the sliding elements 3860 and the housing and magnet to maintain the provision of continuous protection along that interface. According to some modalities, electric gaskets are beryllium extensions or beryllium and copper extensions, or the like (for example, aluminum gaskets), which maintain the electrical connection between 3865 guards and the ground during and after the sliding elements 3860 are moved to the desired positions around the imaging region. According to some modalities, electric gaskets 3867c are provided at the interface between the sliding elements 3860, as illustrated in figure 3F, so that continuous protection is provided between the sliding elements in the arrangements in which the elements
Petition 870190046951, of 20/05/2019, p. 92/196
59/110 slides are joined. Accordingly, 3860 mobile sliding elements can provide configurable protection for the portable MRI system.
[00111] In order to facilitate transport, a 3880 motorized component is provided to allow the portable MRI system to be driven from location to location, for example, using a control such as a joystick or other control mechanism provided in or remote with respect to the MRI system. In this way, the 3800 portable MRI system can be transported to the patient and maneuvered to the bed to perform imaging, as illustrated in figures 3E and 3F. As discussed above, Figure 3E illustrates a portable MRI system 3900 that was transported to a patient's bed to perform a brain scan. Figure 3F illustrates the portable MRI system 3900 that was transported to a patient's bed to perform a scan of the patient's knee.
[00112] The portable MRI systems described here can be operated from a portable electronic device, such as a notepad, tablet, smartphone, etc. For example, the 3875 tablet computer can be used to operate the portable MRI system to run desired imaging protocols and to view the resulting images. The 3875 tablet computer can be connected to a secure cloud to transfer images for data sharing, telemedicine, and / or deep learning of data sets. Any of the techniques for using network connectivity described in US application No. 14 / 846.158, entitled Automatic Configuration of a Low Field Magnetic Resonance Imaging System, filed on September 4, 2015, which is incorporated herein by reference in its entirety, can be used with respect to the portable MRI systems described here.
[00113] Figure 3G illustrates another example of an MRI system
Petition 870190046951, of 20/05/2019, p. 93/196
60/110 portable · according to some modalities of the technology described here. The portable MRI system 4000 can be similar in many ways to the portable MRI systems illustrated in figures 3C-3F. However, the 4060 sliding elements are constructed differently, with the 4065 protection, resulting in electromagnetic protections that are easier and cheaper to manufacture. As discussed above, a noise reduction system can be used to allow the operation of a portable MRI system in unprotected rooms and with varying degrees of protection around the imaging region in the system itself, not including any, or substantially none of the device-level electromagnetic protections for the imaging region.
[00114] It should be appreciated that the electromagnetic protections illustrated in figures 3C to 3G are illustrative and providing protection for an MRI system is not limited to the illustrative electromagnetic protection described above. Electromagnetic protection can be implemented in any suitable way using any of the suitable materials. For example, electromagnetic protection can be formed using interlacing, conductive fabrics, etc. that can provide a movable curtain to protect the imaging region. Electromagnetic protection can be formed using one or more conductive strips (for example, one or more strips of conductive material) coupled to the MRI system as a fixed, mobile or configurable component to protect the imaging region from electromagnetic interference, some examples of which are described in more detail below. Electromagnetic protection can be provided by including materials in doors, sliding elements or moving or fixed part of the housing. Electromagnetic protections can be developed as wired or mobile components, since the packages are not limited in this respect.
Petition 870190046951, of 20/05/2019, p. 94/196
61/110 [00115] Figure 4 illustrates a method of monitoring a patient using low-field MRI to detect changes, according to some modalities. In act 410, the first MR imaging data is acquired by a low-field MRI device from a target anatomy part (eg, a part of the brain, a part of a knee, etc.) from a patient positioned within of the low field MRI device. Placing a patient within the low field device refers to placing the patient with respect to the magnetic components of the low field MRI device, so that a portion of the patient's anatomy is located within the field of view of the field MRI device. down, so that MR image data can be acquired. The term MR image data is used here to refer to MR data including generically, but not limited to, MR data prior to image reconstruction (for example, k space MR data) and MR data that have been processed in some way (for example, post-image reconstruction MR data, such as a three-dimensional volumetric (3D) image). Since both the registration and change detection techniques described here can be performed on any domain (or a combination of domains), the term MR image data is used to refer to the diagnosis of MR data for domain and / or whether image reconstruction (or any other processing) has been carried out. As an illustrative application, MR image data from a patient's brain can be acquired to monitor temporal changes within the brain (for example, changes related to an aneurysm or bleeding within the brain, changes in a tumor or other abnormality of changes in chemical composition, etc.).
[00116] In act 420, subsequent (near) MR image data is acquired from the same or substantially
Petition 870190046951, of 20/05/2019, p. 95/196
62/110 same part of the anatomy included in the first MR image data. The next MR image data can be acquired immediately after the acquisition of the first MR image data, or can be obtained after a desired delay period (for example, after 1, 2, 3, 4, 5, 10 , 15, 20 minutes, etc.). As a result, the next MR image data captures the anatomy part after some related amount of time has passed. The inventors appreciated that low-field MRI facilitates relatively rapid image acquisition, allowing for a temporal sequence of MR image data in relatively quick succession, thereby capturing changes that may be of interest to the physician. The accessibility, availability and / or relatively low cost of the low-field MRI system allows MR data to be acquired over extended periods of time in any time interval necessary to monitor and / or otherwise observe and evaluate the patient .
[00117] As with the first MR image data, the next MR image data can take any shape (for example, a 3D volumetric image, a 2D image, k space MR data, etc.). According to some modalities, the next MR image data (or any subsequent acquired MR image data) are obtained using the same acquisition parameters used to acquire the first MR image data. For example, the same pulse sequence, field of view, SNR and resolution can be used to acquire MR signals from the same part of the patient. In this way, MR image data can be compared to assess changes that have occurred within the anatomy being examined. For example, as described below, MR image data can be used to determine if there is a change in the degree of deviation from the midline
Petition 870190046951, of 20/05/2019, p. 96/196
63/110 days in one patient. As another example, as described below, MR image data can be used to determine whether there is a change in size of an anomaly (for example, bleeding, injury, edema, stroke nucleus, stroke and / or swelling) in a patient. In other modalities, one or more acquisition parameters can be changed to change the acquisition strategy to acquire the next MR image data, as discussed in greater detail below with respect to figure 5.
[00118] The first, next and any subsequent acquired MR image data are referred to as respective MR image data tables. A frame sequence can be acquired and individual frames can be registered in a frame sequence acquired over time. Thus, a frame corresponds to the acquired MR image data, which represents the particular time in which the MR image data was acquired. Frames do not need to include the same amount of MR image data or correspond to the same field of view, but frames generally need to overlap sufficiently so that descriptors of suitable characteristics can be detected (for example, enough material in common among the staff).
[00119] In act 430, the first and next MR image data are recorded together or aligned with each other. Any suitable technique can be used to collect the first and next MmR image data together, or any acquired MR image data pair for which change detection processing is desirable. In the simplest case, the registration can be performed considering that the patient is quiet, so that the MR image data are aligned without transforming
Petition 870190046951, of 20/05/2019, p. 97/196
64/110 sea or deform the MR image data. However, such a simplified technique does not consider the patient's movement, changes resulting from breathing, etc., which may need to be compensated in other ways to avoid attributing observed differences between the images resulting from these factors to biological processes. More sophisticated recording techniques used to align MR image data to compensate for patient movement, breathing, etc., include, but are not limited to, the use of deformation models and / or correlation techniques adapted to data from MR image acquired at different times.
[00120] According to some modalities, the registration of acquired MR image data together involves the determination of a transformation that better aligns the MR image data (for example, in the sense of mean squares). The transformation between the MR image data acquired at different times can include translation, rotation, scale or any suitable linear or non-linear deformation, since the aspects are not limited in this aspect. The transformation can be determined at any desired scale. For example, a transformation can be determined for several identified subregions (for example, volumes including a number of voxels) of the MR image data, or it can be determined for each voxel in the MR image data. The transformation can be determined in any way, for example, using a deformation model that deforms an interlacing or coordinated frame of the first MR image data to the coordinated frame of the next MR image data and vice versa. Any suitable registration technique can be used, since the aspects are not limited in this regard. An illustrative process to record MR image data together, acquired at different times, according to some modality,
Petition 870190046951, of 20/05/2019, p. 98/196
65/110 is discussed in greater detail below with reference to figure 6.
[00121] In act 440, one or more changes are detected in the MR image data registered together. For example, since MR imaging data have been recorded together, differences between MR imaging data can be attributed to changes in the patient's anatomy being examined (for example, morphological changes in anatomy or other changes in biology or physiology of the examined anatomy), such as a change in the size of an aneurysm, increased or reduced bleeding, progress or retraction of a tumor or other tissue anomaly, changes in chemical composition or other biological or physiological changes of interest. The change detection can be carried out in any suitable way. For example, since MR image data were recorded together, change detection can be performed in k space using amplitude and phase information (coherent change detection), or change detection can be performed in the image domain using intensity information (non-coherent change detection). In general terms, the detection of coherent change may be more sensitive, revealing changes in the sub-voxel level. However, although non-coherent change detection can generally be less sensitive, change detection in the image domain can be more robust for joint registration errors.
[00122] In some modalities, the detection of change can be performed by deriving characteristics of each MR frame in a sequence of MR frames and comparing the characteristics with each other. For example, in some modalities, image processing techniques (for example, including the deep learning techniques described here) can be applied to each MR frame in a sequence of two or more MR frames, obtained
Petition 870190046951, of 20/05/2019, p. 99/196
66/110 by representing the patient's brain image, to identify a respective sequence of two or more midline deviation measurements. In turn, the sequence of midline deviation measurements can be used to determine whether there is a change in the degree of deviation from the midline for the patient being monitored. As another example, in some modalities, the image processing techniques (for example, the deep learning techniques described here) can be applied to each MR frame in a sequence of two or more MR frames, obtained by representing the brain image of a patient, to identify a respective sequence of two or more measurements of a size of the anomaly in the patient's brain (for example, bleeding, injury, edema, stroke nucleus, stroke penumbra and / or swelling) . In turn, the sequence of size measurements can be used to determine if there is a change in the size of the abnormality in the brain of a patient being monitored.
[00123] In some modalities, multi-resolution techniques can be used to perform change detection. For example, the first MR image data can correspond to a high resolution baseline image, and the MR image data acquired subsequently can correspond to low resolution images that can be correlated with the baseline image. high resolution. The acquisition of low resolution images can accelerate the frame rate of the change detection process, allowing the acquisition of more data in a shorter short period. Any suitable techniques or criteria can be used to determine what data to acquire for a low resolution image. The particular data to be acquired for a low resolution image can be determined using, for example, wavelets, selective k space sampling, dust filtering
Petition 870190046951, of 20/05/2019, p. 100/196
67/110 lifase, keyframe based techniques, etc. Sparse sampling of the k space over short time intervals (for example, selective sampling that varies with the k space time), as an example, results in a better time resolution.
[00124] The selection of particular data to be acquired can also be determined by detecting changes between the MR image data frames. For example, when a change is detected, a 2D ID or volume selection, having a field of view that includes the location of the detected change, can be selected for acquisition to interrogate a particular part of the anatomy showing change over time.
[00125] Using the detection of coherent change, differences in phase and amplitude in each frame of the acquired MR data are evaluated. For example, the frames recorded in the MR image data set can be subtracted to obtain difference information indicative of the changes that occur in the MR data. According to some modalities, a finite impulse response filter (FIR) is applied to each voxel in the table, which can be used as a reference. Filtering can also be used to provide a forward-facing filter that considers a number of frames through which to perform change detection. For example, a current, previous and next chart can be evaluated using a sliding window to analyze changes through a desired number of charts.
[00126] The inventors recognized that acquiring an entire 3D volume of MR data can take a substantial amount of time. In some modalities, change detection is used to selectively determine particular data (for example, particular lines in the k space) to acquire, so that the MR data used for image reconstruction can be acquired in one
Petition 870190046951, of 20/05/2019, p. 101/196
68/110 less time than it would take to acquire an entire 3D volume. For example, using the sliding window approach described above, an initial 3D volume can be purchased first. Then, at subsequent points in time, instead of reacquiring the full 3D volume, a subset of lines in the selected k space based on the parts of the image that are changing, can be acquired and the previous 3D volume can be updated with the newly acquired data. -acquired.
[00127] In some modalities, a particular feature or area of interest can be identified in advance, and the acquisition sequence can be customized to acquire k space lines that will emphasize the identified feature or area of interest. For example, the acquisition sequence can focus on acquiring just the edges of the k space or any other suitable part of the k space. In some embodiments, the area of interest identified may be a part of the anatomy. For example, to analyze post-surgical bleeding, it may not be necessary to acquire data on the entire anatomy. Instead, selected parts of the k space, which correspond to the anatomy of interest for monitoring, can be sampled multiple times in a relatively short period of time to allow a physician to closely monitor changes in the anatomy of interest over a time scale. lower, providing a high temporal correlation between acquisitions.
[00128] Using non-coherent change detection, the intensity of the voxels in the 3D images reconstructed from the acquired MR data can be compared to assess the changes as they occur over time. The detected changes, evaluated in a coherent way (for example, in k space) or non-coherent (for example, in 3D images) can be carried in any way. For example, changes in MR image data can be found
Petition 870190046951, of 20/05/2019, p. 102/196
69/110 sliced into images displayed to provide a visual indication to a doctor that changes have occurred over time. For example, voxels that undergo changes can be colored which, in turn, can be coded according to the extent to which the change has occurred. That way, a doctor can quickly see hot spots that are undergoing significant changes. Alternatively, or in addition, change detection can be performed by analyzing the regions through which changes are taking place. For example, connected component analysis can be used to find contiguous regions where voxel changes have occurred. That is, the regions of connected voxels that have undergone changes can be emphasized or displayed differently (for example, using color, shading, etc.) to indicate that the changes are taking place in the corresponding regions. Changes detected in the acquired MR image data can be carried in other ways, since the aspects are not limited to this aspect.
[00129] Format and volume analysis can also be performed to determine whether a particular feature of the anatomy of interest is changing (for example, growing or shrinking, progressing or regressing, or otherwise characterizing changes in features). For example, image processing techniques can be used to segment MR image data into regions and to evaluate one or more segment properties such as format, volume, etc. Changes to one or more segment properties can be ported to a doctor via a monitor or otherwise. For example, the size of a tumor can be monitored through a sequence of images to assess whether the tumor is growing or decreasing in size. As another example, bleeding in the brain can be monitored over time
Petition 870190046951, of 20/05/2019, p. 103/196
70/110 where the important change to be assessed is the bleeding volume. In this way, the acquired MR image data can be processed to segment the characteristics of interest (for example, tumor, bleeding, hemorrhage, etc.) and compute the volume of the corresponding characteristic.
[00130] It should be appreciated that the segmented volumes can be analyzed in other ways to characterize the metrics of interest for the segmented volume. For example, 2D and / or 3D format descriptors can be applied to segmented characteristics to characterize any number of aspects or properties of the segmented characteristic including, but not limited to volume, surface area, symmetry, texture, etc. Thus, the detection of change can be performed on the characteristics of interest captured in the acquired MR data to assess how the characteristics are changing over time. The changes detected in the segmented characteristics can be used not only to understand how the characteristic is evolving over time, but characteristics of the particular characteristics can be compared with the information stored to assist in the differentiation between healthy and unhealthy, normal and anomalous and / or to assess the danger of a particular condition. The information obtained from the MR data can also be stored together with the existing information to increase the information deposit that can be used for subsequent data analysis.
[00131] According to some modalities, the techniques can be used to remove changes in the data caused by regular or periodic movement, such as breathing and heartbeat, etc. By determining which parts of the image are changing and which are not, it is possible to focus the acquisition only on the parts of the image that are changing and not to acquire data for the
Petition 870190046951, of 20/05/2019, p. 104/196
71/110 parts of the image that are not changing. By acquiring a smaller data set, related only to the parts of the image that are changing, the acquisition time is compressed. In addition, some changes in the image are caused by periodic events such as breathing and heartbeat. In some modalities, periodic events are modeled based on their periodicity to allow a change detection process to ignore the periodic movements caused by periodic events when determining which parts of the image are changing, and should be the focus of the acquisition.
[00132] According to some modalities, the detection of change can be carried out by detecting the rate of change of MR image data through a sequence of acquired MR image data. As used here, a rate of change refers to any functional form of time. Detecting the rate of change can provide richer data regarding the matter being examined, such as an indication of the seriousness of a bleed, the size of a hemorrhage, an increase in midline deviation, the aggressiveness of the injury, etc. As an example, when a contrast agent is administered, there is a natural and expected way in which the contrast agent is absorbed into the body. The absorption of the contrast agent is detected as an increase in signal which will register as a change having a particular functional shape. The way in which the signal changes, as the contrast agent fades and / or is metabolized, will give rise to a detectable change also in the signal that will have a functional shape over time. The functional form of changes over time can provide information about the type, aggressiveness or other characteristics of an injury or other anomaly that can provide clinically useful and / or critical data. As another example, a stroke victim can be monitored for
Petition 870190046951, of 20/05/2019, p. 105/196
72/110 because a stroke has occurred, changes over the course of the injury caused by the stroke, which differ from expected, can be used to alert staff to unusual changes, provide a measure of drug effectiveness, or provide other information relevant to the condition of the patient. In general, detecting the rate of change can facilitate higher-order analysis of the matter being examined.
[00133] Techniques are available to facilitate faster acquisition of MR data, allowing faster image acquisition for low field MRI. For example, compressed perception techniques, sparse imaging set techniques, and MR digital printing are some examples of techniques that can accelerate MR image acquisition. Additionally, in some modalities, Doppler techniques can be used to analyze multiple image frames over a short period of time to estimate the speeds that can be used to filter out parts of the image that are not changing.
[00134] After the detection of changes in the acquired MR image data is complete, act 420 can be repeated to obtain additional MR image data, immediately or after waiting for a predetermined amount of time, before acquiring the data image of subsequent MR. Subsequently acquired MR image data can be compared with any previously acquired MR image data to detect changes that occurred during any desired time interval (for example, by repeating act 430 and 440). In this way, sequences of MR image data can be obtained and changes detected and ported to facilitate the understanding of temporal changes that occur in the part of the patient's anatomy being monitored, observed and / or evaluated. It should be appreciated that any data
Petition 870190046951, of 20/05/2019, p. 106/196
73/110 of acquired MR images can be recorded and analyzed for changes. For example, data from successive MR images can be compared so that, for example, changes in a relatively small time scale can be detected. The detected change can be ported to a doctor so that the anatomy of interest can be monitored continuously, regularly and / or periodically.
[00135] Additionally, the acquired MR image data can be stored so that a doctor can request that the change detection be performed at desired points of interest. For example, a doctor may be interested in seeing the changes that have occurred in the past hour and may specify that change detection be performed between the MR image data acquired an hour ago and the current MR image data. The physician may specify a time interval, may specify multiple moments of interest, or may select image icons with a time stamp to indicate which MR image data the physician would like to have the change detection performed. In this way, the techniques described here can be used to monitor changes in progress and / or evaluate changes that occurred during any time interval during which MR image data was acquired. The change detection techniques described above can be used to allow monitoring, evaluation and observation of a patient over a period of time, thus allowing MRI to be used as a monitoring tool in ways in which conventional MRI and others modalities cannot be used.
[00136] In some embodiments, acquired MR image data can be used to assess the change with respect to a stored high field MRI scan. That way, a pa
Petition 870190046951, of 20/05/2019, p. 107/196
74/110 can be examined using a high field MRI scanner initially, but subsequent monitoring (which would not be feasible using the high field MRI) will be performed using a low field MRI system, examples of which are provided here. The change detection techniques described here can be applied not only to detect changes between MR image data sets acquired by a low field MRI system, but also to detect changes between MR image data acquired by a high field MRI system (for example, initially) and the MR image data acquired by a low field MRI system (for example, subsequently), regardless of the order in which the high field MR image data and low field MR image data is obtained.
[00137] Figure 5 illustrates a method of changing an acquisition strategy based, at least in part, on the observations made regarding the detection of change. The inventors have developed a multi-acquisition console that allows acquisition parameters to be modified immediately to dynamically update an acquisition strategy implemented by the low-field MRI system. For example, commands for the low field MRI system can be sequenced from the console to achieve dynamic updates to the acquisition process. The inventors appreciated that the ability to dynamically update the acquisition parameters and / or change the acquisition strategy can be exploited to achieve a new paradigm for MRI, allowing the MRI system to be used to monitor a patient and adapt the strategy acquisition based on observations of acquired MR image data (for example, based on change detection information).
Petition 870190046951, of 20/05/2019, p. 108/196
75/110 [00138] In method 500, illustrated in figure 5, acts 510-540 can be similar to acts 410-440 in method 400 illustrated in figure 4 to obtain change detection information in relation to the MR image data obtained by a low field MRI system. In Act 550, at least one acquisition parameter can be updated, changed or modified based on the results of the change detection. The acquisition parameters that can vary are not limited in any way and can include any or combination of field of view, signal to noise ratio (SNR), resolution, pulse sequence type, etc. Some examples of the acquisition parameters that can be changed are described in more detail below.
[00139] According to some modalities, the change detection information can be used to update the acquisition parameters to, for example, increase SNR of MR data obtained from a particular region. For example, based on the characteristics of the joint record (eg transformation properties, deformation models, etc.) and / or changes observed in particular regions, it may be desirable to increase the SNR in those regions to, for example, better evaluate the present subject, to improve an additional change detection or, in other way, to obtain more information regarding the part of the anatomy being monitored and / or observed. Similarly, acquisition parameters can be changed to obtain higher resolution MR data for particular regions of the part of the anatomy being monitored / observed. Detecting the change can reveal that a patient has been moved or the subject of interest is no longer ideally in the field of view. This information can be used to dynamically change the field of view of subsequent image acquisition.
[00140] According to some modalities, the type of sequence
Petition 870190046951, of 20/05/2019, p. 109/196
76/110 of pulse that is applied can be changed based on what was observed in the change detection data obtained from the acquired MR image data. Different pulse sequences may be better for capturing the particular types of information and these differences can be exploited to allow for proper exploration based on observed change detection data. Due, at least in part, to the dynamic capacity of the system developed by the inventors, different pulse sequences can be interleaved, alternated or otherwise used to acquire MR data that capture the information of interest. For example, a rapid rotating echo sequence may have been used to acquire multiple frames of MR image data and the results of change detection may suggest the benefit of changing to a different pulse sequence, for example, a bSSFP sequence , to observe a particular change (for example, to obtain different MR data, to allow for a higher SNR or resolution in a particular region, etc.). Thus, changes that may not be observable using a sequence type can be observed by changing the type of pulse sequence being used.
[00141] As another example, pulse sequences can be chosen for the type of contrast provided (for example, T1, T2, etc.) or type of information that is captured, and the appropriate pulse sequence can be used to obtain data of MR, which can be changed dynamically during the monitoring process. The choice of pulse sequence or combination of pulse sequences used can be guided by the change detection information that is obtained. For example, MR data can be captured using a given pulse sequence and, based on the change detection information obtained (for example, based on information obtained by performing act 540), the pulse sequence can be alte
Petition 870190046951, of 20/05/2019, p. 110/196
77/110 to explore a region using magnetic resonance spectroscopy (MRS). Thus, the exploration of the chemical composition of a part of the anatomy being monitored can be initiated as a result of the changes observed in the MR data.
[00142] It should be appreciated that the acquisition parameters can vary dynamically at any time during the acquisition. That is, a full acquisition needs to be completed before the acquisition strategy changes. As a result, the acquisition parameters can be updated based on partial acquisition and / or partial image reconstruction to facilitate an acquisition strategy that is fully dynamic. The ability to dynamically update any or any combination of acquisition parameters allows MRI to be used as a monitoring and exploration tool, whereas conventional MRI systems cannot be used in this way.
[00143] Some applications, such as diffusion-weighted imaging (DWI), require substantial amounts of energy due to the higher gradient fields required for such applications. In some embodiments, energy savings can be achieved by merging acquisitions into a DWI sequence (or another) with acquisitions that require less energy. Allowing dynamic updating of the acquisition parameters during an acquisition, any combination and interleaving of acquisition sequences to achieve a desired objective (eg low energy consumption, reduced heating, reduced voltage in the gradient coils, etc.) can be accomplished.
[00144] In some modalities, biological or physiological events that unfold over a relatively short period of time can be studied using the change detection techniques described here. For example, for rotation labeling
Petition 870190046951, of 20/05/2019, p. 111/196
78/110 arterial, a complete data set can be obtained initially, and subsequent acquisitions can sparse sample the data. Blood perfusion over time can be monitored for change detection, where changes in the image correspond to the flow of blood into a particular region of the examined anatomy.
[00145] As discussed above, the joint recording of MR image data acquired at different times allows the identification of changes in MR data by reducing the effect of patient movement on the change detection process. The joint registration can be performed with a model for deformation purposes. The deformation interlacing captures changes in shape and distribution over time, which occur from the patient's Sufi movements or from biological morphology. In order to keep the record through the tables, as the examined volume moves or deforms, the space acquisition strategy k can be updated based on new restrictions of the deformed volume. For example, the acquisition parameters that affect the field of view, SNR, resolution, etc., can be updated based on the new restrictions of the deformed volume.
[00146] Figure 6 illustrates a technique 600 for registering frames of MR image data together, according to some modalities. For example, registration technique 600 can be used to align a pair of frames acquired at two separate times. In act 610, one or more characteristic descriptors that appear or are common to the frames being registered together, are detected. Feature descriptors can be any feature present in the MR image between frames that can be reliably detected. Features can include local features such as edges, corners, bumps, etc. and / or may
Petition 870190046951, of 20/05/2019, p. 112/196
79/110 include meeting characteristics such as curves, contours, shape, intensity distributions and / or patterns, etc. Any characteristic that can be reliably detected between the tables can be used as a characteristic descriptor, since the aspects are not limited in this respect. Any suitable technique can be used to determine characteristic descriptors including, but not limited to, SIFT, SURF, U-SURF, CenSurE, BRIED, ORB, and corner detector techniques, such as FAST, Harris, Hessian and ShiTomasi.
[00147] After the characteristic descriptors between the tables have been determined, the process proceeds to Act 620, where the associated sub-regions through the tables are correlated. The correlation calculations between the sub-regions can be performed in any number of dimensions (for example 1D, 2D, 3D), since the aspects are not limited in this respect. After the correlations between the subregions are determined, the process proceeds to act 630, where the warped or deformed model, from frame to frame, is determined based on the correlations between the subregions in different frames. Once the deformation of the model is determined between the frames, the process proceeds to act 640, where the model deformation is used to record data together across multiple frames.
[00148] Once the data are recorded together, the change detection metrics including, but not limited to those discussed above, such as coherent changes, non-coherent changes, and others including changes in position, speed, acceleration or derived vectors can be determined using data recorded together. Other metrics including segmentation and geometric shape descriptors, such as surface area, ripple volume, spherical harmonic base coefficients, etc. also
Petition 870190046951, of 20/05/2019, p. 113/196
80/110 can be determined based on the data recorded in common and, optionally, the metrics can be used to update the acquisition parameters for future acquisitions immediately as discussed above.
[00149] As described above, the inventors have developed techniques for using low field MRI to monitor a patient to determine if there is a change in a degree of midline deviation in the patient's brain. Midline deviation refers to the amount of displacement of the midline of the brain from its normal symmetrical position due to trauma (for example, stroke, hemorrhage, or other injury) and is an important indicator for doctors of the seriousness of the brain trauma. The midline deviation can be characterized as a change in the brain beyond its midline, usually in the direction away from the affected side (for example, one side with an injury).
[00150] In some embodiments, the midline deviation can be measured as the distance between a midline structure of the brain (for example, a point in the septum pellucidum) and a line designated as the midline. The midline can be coplanar with the falx cerebri (also known as falx cerebral), which is a fold in increasing shape of a meningeal layer of the dura mater that descends vertically in the longitudinal fissure between the cerebral hemispheres of the human brain. The midline can be represented as a line that connects the anterior and posterior fixations of the falx cerebri to the internal board of the skull.
[00151] As an example, illustrated in figure 7A, the midline 702 is a line that connects the anterior and posterior fixation points 706a and 706b of the falx cerebri. In this example, the midline deviation can be measured as the distance between the measuring point 706c in the septum pellucidum and the midline 702. This distance is the length
Petition 870190046951, of 20/05/2019, p. 114/196
81/110 to line 704 defined by end points 706c and 706d and which is orthogonal to the middle line 702.
[00152] As another example, illustrated in figure 7B, the middle line 712 is a line that connects the anterior and posterior fixation points 716a and 716b of the falx cerebri. In this example, the deviation from the midline can be measured as the distance between the measurement point 716c on the septum pellucidum and the midline 712. This distance is the length of line 714 defined by the end points 716c and 716d, and which is orthogonal to the midline 712.
[00153] Figure 8 is a flow chart of an illustrative process 800 to determine a degree of change in a patient's midline deviation, according to some modalities of the technology described here. In some embodiments, the entire process 800 can be performed while the patient is inside a low field MRI device, which can be any suitable type described here including, for example, any of the low field MRI devices illustrated in figures from 3A to 3G.
[00154] Process 800 begins in Act 802, where the low-field MRI device acquires the initial MRI data from a target part of the patient's brain. As described here, the term MR image data is used here to refer to MR data including generically, but not limited to, MR data prior to image reconstruction (for example, k-space MR data) and RMs that have been processed in some way (for example, post-image reconstruction MR data, such as a three-dimensional volumetric (3D) image). In some embodiments, the initial MR data may include one or more two-dimensional images of the respective brain slices (for example, two, three, four, five, etc. neighboring slices). When multiple slices are included, the slices can be neighbors. For example, MR data
Petition 870190046951, of 20/05/2019, p. 115/196
Initial 82/110 may include one or more 2D images of one or more respective slices in which the two lateral ventricles are prominent. [00155] Next, in act 804, the initial MR image data is provided as input to a trained statistical classifier in order to obtain the corresponding initial output. In some modalities, before being supplied to the trained statistical classifier, the initial MR image data can be pre-processed, for example, by new sampling, interpolation, related transformation and / or using any other suitable pre-processing technique , since the technology aspects described here will not be limited in this regard.
[00156] In some modalities, the output of the trained statistical classifier may indicate one or more initial locations, in the initial MR data, of one or more landmarks associated with at least one midline structure of the patient's brain. This location or locations can be identified from the output of the statistical classifier trained in act 806 of process 800. The output can specify the locations directly or indirectly. In the latter case, the locations can be derived from the information included in the output of the trained statistical classifier.
[00157] For example, in some modalities, the output of the trained statistical classifier can indicate the locations of the fixation points of the anterior and posterior falx cerebri and the location of a measurement point in the septum pellucidum. When the initial MR data includes a 2D image of a corresponding slice, the output of the trained statistical classifier can indicate the locations of the landmarks (for example, fixation points of the falx cerebri and measurement point in the septum pellucidum) within the 2D image . As described above, the locations of the fixation points of the falx cerebri and the measurement point in the septum pellucidum can be used to perform
Petition 870190046951, of 20/05/2019, p. 116/196
83/110 a midline deviation measurement.
[00158] In some modalities, the trained statistical classifier may be a neural network statistical classifier. For example, the statistical training classifier may include a convolutional neural network (for example, as illustrated in figures 9A and 9B), a convolutional neural network and a recurrent neural network, such as a short term, long memory network (for example, example, as illustrated in figures 9A and 9C), a fully convolutional neural network (for example, as illustrated in figure 10), and / or any other suitable type of neural network. The trained statistical classifier can be implemented in software, in hardware, or using any suitable combination of software and hardware. In some embodiments, one or more machine learning software libraries can be used to implement the trained statistical classifier including, but not limited to Theano, Torch, Caffe, Keras and TensorFlow. These libraries can be used to train a statistical classifier (for example, a neural network) and / or using a trained statistical classifier. Aspects of the training of the trained statistical classifier used in acts 804 and 806 are described in greater detail below. It should be appreciated that the trained statistical classifier is not limited to being a neural network and can be any other suitable type of statistical classifier (for example, a support vector machine, a graphic model, a Bayesian classifier, a tree classifier decision making, etc.), since aspects of the technology described here are not limited to this aspect.
[00159] As discussed above, in some modalities, the trained statistical classifier can be a convolutional neural network. Figures 9A and 9B illustrate an illustrative example of such a convolutional neural network. As illustrated in figure 9A, an input image (a 256 x 256 image in this example) is provided as an input for
Petition 870190046951, of 20/05/2019, p. 117/196
84/110 the convolutional neural network, which processes the incoming image through an alternating series of convolutional and ensemble layers. In this example, the convolutional neural network processes the input image using two convolutional layers to obtain 32 feature maps of 256 x 256. Then, after an application of a set layer (for example, a maximum set layer), two additional convolutional layers are applied to obtain 64 feature maps of 128 x 128. Next, after an application of another set layer (for example, maximum set), two additional convolutional layers are applied to obtain 128 maps of 64 x 64 features. Next, after applying another set layer and another convolutional layer, the resulting 256 32 x 32 feature maps are provided as input to the part of the neural network illustrated in Figure 9B. In this part, after an additional convolution, the feature maps are processed through at least one fully connected layer to generate forecasts. The predictions may, in some modalities, indicate the locations of the fixation points of the falx cerebri (for example, posterior and anterior fixation points, and a measurement point in the septum pellucidum).
[00160] Figures 9A and 9C illustrate another illustrative example of a neural network that can be used as the trained statistical classifier, in some modalities. The neural network of figures 9A and 9C have a part of the convolutional neural network (illustrated in figure 9A, which was described above) and a part of recurrent neural network (illustrated in figure 90), which can be used to model the time constraints among the input images provided as inputs to the neural network over time. The part of the recurrent neural network can be implemented with a short-term, long-term memory (LSTM) neural network. Such a neural network architecture can be used
Petition 870190046951, of 20/05/2019, p. 118/196
85/110 to process a series of images obtained by a low field MRI device while performing a monitoring task. A series of images obtained by the low-field MRI apparatus can be provided as input to the CNN-LSTM neural network, within which the characteristics derived from at least one image obtained previously can be combined with characteristics obtained from an image obtained later to generate forecasts. [00161] In some embodiments, the neural networks illustrated in figures 9A to 9C may use a core size of 3 with a stride of 1 for convolutional layers, a core size of 2 for joint layers, and a variation escalation initiator.
[00162] In some embodiments, the neural networks illustrated in figures 9A to C can be used to process a single image (for example, a single slice) at a time. In other embodiments, the neural networks illustrated in figures 9A to 9C can be used to process multiple slices (for example, multiple neighboring slices) at the same time. In this way, the characteristics used to predict point locations (for example, locations of fixation points of the falx cerebri and a measurement point in the septum pellucidum) can be composed using information from a single slice or from multiple neighboring slices .
[00163] In some modalities, when multiple slices are being processed by the neural network, the convolations can be two-dimensional (2D) or three-dimensional (3D) convolutions. In some embodiments, processing can be slice-based so that characteristics are calculated for each slice using the slice information and one or more of its neighboring slices (only from the slice itself or the slice itself and one or more of its neighbors). In other modalities, processing poPetição 870190046951, of 20/05/2019, p. 119/196
86/110 to be a fully 3D processing channel, so that the characteristics for multiple slices are computed simultaneously using data present in all slices.
[00164] In some modalities, instead of using a convolutional neural network architecture with one or more fully connected output layers, as illustrated in figures 9A to 9C, a fully convolutional neural network architecture can be employed. In such an architecture, the output is a result of a single channel having the same dimensionality as the input. In this approach, a map of point locations (for example, fixation points of the falx cerebri) is created by introducing intensity profiles of Gaussian nucleus at point locations, with the trained neural network, to return these profiles using error loss medium square.
[00165] Figure 10 illustrates two fully convolutional neural network architectures, which can be used in some modalities. The first architecture, with processing involving the processing path (a), includes three parts: (1) an output compression part comprising a series of alternating convolutional or assembly layers; (2) a piece of short term, long memory (indicated by path (a)); and (3) an inlet expansion part comprising a series of alternating convolutional and non-convolutional layers. This type of architecture can be used to model temporal constraints, as can the neural network architecture of figures 9A and 9C. The second architecture, with processing involving the processing path (b), includes three parts: (1) an output compression part comprising a series of alternating convolutional and ensemble layers; (2) a part of the convolutional network (indicated by route (b)); and (3) an entrance expansion part comprising a series of layers
Petition 870190046951, of 20/05/2019, p. 120/196
87/110 convolutional and non-convolutional alternating and a center layer of mass. The center of mass layer computes the estimate as a center of mass computed from the location estimates returned at each location.
[00166] In some embodiments, the neural networks illustrated in figure 10 may use a core size equal to 3 for convolutional layers with pitch equal to 1, a core size equal to 2 for the set layers, a core size 6 with step 2 for non-convolution layers, and a variation scaling initializer. In some modalities, when multiple slices are being processed by one or more neural networks illustrated in figure 10, the convolutions can be two-dimensional (2D) or three-dimensional (3D) convolutions. In some embodiments, processing can be slice-based, so that characteristics are calculated for each slice using slice information and one or more of its neighboring slices. In other modalities, processing can be a fully 3D processing channel, so that the characteristics for multiple slices are computed simultaneously using the data present in all slices.
[00167] It should be appreciated that the neural network architectures illustrated in figures 9A to 9C and figure 10 are illustrative and that variations of these architectures are possible. For example, one or more other layers of neural network (for example, a convolutional layer, a non-convolutional layer, a rectified linear unit layer, an ascending sampling layer, a concatenated layer, a filler layer, etc.) be introduced in any of the neural network architectures of figures 9A to 9C and 10, as one or more additional layers and / or instead of one or more layers that are part of the illustrated architectures. As another example, the dimensionality of one or more layers can
Petition 870190046951, of 20/05/2019, p. 121/196
88/110 vary and / or the core size for one or more convolutional, joint and / or non-convolutional layers may vary.
[00168] Next, process 800 proceeds to act 808, where the next MR image data is acquired. The next MR image data is acquired after the initial MR data is acquired. Thus, although in some modalities acts 804 and 806 can be performed after act 808 is performed, act 808 is usually performed after act 802. The next MR image data can be acquired immediately after the acquisition of the initial MR image data, or can be obtained after a desired delay period (for example, in 1, 2, 3, 4, 5, 10, 15, 20 minutes, in an hour, in two hours, etc. ). As with the initial MR image data, the next MR image data can take any shape (for example, a 3D volumetric image, a 2D image, k-space data and MR, etc.). In some embodiments, the initial MR data and the next MR image data are of the same type. For example, each of the initial and close MR data can include one or more two-dimensional images of one or more respective brain slices (for example, neighbors). For example, the initial MR data can include multiple images of neighboring slices taken at first and the next MR data can include multiple images of the same neighboring slices taken at a second time after the first moment.
[00169] Next, process 800 proceeds to act 810 where the next MR image data is provided as input to the trained statistical classifier to obtain the next corresponding output. In some modalities, before being provided to the trained statistical classifier, the next MR image data can be pre-processed, for example, by resampling,
Petition 870190046951, of 20/05/2019, p. 122/196
89/110 interpellation, finite transformation, and / or using any other suitable pre-processing technique, since the aspects of the technology described here are not limited to this aspect. The next MR image data can be pre-processed in the same way as the initial MR data was pre-processed.
[00170] In some modalities, the next output of the trained statistical classifier may indicate one or more updated locations, in the next MR data, of one or more landmarks associated with at least one midline structure of the patient's brain. This location or locations can be identified from the output of the statistical classifier trained in act 812 of process 800. The output can specify the locations directly or indirectly. In the latter case, the locations can be derived from the information included in the output of the trained statistical classifier.
[00171] For example, in some modalities, the output of the trained statistical classifier obtained in act 812 can indicate the updated locations of the fixation points of the anterior and posterior falx cerebri and the updated location of a measurement point in the septum pellucidum. When the next MR data includes a 2D image of a corresponding slice, the corresponding output from the trained statistical classifier can indicate the updated locations of the landmarks (for example, fixation points of the falx cerebri and measurement point in the septum pellucidum) within the image in 2D. As described above, the updated locations of the fixation points of the falx cerebri and the measurement point in the septum pellucidum can be used to perform a new midline deviation measurement or an updated midline deviation measurement.
[00172] Next, process 800 proceeds to act 814, where the degree of change in the midline deviation is determined using the initial and updated locations of the landmarks associated with the structures.
Petition 870190046951, of 20/05/2019, p. 123/196
90/110 midline tures that were obtained in acts 806 and 812, respectively. For example, in some embodiments, the initial locations of the fixation points of the falx cerebri and the measurement point in the septum pellucidum can be used to determine (for example, calculate) an amount of initial midline deviation. The updated locations of the fixation points of the falx cerebri and the measurement point in the septum pellucidum can be used to determine an updated midline deviation amount. The initial and updated midline deviation quantities can be used to determine (for example, by assessing their difference) the degree of change in the patient's midline deviation during the time period between the acquisition of the initial MR data and the next MR data.
[00173] Next, process 800 proceeds to decision block 816, where it is determined whether it is necessary to carry out a new determination of the degree of change in the deviation from the midline. This determination can be made in any appropriate way (for example, by determining whether a limit number of interactions has been carried out, based on a schedule, based on manual input provided by a doctor, etc.), since aspects of the technology described here are not limited to this aspect. When it is determined that a new determination of the degree of change in the midline deviation needs to be performed, then process 800 returns to block 808 and acts 808-814 are repeated again (with newly obtained MR data being compared with the most recently obtained MR data). On the other hand, when it is determined that a new determination of the degree of change in the midline deviation should not be carried out, process 800 is terminated.
[00174] It should be appreciated that the 800 process is illustrative and that there are variations. For example, in some modalities, the classifi
Petition 870190046951, of 20/05/2019, p. 124/196
91/110 trained statistician can be trained, as a multitasking model, so that its output can be used not only to identify one or more locations associated with at least one midline structure of the patient's brain, but also to segment the ventricles. As described here, the measurement point to compare in the midline is found in the septum pellucidum and is therefore beneficial if using the lateral ventricle labels to train a multitasking model, since such a model will identify the location of the septum pellucidum of more accurately. The symmetry or asymmetry of the segmented lateral ventricles can help to identify the location of the septum pellucidum more precisely. Such a model can be trained if the training data includes the lateral ventricle labels in addition to the measurement point labels on the septum pellucidum and the fixation points of the falx cerebri.
[00175] The trained statistical classifier can be trained in any suitable way. In modalities, where the trained statistical classifier is a neural network, the neural network can be trained using any neural network training technique including, but not limited to, gradient reduction technique, stochastic gradient reduction, return propagation and / or any other suitable interactive optimization technique. In modalities where the neural network comprises a recurrent neural network, the training technique can employ the reduction of the stochastic gradient and the propagation of the return over time.
[00176] In some modalities, the trained statistical classifier can be trained using training data comprising labeled patient scans. For example, the classifier can be trained using training data comprising labeled patient scans, displaying the deviation in the midline (for example, stroke patients and / or patients with cannula)
Petition 870190046951, of 20/05/2019, p. 125/196
92/110 cer). Scans can be recorded manually by one or more clinical specialists. In some modalities, annotations may include indications of the locations of the fixation points of the falx cerebri and measurement points in the septum pellucidum. In some embodiments, annotations may include a line representing the midline (instead of or in addition to the location indications of falx cerebri location points). If there is no midline deviation in a particular scan, no indication of the midline (a line or attachment points) can be provided.
[00177] The inventors appreciated that there is an inherent ambiguity in the location of the measurement point. Specifically, slight deviations from the measurement point along the septum pellucidum can be tolerated, but deviations from the measurement point perpendicular to the septum pellucidum are not allowed. Accordingly, in some modalities, the training data can be increased by generating additional locations allowed for the location of the measurement point along the septum pellucidum.
[00178] As described above, the inventors have also developed low-field MRI techniques to determine if there is a change in the size of the anomaly (e.g., bleeding, injury, edema, stroke nucleus, stroke penumbra and / or swelling) in a patient's brain. In fact, MRI is an important and accurate method of detecting acute hemorrhage in patients with symptoms of acute focal stroke, and it is more accurate than CT scans to detect chronic inter-cerebral hemorrhages. Some studies have identified that MRI imaging is better than CT imaging for the detection of acute ischemia and can accurately detect acute and chronic hemorrhage. As a result, MRI can be a preferred imaging method for obtaining
Petition 870190046951, of 20/05/2019, p. 126/196
93/110 accurate agnostic of patients suspected of having an acute stroke and to monitor anomalies associated with a stroke.
[00179] Accordingly, in some modalities, low field MRI monitoring techniques can be combined with machine learning techniques to continuously monitor the size of the anomaly and detect changes in its size over time. In such modalities, low-field MRI monitoring allows a sequence of images of a patient's brain to be obtained and the machine learning techniques described here (for example, deep learning techniques, such as convolutional neural networks) can be used to determine, from the sequence of images, a corresponding sequence of anomaly sizes. For example, the deep learning techniques developed by the inventors can be used to segment (for example, identify the contours of) hemorrhages in MRI images, to identify points that specify the main geometric axes of a 2D or 3D boundary region (for example , box), identify a maximum bleeding diameter and a maximum orthogonal bleeding diameter that is orthogonal to the maximum diameter and / or perform any other processing in addition to identifying the bleeding size.
[00180] In some modalities, the volume of the anomaly can be identified using the so-called ABC / 2 formula for spherical or ellipsoid anomalies. The value A represents the length of a maximum diameter of the anomaly (for example, length of diameter 110 shown in figure 11A), the value B represents the length of a maximum orthogonal diameter of the anomaly that is orthogonal to the maximum diameter (for example, length diameter 1104 shown in figure 11 A), and the value C is the total number of slices with anomaly observed in the vertical plane multiplied by the slice thickness. The values A, B and C can then be multiplied and the product can be divided
Petition 870190046951, of 20/05/2019, p. 127/196
94/110 followed by 2 in order to estimate the volume of the anomaly. It should be appreciated that the length of the maximum diameter A and the length of the maximum orthogonal diameter B can be used to estimate the size (for example, the volume) of an anomaly in any suitable way, since the aspects of the technology described here do not are limited to that aspect.
[00181] Accordingly, in some modalities, the machine learning techniques described here can be applied to process MRI images to identify, within the MRI images, a first maximum diameter of an anomaly and a second maximum diameter. The first and second maximum diameters, in turn, can be used to estimate the size of the anomaly using the ABC / 2 technique or in any other appropriate way. For example, as illustrated in figure 11B, the machine learning techniques described here are used to identify the first diameter 11106 of an anomaly and the second diameter 1108 of the anomaly orthogonal to the first diameter. The lengths of diameters 1106 and 1108 can be used to estimate the size of the anomaly illustrated in figure 11B (a hemorrhage from the ganglia of the right intraparenchymal base).
[00182] As another example, illustrated in figure 11C, the machine learning techniques described here are used to identify the first diameter 1110 of the hematoma and the second diameter 1112 of the hematoma, orthogonal to the first diameter. The lengths of diameters 1110 and 1112 can be used to estimate the size of the hematoma illustrated in figure 11C (a right parieto-temporal intraparenchymal hematoma). As another example, illustrated in figure 11 D, the machine learning techniques described here are used to identify the first bleeding diameter 1114 and the second bleeding diameter 1116, orthogonal to the first diameter
Petition 870190046951, of 20/05/2019, p. 128/196
95/110 tro. The lengths of diameters 1114 and 1116 can be used to estimate the size of the hematoma illustrated in figure 11D (a right parieto-temporal intraparenchymal hematoma). As another example, illustrated in figure 11E, the machine learning techniques described here are used to identify the first bleeding diameter 1118 and the second bleeding diameter 1120, orthogonal to the first diameter. The lengths of diameters 1118 and 1120 can be used to estimate the size of the hemorrhage illustrated in figure 11E (intraparenchymal hemorrhage in the right parietal lobe with mild edema surrounding it). As another example, illustrated in figure 11F, the machine learning techniques described here are used to identify the first bleeding diameter 1122 and the second bleeding diameter 1124, orthogonal to the first diameter. The lengths of diameters 1122 and 1124 can be used to estimate the size of the hemorrhage illustrated in figure 11F (frontal lobe hemorrhagic contusions).
[00183] In some modalities, changes in the size of an anomaly can be monitored. The size of the anomaly (for example, a hemorrhage, an injury, an edema, a stroke nucleus, a penumbra of a stroke and / or swelling) can be monitored by identifying the size of the anomaly in a series of images taken at different times. For example, as illustrated in figure 12A, the size of the bleeding in a first MRI image, obtained at first, can be determined based on the lengths of diameters 1202 and 1204, which are identified using machine learning techniques described here (for example, using a neural network having an architecture illustrated in figure 14 or figure 15). As shown in figure 12B, the size of the hemorrhage in a second MRI image obtained in a second moment (occurring at least a limit amount of
Petition 870190046951, of 20/05/2019, p. 129/196
96/110 time after the first moment) can be determined based on the lengths of diameters 1206 and 1208, which are also identified using the machine learning techniques described here. Comparing the lengths of the diameters (and / or the bleeding sizes derived from them), as illustrated in figure 12C, allows you to determine whether the bleeding size has changed (for example, has it been smaller or larger ) And, if so the case, the difference by which the size was changed.
[00184] Figure 13 is a flowchart of an illustrative process 1300 to determine a degree of change in the size of an anomaly (for example, hemorrhage, injury, edema, stroke nucleus, stroke penumbra and / or swelling) in a patient's brain, according to some modalities of the technology described here. In some embodiments, the entire 1300 process can be performed while the patient is inside a low field MRI device, which can be of any suitable type described here including, for example, any of the low field MRI devices illustrated in figures 3A to 3G). Although, for the sake of clarity, the 1300 process is described with respect to detecting a change in the size of a hemorrhage, it should be appreciated that the 1300 process can be applied to detect changes in the size of any suitable type of anomaly ( for example, a hemorrhage, an injury, an edema, a stroke nucleus, a penumbra of a stroke and / or swelling), since the aspects of the technology described here are not limited to this aspect. Similarly, the neural network architectures described in figures 14 and 15 can be applied to detect changes in the size of any suitable type of anomaly, however, they are not limited, however, to be applied exclusively to detect changes in the size of hemorrhage.
[00185] Process 1300 starts at act 1302, where the device
Petition 870190046951, of 20/05/2019, p. 130/196
97/110 low-field MRI acquires initial MRI data from a target part of the patient's brain. As described here, the term MR image data is used here to refer to MR data generically including, but not limited to, MR data prior to image reconstruction (for example, k-space MR data) and RMs that have been processed in some way (for example, post-image reconstruction RM data such as three-dimensional (3D) volumetric image). In some embodiments, the initial MR data may include one or more two-dimensional images of the respective brain slices (for example, two, three, four, five, etc. neighboring slices). When multiple slices are included, the slices can be neighbors.
[00186] Then, in act 1304, the initial MR image data is provided as input to a trained statistical classifier in order to obtain the corresponding initial output. In some modalities, before being provided to the trained statistical classifier, the initial MR image data can be pre-processed, for example, by resampling, interpolation, related transformation, and / or by using any other pre-processing techniques appropriate, since the aspects of the technology described here are not limited to that aspect.
[00187] In some modalities, the output of the trained statistical classifier can be used to identify, in act 1306, initial values of the characteristics indicative of the size of a hemorrhage in the patient's brain. In some embodiments, the features may have a first maximum bleeding diameter in a first direction and a second maximum bleeding diameter in a second direction, which is orthogonal to the first direction. The values can indicate the initial lengths of the diameters and / or the initial end points of the diameters (from where the initial lengths can
Petition 870190046951, of 20/05/2019, p. 131/196
98/110 be derived). In some embodiments, the characteristics may be the corners of a boundary box that limits the perimeter of the bleeding and the initial values may be local to the corners. In some modalities, the characteristics may specify the bleeding limit and the initial values may be the locations of one or more points along the segmented limit. The output of the trained statistical classifier can specify the initial values directly or indirectly. In the latter case, the values can be derived from the information included in the output of the trained statistical classifier.
[00188] In some embodiments, the initial values of the characteristics obtained in act 1306 can be used to obtain an initial estimate of the size of the hemorrhage. For example, when initial values can be used to determine the initial lengths of the maximum orthogonal bleeding diameters, the initial lengths can be used to estimate the initial bleeding volume (for example, according to method ABC / 2 described above) . As another example, when the initial values specify the limit of a hemorrhage, the limit information can be used to estimate the initial bleeding area on the slice (for example, using a polygonal approximation or in any other suitable way).
[00189] In some modalities, the trained statistical classifier may be a neural network statistical classifier. For example, the statistical training classifier may include a fully convolutional neural network (for example, as illustrated in figures 10 and 14) or a convolutional neural network (for example, as illustrated in figures 9A to 9C and 15), and / or any other suitable type of neural network. The trained statistical classifier can be implemented in software, in hardware, or using any suitable combination of software and hardware. In some modalities, one or more bibli
Petition 870190046951, of 20/05/2019, p. 132/196
99/110 machine learning software loops can be used to implement the trained statistical classifier including, but not limited to Theano, Torch, Caffe, Keras and TensorFlow. These libraries can be used to train a statistical classifier (for example, a neural network) and / or use a trained statistical classifier. The trained statistical classifier can be trained using any suitable training technique including any neural network training technique (for example, gradient reduction) described above. It should also be appreciated that the trained statistical classifier is not limited to being a neural network and can be any other suitable type of statistical classifier (for example, a support vector machine, a graphic model, a Bayesian classifier, a classifier of decision tree, etc.), since aspects of the technology described here are not limited to that aspect.
[00190] In some modalities, the trained statistical classifier can be one of the neural networks described above with reference to figures 9A to 90 or figure 10. Such trained statistical classifier can identify point locations in the MRI image data. For example, such a trained statistical classifier can be used to identify the locations of the end points of the first and second orthogonal diameters of a hemorrhage. As another example, such a trained statistical classifier can be used to identify the locations of the corners of a hemorrhage boundary box.
[00191] In other modalities, the trained statistical classifier can be a totally convolutional neural network having an architecture as illustrated in figure 14. Such trained statistical classifier can be used to identify the limits of hemorrhage. Training, like a neural network, can involve images of
Petition 870190046951, of 20/05/2019, p. 133/196
100/110 training with zero filling, using size 3 and step 1 convolute cores, using a maximum set core with a size equal to 2, and deconvolution cores (upscale and convolution) with size equal to 6 and size equal to 2 The output of the neural network can identify the limit of bleeding.
[00192] In other additional modalities, the trained statistical classifier can be a convolutional neural network having an architecture as illustrated in figure 15. Such trained statistical classifier can be used to identify the limits of hemorrhage by classifying individual voxels, an approach that has the advantage of having a greater invariance at the injury site. The neural network uses convoluted nuclei with size equal to 5 and step 1 in the first layer and nuclei with size equal to 3 in the subsequent layers. This building block can be repeated for different sizes of entrance neighborhood (25 as illustrated, 20, 15 or more, 30, 35). Larger neighborhoods use a larger initial core size (for example, 7). Feature maps are mixed in the last feature layer and combined to result in a single forecast.
[00193] It should be appreciated that the neural network architectures illustrated in figures 14 and 15 are illustrative and that variations of these architectures are possible. For example, one or more other layers of neural network (for example, a convolutional layer, a convolutional layer, a rectified linear unit layer, an ascending sampling layer, a concatenated layer, a filler layer, etc.) introduced in any of the neural network architectures of figures 14 and 15, as one or more additional layers and / or instead of one or more layer parts of the illustrated architectures. As another example, the dimensionality of one or more layers may vary and / or the core size for
Petition 870190046951, of 20/05/2019, p. 134/196
101/110 one or more convolutional, joint and / or convolutional layers may vary.
[00194] In some modalities, when multiple slices are being processed by the neural network, convolutions can be two-dimensional (2D) or three-dimensional (3D) convolutions. In some modalities, the processing can be based on slice, so that the characteristics are calculated for each slice using the slice information and one or more of its neighboring slices (the slice itself or the slice itself and one or more of your neighboring slices). In other modalities, processing can be a fully 3D processing channel, so that the characteristics for multiple slices are computed simultaneously using data present in all slices.
[00195] Next, process 1300 proceeds to act 1308, where the next MR image data is acquired. The next MR image data is acquired after the initial MR data is acquired. Thus, although in some modalities, acts 1304 and 1306 are performed after act 1308 is performed, act 1308 is generally performed after act 1302. The following MR image data can be acquired immediately after the acquisition of the initial MR image data, or can be obtained after a desired delay period (for example, in 1, 2, 3, 4, 5, 10, 15, 20 minutes, in an hour, in two hours, etc. ). As with the initial MR image data, the next MR image data can take any shape (for example, a 3D volumetric image, a 2D image, k-space MR data, etc.). In some embodiments, the initial MR data and the next MR image data are of the same type. For example, each of the initial and close MR data can include one or more two-dimensional images of one or more slices of brain
Petition 870190046951, of 20/05/2019, p. 135/196
102/110 bro respective (for example, neighbors). For example, the initial MR data can include multiple images of neighboring slices taken at a first time and the next MR data can include multiple images of the same neighboring slices obtained at a second time after the first moment.
[00196] Next, process 1300 continues in act 1310 where the next MR image data is provided as input to the trained statistical classifier to obtain the next corresponding output. In some modalities, before being provided to the trained statistical classifier, the next MR image data can be pre-processed, for example, by resampling, interpellation, related transformation, and / or using any other pre-processing technique appropriate, since the aspects of the technology described here are not limited to this aspect. The next MR image data can be pre-processed in the same way as the initial MR data was pre-processed.
[00197] In some modalities, the output of the trained statistical classifier can be used to identify, in act 1312, the updated values of the characteristics indicative of the size of a hemorrhage in the patient's brain. In some embodiments, the characteristics may be a first maximum bleeding diameter in a first direction and a maximum bleeding diameter in a second direction, which is orthogonal to the first direction. The updated values can indicate the updated lengths of the diameters and / or end points of the diameters (from which the lengths can be derived). In some embodiments, the characteristics may be corners of a boundary box that limits the perimeter of the bleed and the updated values may be updated locations of the corners. In some modalities, the characteristics can specify the bleeding limit and the updated values can be current locations.
Petition 870190046951, of 20/05/2019, p. 136/196
103/110 of one or more points along the segmented limit. The output of the trained statistical classifier can specify the updated values, directly or indirectly. In the latter case, the values can be derived from the information included in the output of the trained statistical classifier.
[00198] In some modalities, the updated values of the characteristics obtained in act 1306 can be used to obtain an updated estimate of the size of the hemorrhage. For example, when the updated values can be used to determine the updated lengths of the maximum orthogonal bleeding diameters, the updated lengths can be used to estimate the bleeding volume (for example, according to the ABC / 2 method described above). As another example, when the updated values specify the limit of a hemorrhage, the limit information can be used to estimate the updated bleeding area on the slice.
[00199] Next, process 1300 proceeds to act 1314, where it is determined whether the size of the bleeding has been changed and, if so, how much. The determination can be made using the initial and updated values obtained in acts 1306 and 1312, respectively. For example, in some embodiments, the initial values obtained in act 1306 can be used to obtain an initial estimate of the size (eg volume, area, etc.) for bleeding and the updated values obtained in act 1312 can be used to get an updated size estimate. In turn, initial and updated size estimates can be used to determine whether the size of the bleeding has changed (for example, by assessing its difference) and, if so, by how much.
[00200] Next, process 1300 proceeds to decision block 1316, where it is determined whether it is necessary to continue monitoring the
Petition 870190046951, of 20/05/2019, p. 137/196
104/110 size of the bleeding for any changes. This determination can be made in any appropriate way (for example, by determining whether a limit number of interactions has been made, based on a schedule, based on manual input provided by a doctor, etc.), since aspects of technology described here are not limited in that regard. When it is determined that monitoring should continue, process 1300 returns to block 1308 and acts 1308-1314 are repeated again (with the newly obtained MR data being compared with the most recently obtained MR data). On the other hand, when it is determined that monitoring does not need to continue, process 1300 is terminated.
[00201] Figure 16 is a diagram of an illustrative computer system in which the modalities described here can be implemented. An illustrative implementation of a computer system 1600 that can be used with respect to any of the modalities of the description provided here, is illustrated in figure 16. For example, the processes described with reference to figures 8 and 13 can be implemented in and / or using the computer system 1600. The computer system 1600 may include one or more 1610 processors and one or more articles of manufacture comprising the non-transitory computer-readable storage medium (for example, 1620 memory and one or more media) 1630 non-volatile storage). Processor 1610 can control data writing to and reading data from memory 1620 and non-volatile storage device 1630 in any suitable manner, since aspects of the description provided here are not limited in this regard. To perform any functionality described here, the 1610 processor can execute one or more executable instructions per processor stored in one or more storage media
Petition 870190046951, of 20/05/2019, p. 138/196
105/110 non-transitory computer-readable (eg, 1620 memory), which can serve as a non-transitory computer-readable storage medium by storing executable instructions per processor for execution by the 1610 processor.
[00202] Having thus described the various aspects and modalities of the technology presented in the description, it should be appreciated that various changes, modifications and improvements will occur promptly to those skilled in the art. Such changes, modifications and improvements must be contained in the spirit and scope of the technology described here. For example, those skilled in the art will readily envision a variety of other means and / or structures to perform the function and / or obtain the results and / or one or more of the advantages described here, and each of said variations and / or modifications is considered included in the scope of the modalities described here. Those skilled in the art will recognize, or be able to determine, using no more than routine experimentation, many equivalences to the specific modalities described here. It must therefore be understood that the above modalities are presented by way of example only and that, within the scope of the appended claims and their equivalences, the inventive modalities can be practiced in another way, in addition to the specifically described. In addition, any combination of two or more characteristics, systems, articles, materials, kits and / or methods described here, if such characteristics, systems, articles, materials, kits and / or methods are not mutually inconsistent, is included in the scope of this description.
[00203] The modalities described above can be implemented in any of the various ways. One or more aspects and modalities of the present description, involving the performance of processes or methods, may use program instructions executable by a device (for example, a computer, a process
Petition 870190046951, of 20/05/2019, p. 139/196
106/110 or other device) to perform, or control the performance of processes or methods. In this regard, several inventive concepts can be embodied as a computer-readable storage medium (or multiple computer-readable storage media) (for example, a computer memory, one or more diskettes, compact discs, optical discs, magnetic tapes , flash memories, circuit configurations in Field Programmable Port Sets or other semiconductor devices or other tangible computer storage media) encoded with one or more programs that, when run on one or more computers or other processors, perform the methods that implement one or more of the various modalities described above. The computer-readable medium or media can be transportable, so that the program or programs implement various aspects described above. In some embodiments, the computer-readable medium may be non-transitory.
[00204] The terms program or software are used here in a generic sense to refer to any type of code or computer set of computer executable instructions that can be used to program a computer or other processor to implement various aspects as described above . Additionally, it should be appreciated that, according to one aspect, one or more computer programs that, when executed perform the methods of the present description, do not need to reside in a single computer or processor, but can be distributed in a modular way among several different computers or processors to implement various aspects of this description.
[00205] Computer executable instructions can be in many forms, such as program modules, executed by one or more computers or other devices. Generally, the
Petition 870190046951, of 20/05/2019, p. 140/196
107/110 programs include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement the particular abstract data types. Typically, the functionality of the program modules can be combined or distributed as desired in various modalities.
[00206] In addition, data structures can be stored in a computer-readable medium in any suitable form. For the sake of simplicity of illustration, data structures can be illustrated as having fields that are related by location in the data structure. Such a relationship can likewise be achieved by designating storage for fields with locations in a computer-readable medium that creates the relationship between fields. However, any suitable mechanism can be used to establish a relationship between information in the fields of a data structure, including through the use of pointers, tags or other mechanisms that establish the relationship between data elements.
[00207] When implemented in software, the software code can be executed on any suitable processor or collection of processors, whether provided on a single computer or distributed among multiple computers.
[00208] Additionally, it should be appreciated that a computer can be embodied in any one of several ways, such as a rack-mounted computer, a desktop computer, a laptop computer, or a tablet computer, as non-limiting examples. In addition, a computer can be embedded in a device not generally considered to be a computer, but with the appropriate processing capabilities, including a Personal Digital Assistant (PDA), a smartphone or any other suitable portable or fixed electronic device.
Petition 870190046951, of 20/05/2019, p. 141/196
108/110 [00209] In addition, a computer can have one or more input and output devices. These devices can be used, among other things, to present a user interface. Examples of output devices that can be used to provide a user interface include printers or display screens for visual output presentation and speakers or other sound generation devices for audible output presentation. Examples of input devices that can be used for a user interface include keyboards, and pointing devices, such as a mouse, touchpad and scanning tablets. As another example, a computer can receive input information through speech recognition or in other audible formats.
[00210] Such computers can be interconnected by one or more networks in any suitable form, including a local area network or a wide area network, such as a corporate network, and smart network (IN) or the Internet. Such networks may be based on any suitable technology and may operate according to any suitable protocol and may include wireless networks, wired networks or fiber optic networks.
[00211] Furthermore, as described, some aspects can be embodied as one or more methods. The acts performed as part of the method can be ordered in any appropriate way. Accordingly, the modalities can be constructed in which the acts are performed in a different order to that illustrated, which can include performing some acts simultaneously, although illustrated as sequential acts in the illustrative modalities.
[00212] All definitions, as defined and used here, must be understood as prevailing over dictionary definitions, definitions in documents incorporated by reference, and / or normal meanings of defined terms.
Petition 870190046951, of 20/05/2019, p. 142/196
109/110 [00213] The indefinite articles one, one, as used herein in the specification and in the claims, unless clearly indicated to the contrary, are to be understood to mean at least one.
[00214] The phrase and / or, as used here in the specification or in the claims, must be understood as meaning either one or both of the elements joined in this way, that is, elements that are present together in some cases and separately present in other cases . The multiple elements listed with and / or should be considered in the same way, that is, one or more of the elements joined in this way. Other elements may, optionally, be present in addition to the elements specifically identified by the clause and / or, whether it is related or not to the elements specifically identified. Thus, as a non-limiting example, a reference to A and / or B, when used in conjunction with the broad-meaning language as comprising, may refer, in one embodiment, to A only (optionally including elements in addition to B) ; in another modality, B only (optionally including elements in addition to A); and in yet another modality, both A and B (optionally including other elements); etc.
[00215] As used here in the specification and in the claims, the phrase at least one with reference to a list of one or more elements, must be understood as meaning at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed in the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows the elements to be optionally present in addition to the specified elements.
Petition 870190046951, of 20/05/2019, p. 143/196
110/110 identified within the list of elements to which the phrase at least one refers, related or not to those specifically identified elements. Thus, a non-limiting example, at least one of A and B ”(or, equivalently, at least one of A or B, or, equivalently, at least one of A and / or B) may refer , in one embodiment, to at least one, optionally including more than one, A, without any B present (and, optionally, including elements other than B); in another embodiment, at least one, optionally including more than one, B, without any A present (and, optionally, including elements other than A); in another additional embodiment, at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and, optionally, including other elements); etc.
[00216] In addition, the phraseology and terminology used here are for description purposes and should not be considered limiting. The use of including, comprising or possessing, containing, involving and variations thereof, must encompass the items listed hereinafter and their equivalences in addition to additional items.
[00217] In the claims, in addition to the above specification, all transition phrases as comprising, including, bearing, possessing, containing, involving, maintaining, composed of and the like should be understood as comprehensive, that is, including, but not limited to limited to. Only traditional phrases consisting of and consisting essentially of must be closed or semi-closed transition phrases, respectively.

权利要求:
Claims (68)
[1]
1. Method of detecting change in the degree of midline deviation in a patient's brain positioned inside a low field magnetic resonance imaging (MRI) device, the method characterized by the fact that it comprises:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance imaging (MR) data from the patient's brain;
provide the first MR data as input to a trained statistical classifier to obtain the first corresponding output;
identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain;
acquiring the second MR imaging data for the patient's brain following the acquisition of the first MR imaging data;
providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output;
identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location of at least one landmark and at least one updated location of at least one landmark.
[2]
2. Method, according to claim 1, characterized by the fact that, from the first exit, it identifies at least one
Petition 870190046951, of 20/05/2019, p. 145/196
2/22 initial location of at least one landmark associated with at least one midline structure of the patient's brain, comprising:
identify an initial location of an anterior fixation point for a falx cerebri;
identify an initial location of a posterior fixation point for the falx cerebri; and identifying an initial location of a measurement point in a septum pellucidum.
[3]
3. Method, according to claim 2, characterized by the fact that it determines an initial amount of midline deviation, using the initial identified locations of the anterior fixation point of the falx cerebri, the posterior fixation point of the falx cerebri, and the measuring point at septum pellucidum.
[4]
4. Method, according to claim 2, characterized by the fact that, starting from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain, comprises:
identify an updated location of the anterior fixation point of the falx cerebri;
identify an updated location of the posterior fixation point of the falx cerebri; and identify an updated location of the measurement point in the septum pellucidum.
[5]
5. Method, according to claim 4, characterized by the fact that the degree of change in the midline deviation is determined using the initial and updated locations identified from the previous fixation point of the falx cerebri, the fixation point posterior view of the falx cerebri and the measuring point at the septum pel
Petition 870190046951, of 20/05/2019, p. 146/196
3/22 lucidum.
[6]
6. Method, according to claim 5, characterized by the fact of determining the degree of change in the midline deviation, comprising:
determine an initial amount of midline deviation using the identified initial locations of the anterior fixation point of the falx cerebri, the posterior fixation point of the falx cerebri, and the measurement point in the septum pellucidum;
determine an updated amount of midline deviation using the updated locations identified from the anterior fixation point of the falx cerebri, the posterior fixation point of the falx cerebri, and the measurement point in the septum pellucidum; and determining the degree of change in the midline deviation using the determined initial and updated quantities of the midline deviation.
[7]
7. Method, according to claim 1, characterized by the fact that the trained statistical classifier comprises a multilayered neural network.
[8]
8. Method, according to claim 1, characterized by the fact that the trained statistical classifier comprises a convolutional neural network.
[9]
9. Method, according to claim 1, characterized by the fact that the trained statistical classifier comprises a fully convolutional neural network.
[10]
10. Method, according to claim 1, characterized by the fact that the trained statistical classifier comprises a convolutional neural network and a recurrent neural network.
[11]
11. Method according to claim 10, characterized in that the recurrent neural network comprises a neural network of long-term short-term memory.
Petition 870190046951, of 20/05/2019, p. 147/196
4/22
[12]
12. Method according to claim 1, characterized in that the second MR image data is obtained within one hour from the first MR image data.
[13]
13. Method according to claim 1, characterized in that it additionally comprises repeating the acquisition of MR image data to obtain a sequence of MR image data frames.
[14]
14. Method, according to claim 1, characterized by the fact that the frame sequence is acquired over a period of time greater than one hour, while the patient remains positioned inside the low field magnetic resonance imaging device.
[15]
15. Method, according to claim 13, characterized by the fact that the sequence of frames is acquired over a period of time greater than two hours, while the patient remains positioned inside the low field magnetic resonance imaging device.
[16]
16. Method, according to claim 13, characterized by the fact that the sequence of frames is acquired over a period of time greater than five hours, while the patient remains positioned inside the low field magnetic resonance imaging device.
[17]
17. Low-field magnetic resonance imaging device, characterized by the fact that it is configured to detect changes in the degree of midline deviation in the brain of a patient positioned inside a magnetic resonance imaging (MRI) device low field, the low field MRI device comprising:
a plurality of magnetic components, including:
a BO magnet configured to produce, at least in
Petition 870190046951, of 20/05/2019, p. 148/196
5/22 part, a BO magnetic field;
at least one gradient magnet configured to spatially encode MRI data; and at least one radio frequency coil configured to stimulate a magnetic resonance response and detect magnetic components configured to, when operated, acquire magnetic resonance image data; and at least one controller configured to operate the plurality of magnet components so that, while the patient remains positioned within the low field magnetic resonance device, acquire the first magnetic resonance imaging (MR) data from the patient's brain, and acquire second MR image data for the part of the patient's brain following the acquisition of the first MR image data;
where at least one controller is additionally configured to perform:
providing the first and second MR data as input to a trained statistical classifier to obtain corresponding first and second outputs;
identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain;
identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location of at least one landmark and at least one updated location of at least one landmark.
[18]
18. At least one readable storage medium per
Petition 870190046951, of 20/05/2019, p. 149/196
6/22 non-transitory computer storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform a method of detecting change in the degree of line deviation average in the brain of a patient positioned inside a low field magnetic resonance imaging (MRI) device, the method characterized by the fact that it comprises:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance imaging (MR) data from the patient's brain;
provide the first MR data as input to a trained statistical classifier to obtain the first corresponding output;
identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain;
acquiring second MR image data from the patient's brain following the acquisition of the first MR image data;
providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output;
identify, from the second exit, at least one updated location of at least one landmark associated with at least one midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location of at least one landmark and at least one updated location of at least
Petition 870190046951, of 20/05/2019, p. 150/196
7/22 a milestone.
[19]
19. System, characterized by the fact of understanding:
at least one computer hardware processor; and at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform a method to detect the change the degree of deviation from the midline in a patient's brain positioned within a low-field magnetic resonance imaging (MRI) device, the method characterized by the fact that it understands while the patient remains positioned within the low-level MRI device field:
acquire the first magnetic resonance imaging (MR) data from the patient's brain;
provide the first MR data as input to a trained statistical classifier to obtain the first corresponding output;
identify, from the first exit, at least one initial location of at least one landmark associated with at least one midline structure of the patient's brain;
acquiring the second MR image data from the patient's brain following the acquisition of the first MR image data;
providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output;
identify, from the second exit, at least one updated location of at least one landmark associated with at least mePetition 870190046951, of 20/05/2019, p. 151/196
8/22 us a midline structure of the patient's brain; and determining a degree of change in the midline deviation using at least one starting location of at least one landmark and at least one updated location of at least one landmark.
[20]
20. Method for determining the change in size of an anomaly in a patient's brain positioned inside a low-field magnetic resonance (MRI) imaging device, characterized by the fact that it comprises:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance imaging (MR) data from a patient's brain;
provide the first MR image data as input to a trained statistical classifier to obtain the first corresponding output;
identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain;
acquiring the second MR image data from the patient's brain following the acquisition of the first MR image data;
providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output;
identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain;
determine the change in the size of the anomaly using at least one initial value of at least one characteristic
Petition 870190046951, of 20/05/2019, p. 152/196
9/22 and the at least one updated value of at least one feature.
[21]
21. Method, according to claim 20, characterized by the fact that, using the first output, it identifies the at least one initial value of the at least one characteristic indicative of the size of the brain anomaly in the patient, comprising:
identify a region in the MR image data including the anomaly.
[22]
22. Method, according to claim 20, characterized by the fact that, using the first output, it identifies at least one initial value of at least one characteristic indicative of the size of the anomaly in the patient's brain, comprising:
identify one or more first values indicative of a first anomaly diameter.
[23]
23. Method according to claim 22, characterized by the fact that it additionally comprises identifying one or more second values indicative of a second diameter of anomaly orthogonal to the first diameter.
[24]
24. Method according to claim 20, characterized in that it determines the change in the size of the anomaly using at least one initial value of the at least one characteristic and at least one updated value of the at least one characteristic comprising:
determine an initial size of the anomaly using at least one value of at least one characteristic;
determine an updated anomaly size using at least one updated value of at least one characteristic; and determine the change in the size of the anomaly using the initial and updated sizes determined for the anomaly.
[25]
25. Method, according to claim 20, characterized by the fact that the trained statistical classifier comprises
Petition 870190046951, of 20/05/2019, p. 153/196
10/22 a multilayered neural network.
[26]
26. Method, according to claim 20, characterized by the fact that the trained statistical classifier comprises a convolutional neural network.
[27]
27. Method, according to claim 20, characterized by the fact that the trained statistical classifier comprises a fully convolutional neural network.
[28]
28. Method according to claim 20, characterized in that the second MR image data is obtained within one hour from the first MR image data.
[29]
29. Method according to claim 20, characterized in that it further comprises repeating the acquisition of MR image data to obtain a sequence of MR image data frames.
[30]
30. Method according to claim 20, characterized in that the frame sequence is acquired over a period of time greater than one hour, while the patient remains positioned within the low-field magnetic resonance imaging device.
[31]
31. Method, according to claim 30, characterized by the fact that the sequence of frames is acquired over a period of time greater than two hours, while the patient remains positioned inside the low field magnetic resonance imaging device.
[32]
32. Method according to claim 30, characterized in that the sequence of frames is acquired over a period of time greater than five hours, while the patient remains positioned within the low-field magnetic resonance imaging device.
Petition 870190046951, of 20/05/2019, p. 154/196
11/22
[33]
33. Method according to claim 20, characterized in that the anomaly comprises hemorrhage.
[34]
34. Method, according to claim 20, characterized in that the anomaly comprises hemorrhage, injury, edema, a stroke nucleus, a stroke and / or swelling penumbra.
[35]
35. Low-field magnetic resonance imaging (MRI) device, configured to determine a change in the size of an anomaly in a patient's brain, the low-field MRI device being characterized by the fact that it comprises:
a plurality of magnetic components, including:
a magnet B0 configured to produce, at least in part, a magnetic field B0;
at least one gradient magnet configured to spatially encode MRI data; and at least one radio frequency coil configured to stimulate a magnetic resonance response and detect magnetic components configured to, when operated, acquire magnetic resonance image data; and at least one controller configured to operate the plurality of magnet components so that, while the patient remains positioned within the low field magnetic resonance device, acquire the first magnetic resonance imaging (MR) data from the patient's brain, and acquire the second MR image data from the patient's brain following the acquisition of the first MR image data;
where at least one controller is additionally configured to perform:
provide the first and second image data of
Petition 870190046951, of 20/05/2019, p. 155/196
12/22
MR as input to a statistical classifier trained to obtain the corresponding first and second outputs;
identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain;
acquiring the second MR imaging data for the part of the patient's brain subsequent to the acquisition of the first MR imaging data;
identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain;
determine the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[36]
36. At least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform the method of determining change in the size of an anomaly in a patient's brain positioned inside a low-field magnetic resonance (MRI) imaging device, characterized by the fact that it comprises:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance imaging (MR) data from the patient's brain;
provide the first MR image data as input to a trained statistical classifier to obtain the first corresponding output;
Petition 870190046951, of 20/05/2019, p. 156/196
13/22 identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain;
acquiring the second MR image data from the patient's brain following the acquisition of the first MR image data;
providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output;
identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain;
determine the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[37]
37. System, characterized by the fact that it comprises:
at least one computer hardware processor;
at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one hardware processor to perform the method of determining the change in size of a anomaly in a patient's brain positioned inside a low field magnetic resonance imaging (MRI) device, the method comprising:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance imaging (MR) data from the patient's brain;
provide the first MR image data as
Petition 870190046951, of 20/05/2019, p. 157/196
14/22 registered in a statistical classifier trained to obtain the first corresponding output;
identify, using the first output, at least an initial value of at least one characteristic indicative of a size of an abnormality in the patient's brain;
acquiring the second MR image data from the patient's brain following the acquisition of the first MR image data;
providing the second MR image data as input to the trained statistical classifier to obtain the corresponding second output;
Identify, using the second output, at least one updated value of at least one characteristic indicative of the size of the anomaly in the patient's brain; and determining the change in the size of the anomaly using at least one initial value of at least one characteristic and at least one updated value of at least one characteristic.
[38]
38. Method of detecting change in the biological matter of a patient positioned inside a low-field magnetic resonance imaging (MRI) device, the method characterized by the fact that it comprises:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance image data from a part of the patient;
acquire the second MRI data from the patient following the acquisition of the first MRI data;
align the first MRI image data and the second MRI image data
Petition 870190046951, of 20/05/2019, p. 158/196
15/22 ca; and compare the first MRI image data and the second MRI image data to detect at least one change in biological matter on the part of the patient.
[39]
39. Method according to claim 38, characterized in that it additionally comprises modifying at least one acquisition parameter, based on at least one change in biological matter on the part of the patient.
[40]
40. Method, according to claim 39, characterized in that it additionally comprises acquiring the third magnetic resonance image data from the patient using at least one modified acquisition parameter.
[41]
41. Method, according to claim 40, characterized in that the at least one acquisition parameter is modified in order to change at least one among resolution, signal-to-noise ratio and field of view of the third resonance image data magnetic.
[42]
42. Method according to claim 38, characterized in that it additionally comprises repeating the acquisition of the magnetic resonance image data to obtain a sequence of magnetic resonance image data frames.
[43]
43. The method of claim 42, characterized in that the frame sequence is acquired over a period of time greater than one hour, while the patient remains positioned within the low-field magnetic resonance imaging device.
[44]
44. Method according to claim 42, characterized in that the sequence of frames is acquired over a period of time greater than two hours while the patient remains
Petition 870190046951, of 20/05/2019, p. 159/196
16/22 is positioned inside the low field magnetic resonance imaging device.
[45]
45. Method according to claim 42, characterized in that the frame sequence is acquired over a period of time greater than five hours, while the patient remains positioned within the low field magnetic resonance imaging device.
[46]
46. Method according to claim 42, characterized by the fact that it further comprises:
align at least two of the frame strings; and comparing at least two frames aligned to detect at least one change in biological matter on the part of the patient.
[47]
47. Method according to claim 46, characterized in that at least one change is used to compute a change in volume and / or quantity of biological matter between the at least two frames in the frame sequence.
[48]
48. Method according to claim 46, characterized in that a first frame in the sequence of frames corresponds to the magnetic resonance image data of a first region on the part of the patient and a second frame in the sequence of frames corresponds to the data of magnetic resonance image of a sub-region of the first region.
[49]
49. Method, according to claim 48, characterized by the fact that the subregion is selected based on where changes in biological matter were detected.
[50]
50. Method according to claim 47, characterized in that it additionally comprises detecting a rate of change of biological matter on the part of the patient.
[51]
51. Method according to claim 38, characterized by the fact that the first resonance image data is compared
Petition 870190046951, of 20/05/2019, p. 160/196
17/22 magnetic resonance and the second magnetic resonance image data aligned to detect at least one change in biological matter on the part of the patient, comprising:
provide the first magnetic resonance image data aligned as input to a trained statistical classifier to obtain the first corresponding output; and providing the second aligned MRI image data recorded in the statistical classifier to obtain the corresponding second output.
[52]
52. Method according to claim 51, characterized in that the comparison additionally comprises:
determine a change in the size of a hemorrhage using the first outlet and the second outlet.
[53]
53. Method, according to claim 51, characterized in that the comparison additionally comprises:
determine a degree of change in a midline deviation using the first exit and the second exit.
[54]
54. Low-field magnetic resonance imaging device configured to detect the change in the biological matter of a patient positioned within the low-field magnetic resonance imaging device, characterized by the fact that it comprises:
a plurality of magnetic components, including:
a BO magnet configured to produce, at least in part, a BO magnetic field;
at least one gradient magnet configured to spatially encode MRI data; and at least one radio frequency coil configured to stimulate an MRI response and detect magnetic components configured to, when operated, acquire
Petition 870190046951, of 20/05/2019, p. 161/196
18/22 laugh MRI image data; and at least one controller configured to operate the plurality of magnet components so that, while the patient remains positioned within the low field MRI device, acquiring the first MRI data from a part of the patient, and acquiring the second magnetic resonance imaging data from the patient following the acquisition of the first magnetic resonance imaging data, the at least one controller additionally configured to align the first magnetic resonance imaging data and the second magnetic resonance imaging data, and compare the first magnetic resonance imaging data and the second magnetic resonance imaging data aligned to detect at least one change in biological matter on the part of the patient.
[55]
55. Low-field magnetic resonance imaging device according to claim 54, characterized in that the at least one controller is configured to modify at least one acquisition parameter based on at least one change in biological matter from the patient.
[56]
56. Low-field magnetic resonance imaging device according to claim 54, characterized in that a controller is configured to acquire the third magnetic resonance image data from the patient using at least one parameter of modified acquisition.
[57]
57. Low-field magnetic resonance imaging device according to claim 56, characterized by the fact that the at least one acquisition parameter is modified in order to change at least one among resolution, signal to noise ratio and field of view of third magnetic resonance image data.
Petition 870190046951, of 20/05/2019, p. 162/196
19/22
[58]
58. Low field magnetic resonance imaging device according to claim 54, characterized in that the at least one controller is configured to acquire magnetic resonance image data to obtain a sequence of image data frames magnetic resonance imaging.
[59]
59. Low-field magnetic resonance imaging device according to claim 58, characterized in that the frame sequence is acquired over a period of time over an hour while the patient remains positioned within the imaging device low-field magnetic resonance imaging.
[60]
60. Low-field magnetic resonance imaging device according to claim 58, characterized in that the frame sequence is acquired over a period of time greater than two hours while the patient remains positioned within the imaging device low-field magnetic resonance imaging.
[61]
61. Low-field magnetic resonance imaging device according to claim 58, characterized in that the frame sequence is acquired over a period of time greater than five hours, while the patient remains positioned within the imaging device. low field magnetic resonance imaging.
[62]
62. Low-field magnetic resonance imaging device according to claim 58, characterized in that the at least one controller is configured to:
align at least two of the frame strings; and compare the at least two tables aligned with at least
Petition 870190046951, of 20/05/2019, p. 163/196
20/22 minus a change in biological matter on the part of the patient.
[63]
63. Low-field magnetic resonance imaging device according to claim 62, characterized in that the at least one change is used to compute a change in volume and / or quantity of biological matter between at least two frames following frames.
[64]
64. Low-field magnetic resonance imaging device according to claim 62, characterized in that a first frame in the frame sequence corresponds to the magnetic resonance image data from a first region on the patient's part and a second frame in the frame sequence corresponds to the magnetic resonance image data of a sub-region of the first region.
[65]
65. Low-field magnetic resonance imaging device, according to claim 64, characterized by the fact that the subregion is selected based on where changes in biological matter are detected.
[66]
66. Low-field magnetic resonance imaging device according to claim 62, characterized in that the at least one controller is configured to detect a rate of change of biological matter on the part of the patient.
[67]
67. At least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform a method of detecting change in the biological matter of a patient positioned within a low field magnetic resonance imaging (MRI) device, the method characterized by
Petition 870190046951, of 20/05/2019, p. 164/196
21/22 address:
while the patient remains positioned inside the low field magnetic resonance imaging (MRI) device:
acquire the first magnetic resonance image data from a part of the patient;
acquire the second MRI data from the patient following the acquisition of the first MRI data;
align the first magnetic resonance image data and the second magnetic resonance image data; and comparing the first magnetic resonance imaging data and the second magnetic resonance imaging data aligned to detect at least one change in biological matter on the part of the patient.
[68]
68. System, characterized by the fact that it comprises:
at least one computer hardware processor; and at least one non-transitory computer-readable storage medium storing executable instructions per processor that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform a change detection method in the biological matter of a patient positioned inside a low field magnetic resonance imaging (MRI) device, the method comprising:
while the patient remains positioned inside the low field MRI device:
acquire the first magnetic resonance image data from a part of the patient;
Petition 870190046951, of 20/05/2019, p. 165/196
22/22 acquire the second magnetic resonance image data from the patient subsequent to the acquisition of the first magnetic resonance image data;
align the first magnetic resonance image data and the second magnetic resonance image data; and comparing the first magnetic resonance imaging data and the second magnetic resonance imaging data aligned to detect at least one change in biological matter on the part of the patient.

类似技术:

---

公开号 | 公开日 | 专利标题

---

BR112019010225A2|2019-08-20|systems and methods for automated detection in magnetic resonance imaging

---

US20190154783A1|2019-05-23|Tensor field mapping

---

US10359486B2|2019-07-23|Rapid determination of a relaxation time

---

US11156688B2|2021-10-26|Correcting for hysteresis in magnetic resonance imaging

---

KR102354723B1|2022-01-24|tensor field mapping

---

CN103945760A|2014-07-23|Apparatus and method for influencing and/or detecting magnetic particles having a large field of view

---

CN103997958A|2014-08-20|Removal of background in MPI

---

WO2017007663A1|2017-01-12|Field-invariant quantitative magnetic-resonance signatures

---

EP3320358A1|2018-05-16|Field-invariant quantitative magnetic-resonance signatures

---

US20210080528A1|2021-03-18|Gradient coil apparatus and methods for mri

---

Chen et al.2020|Excitation Strategy Analysis of Magnetic Induction Tomography for Intracranial Hemorrhage Detection

同族专利:

---

公开号 | 公开日

---

US10534058B2|2020-01-14|

---

US20200341095A1|2020-10-29|

---

US10955504B2|2021-03-23|

---

US10585156B2|2020-03-10|

---

AU2017363608A1|2019-05-23|

---

US10718842B2|2020-07-21|

---

TW201825047A|2018-07-16|

---

US20190033415A1|2019-01-31|

---

WO2018098141A1|2018-05-31|

---

EP3545494B1|2022-01-05|

---

US20180144467A1|2018-05-24|

---

MX2019005955A|2019-07-10|

---

TWI704903B|2020-09-21|

---

US20190033414A1|2019-01-31|

---

CA3043038A1|2018-05-31|

---

EP3545494A4|2020-05-27|

---

US10416264B2|2019-09-17|

---

EP3545494A1|2019-10-02|

---

EP3968278A1|2022-03-16|

---

US10816629B2|2020-10-27|

---

JP2019535424A|2019-12-12|

---

KR20190087455A|2019-07-24|

---

IL266748D0|2019-07-31|

---

CN109983474A|2019-07-05|

---

US20180143281A1|2018-05-24|

---

US20180143275A1|2018-05-24|

---

US20190011521A1|2019-01-10|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US5900793A|1997-07-23|1999-05-04|Odin Technologies Ltd|Permanent magnet assemblies for use in medical applications|

---

ES2337556T3|2001-01-23|2010-04-27|Health Discovery Corporation|ANALYSIS OF IMAGES ASSISTED BY ERDENADOR.|

---

US7734075B2|2001-10-04|2010-06-08|Siemens Medical Solutions Usa, Inc.|Contrast-invariant registration of cardiac and renal magnetic resonance perfusion images|

---

US8272582B2|2001-11-26|2012-09-25|Gillette Thomas D|Systems and methods for producing ozonated water on demand|

---

WO2004031348A2|2002-09-24|2004-04-15|The Government Of The United States Of America As Represented By The Secretary Of The Department Of Health And Human Services|Method for convection enhanced delivery of therapeutic agents|

---

JP2005118098A|2003-10-14|2005-05-12|Hitachi Medical Corp|Magnetic resonance imaging apparatus|

---

US7711405B2|2004-04-28|2010-05-04|Siemens Corporation|Method of registering pre-operative high field closed magnetic resonance images with intra-operative low field open interventional magnetic resonance images|

---

JP4528322B2|2007-09-28|2010-08-18|富士フイルム株式会社|Image display device, image display method, and image display program|

---

EP2194486A1|2008-12-04|2010-06-09|Koninklijke Philips Electronics N.V.|A method, apparatus, and computer program product for acquiring medical image data|

---

US10303986B2|2009-04-07|2019-05-28|Kayvan Najarian|Automated measurement of brain injury indices using brain CT images, injury data, and machine learning|

---

US20110210734A1|2010-02-26|2011-09-01|Robert David Darrow|System and method for mr image scan and analysis|

---

GB201009101D0|2010-06-01|2010-07-14|Nordic Bioscience As|Computer based analysis of MRI images|

---

EP2468207A1|2010-12-21|2012-06-27|Renishaw Limited|Method and apparatus for analysing images|

---

US8812077B2|2010-12-22|2014-08-19|Viewray Incorporated|System and method for image guidance during medical procedures|

---

JP5387629B2|2011-07-30|2014-01-15|株式会社デンソー|DCDC converter control device|

---

US10143377B2|2012-05-02|2018-12-04|Augusta University Research Institute, Inc.|Single channel imaging measurement of dynamic changes in heart or respiration rate|

---

US9165360B1|2012-09-27|2015-10-20|Zepmed, Llc|Methods, systems, and devices for automated analysis of medical scans|

---

WO2014152919A1|2013-03-14|2014-09-25|Arizona Board Of Regents, A Body Corporate Of The State Of Arizona For And On Behalf Of Arizona State University|Kernel sparse models for automated tumor segmentation|

---

EP2972451A4|2013-03-15|2016-11-30|Synaptive Medical Barbados Inc|System and method for magnetic resonance image acquisition|

---

US9788795B2|2013-04-12|2017-10-17|The Research Foundation For The State University Of New York|Magnetic resonance imaging method|

---

US20140364720A1|2013-06-10|2014-12-11|General Electric Company|Systems and methods for interactive magnetic resonance imaging|

---

US9655563B2|2013-09-25|2017-05-23|Siemens Healthcare Gmbh|Early therapy response assessment of lesions|

---

CA2944829A1|2014-05-23|2015-11-26|Ting Chen|Systems and methods for detection of biological structures and/or patterns in images|

---

US20170140551A1|2014-06-30|2017-05-18|Universität Bern|Method for segmenting and predicting tissue regions in patients with acute cerebral ischemia|

---

JP6128691B2|2014-07-10|2017-05-17|富士フイルム株式会社|MEDICAL IMAGE MEASUREMENT DEVICE AND METHOD, AND PROGRAM|

---

TWI542328B|2014-07-31|2016-07-21|國立中央大學|Method for detecting and quantifying cerebral infarct|

---

JP6684804B2|2014-09-05|2020-04-22|ハイパーファイン リサーチ，インコーポレイテッド|Noise suppression method and apparatus|

---

KR101676789B1|2014-10-21|2016-11-16|삼성전자주식회사|Magnetic resonance imaging apparatus and processing method thereof|

---

WO2016077438A2|2014-11-11|2016-05-19|Hyperfine Research, Inc.|Pulse sequences for low field magnetic resonance|

---

WO2016077417A1|2014-11-11|2016-05-19|Hyperfine Research, Inc.|Low field magnetic resonance methods and apparatus|

---

CN106462802B|2014-11-14|2019-08-06|谷歌有限责任公司|Generate the natural language description of image|

---

US9874620B2|2015-02-05|2018-01-23|Ohio State Innovation Foundation|Low field magnetic resonance imaging scanner for cardiac imaging|

---

JP6800164B2|2015-04-13|2020-12-16|ハイパーファイン リサーチ，インコーポレイテッド|Magnetic coil power method and equipment|

---

WO2016183284A1|2015-05-12|2016-11-17|Hyperfine Research, Inc.|Radio frequency coil methods and apparatus|

---

US10194829B2|2015-07-07|2019-02-05|Q Bio, Inc.|Fast scanning based on magnetic resonance history|

---

US10909675B2|2015-10-09|2021-02-02|Mayo Foundation For Medical Education And Research|System and method for tissue characterization based on texture information using multi-parametric MRI|

---

WO2017106645A1|2015-12-18|2017-06-22|The Regents Of The University Of California|Interpretation and quantification of emergency features on head computed tomography|

---

CA3123141A1|2016-03-22|2017-09-28|Hyperfine Research, Inc.|Methods and apparatus for magnetic field shimming|

---

CN109310396B|2016-06-20|2021-11-09|蝴蝶网络有限公司|Automatic image acquisition for assisting a user in operating an ultrasound device|

---

US10210609B2|2016-09-07|2019-02-19|International Business Machines Corporation|Integrated deep learning and clinical image viewing and reporting|

---

TWI685668B|2016-09-29|2020-02-21|美商超精細研究股份有限公司|Magnetic resonance imaging system, and tuning system for use with the magnetic resonance imaging system|

---

US10585153B2|2016-11-22|2020-03-10|Hyperfine Research, Inc.|Rotatable magnet methods and apparatus for a magnetic resonance imaging system|

---

US10539637B2|2016-11-22|2020-01-21|Hyperfine Research, Inc.|Portable magnetic resonance imaging methods and apparatus|

---

EP3968278A1|2016-11-22|2022-03-16|Hyperfine, Inc.|Systems and methods for automated detection in magnetic resonance images|

---

US10627464B2|2016-11-22|2020-04-21|Hyperfine Research, Inc.|Low-field magnetic resonance imaging methods and apparatus|

---

WO2018231762A1|2017-06-13|2018-12-20|Bostongene, Corporation|Systems and methods for identifying cancer treatments from normalized biomarker scores|

---

US10761156B2|2018-04-20|2020-09-01|Hyperfine Research, Inc.|Deployable guard for portable magnetic resonance imaging devices|

---

BR112020021579A2|2018-05-21|2021-01-19|Hyperfine Research, Inc.|RADIOFREQUENCY COIL SIGNAL CHAIN FOR A LOW FIELD MAGNETIC RESONANCE IMAGE PRODUCTION SYSTEM|

---

AU2019274404A1|2018-05-21|2020-11-12|Hyperfine, Inc.|B0 magnet methods and apparatus for a magnetic resonance imaging system|

---

TW202015621A|2018-07-19|2020-05-01|美商超精細研究股份有限公司|Methods and apparatus for patient positioning in magnetic resonance imaging|

---

US20200033431A1|2018-07-30|2020-01-30|Jo Schlemper|Deep learning techniques for magnetic resonance image reconstruction|

---

TW202012951A|2018-07-31|2020-04-01|美商超精細研究股份有限公司|Low-field diffusion weighted imaging|

---

TW202014726A|2018-07-31|2020-04-16|美商超精細研究股份有限公司|Medical imaging device messaging service|

---

CA3107776A1|2018-08-15|2020-02-20|Hyperfine Research, Inc.|Deep learning techniques for suppressing artefacts in magnetic resonance images|

---

WO2020131250A2|2018-12-19|2020-06-25|Hyperfine Research, Inc.|System and methods for grounding patients during magnetic resonance imaging|

---

CN113498479A|2018-12-28|2021-10-12|海珀菲纳股份有限公司|Correcting magnetic hysteresis in magnetic resonance imaging|

---

KR20210137142A|2019-03-12|2021-11-17|하이퍼파인, 인크.|Systems and methods for magnetic resonance imaging of infants|

---

US20200289019A1|2019-03-14|2020-09-17|Hyperfine Research, Inc.|Deep learning techniques for generating magnetic resonance images from spatial frequency data|

---

US20200337587A1|2019-04-26|2020-10-29|Hyperfine Research, Inc.|Techniques for dynamic control of a magnetic resonance imaging system|

---

EP3965653A1|2019-05-07|2022-03-16|Hyperfine, Inc.|Systems, devices, and methods for magnetic resonance imaging of infants|EP2387708B1|2009-01-16|2019-05-01|New York University|Automated real-time particle characterization and three-dimensional velocimetry with holographic video microscopy|

---

EP3068298A4|2013-11-15|2018-02-21|New York University|Self calibrating parallel transmission by spin dynamic fingerprinting|

---

JP6684804B2|2014-09-05|2020-04-22|ハイパーファイン リサーチ，インコーポレイテッド|Noise suppression method and apparatus|

---

WO2016077417A1|2014-11-11|2016-05-19|Hyperfine Research, Inc.|Low field magnetic resonance methods and apparatus|

---

WO2016077438A2|2014-11-11|2016-05-19|Hyperfine Research, Inc.|Pulse sequences for low field magnetic resonance|

---

EP3218690B1|2014-11-12|2022-03-09|New York University|Colloidal fingerprints for soft materials using total holographic characterization|

---

JP6800164B2|2015-04-13|2020-12-16|ハイパーファイン リサーチ，インコーポレイテッド|Magnetic coil power method and equipment|

---

WO2016183284A1|2015-05-12|2016-11-17|Hyperfine Research, Inc.|Radio frequency coil methods and apparatus|

---

US10561337B2|2015-08-04|2020-02-18|University Of Virginia Patent Foundation|Rapid 3D dynamic arterial spin labeling with a sparse model-based image reconstruction|

---

KR20180056421A|2015-09-18|2018-05-28|뉴욕 유니버시티|Holographic detection and characterization of large impurity particles in fine slurry|

---

CA3123141A1|2016-03-22|2017-09-28|Hyperfine Research, Inc.|Methods and apparatus for magnetic field shimming|

---

US10670677B2|2016-04-22|2020-06-02|New York University|Multi-slice acceleration for magnetic resonance fingerprinting|

---

TWI685668B|2016-09-29|2020-02-21|美商超精細研究股份有限公司|Magnetic resonance imaging system, and tuning system for use with the magnetic resonance imaging system|

---

US10585153B2|2016-11-22|2020-03-10|Hyperfine Research, Inc.|Rotatable magnet methods and apparatus for a magnetic resonance imaging system|

---

EP3968278A1|2016-11-22|2022-03-16|Hyperfine, Inc.|Systems and methods for automated detection in magnetic resonance images|

---

US10539637B2|2016-11-22|2020-01-21|Hyperfine Research, Inc.|Portable magnetic resonance imaging methods and apparatus|

---

US10627464B2|2016-11-22|2020-04-21|Hyperfine Research, Inc.|Low-field magnetic resonance imaging methods and apparatus|

---

US10390727B2|2017-04-21|2019-08-27|The Charles Stark Draper Laboratory, Inc.|Apparatus and method for imaging currents using nanoparticles and low-field magnetic resonance imaging |

---

US10753997B2|2017-08-10|2020-08-25|Siemens Healthcare Gmbh|Image standardization using generative adversarial networks|

---

US10365340B1|2018-03-01|2019-07-30|Siemens Medical Solutions Usa, Inc.|Monitoring dynamics of patient brain state during neurosurgical procedures|

---

US10761156B2|2018-04-20|2020-09-01|Hyperfine Research, Inc.|Deployable guard for portable magnetic resonance imaging devices|

---

AU2019274404A1|2018-05-21|2020-11-12|Hyperfine, Inc.|B0 magnet methods and apparatus for a magnetic resonance imaging system|

---

BR112020021579A2|2018-05-21|2021-01-19|Hyperfine Research, Inc.|RADIOFREQUENCY COIL SIGNAL CHAIN FOR A LOW FIELD MAGNETIC RESONANCE IMAGE PRODUCTION SYSTEM|

---

TW202015621A|2018-07-19|2020-05-01|美商超精細研究股份有限公司|Methods and apparatus for patient positioning in magnetic resonance imaging|

---

TW202012951A|2018-07-31|2020-04-01|美商超精細研究股份有限公司|Low-field diffusion weighted imaging|

---

TW202014726A|2018-07-31|2020-04-16|美商超精細研究股份有限公司|Medical imaging device messaging service|

---

WO2020036855A1|2018-08-12|2020-02-20|The Trustees Of Columbia University In The City Of New York|System, method and computer-accessible medium for tissue fingerprinting|

---

EP3833243A1|2018-08-12|2021-06-16|The Trustees of Columbia University in the City of New York|System, method, and computer-accessible medium for magnetic resonance value driven autonomous scanner|

---

WO2020056274A1|2018-09-14|2020-03-19|The Johns Hopkins University|Machine learning processing of contiguous slice image data|

---

US11037030B1|2018-10-29|2021-06-15|Hrl Laboratories, Llc|System and method for direct learning from raw tomographic data|

---

EP3663785A1|2018-12-07|2020-06-10|Koninklijke Philips N.V.|Functional magnetic resonance imaging artifact removal by means of an artificial neural network|

---

US20210313046A1|2018-12-14|2021-10-07|The Board Of Trustees Of The Leland Stanford Junior University|Qualitative and quantitative MRI using deep learning|

---

WO2020131250A2|2018-12-19|2020-06-25|Hyperfine Research, Inc.|System and methods for grounding patients during magnetic resonance imaging|

---

US10832392B2|2018-12-19|2020-11-10|Siemens Healthcare Gmbh|Method, learning apparatus, and medical imaging apparatus for registration of images|

---

CN113498479A|2018-12-28|2021-10-12|海珀菲纳股份有限公司|Correcting magnetic hysteresis in magnetic resonance imaging|

---

US20200289019A1|2019-03-14|2020-09-17|Hyperfine Research, Inc.|Deep learning techniques for generating magnetic resonance images from spatial frequency data|

---

WO2020223434A1|2019-04-30|2020-11-05|The Trustees Of Columbia University In The City Of New York|Classifying neurological disease status using deep learning|

---

WO2020232124A1|2019-05-13|2020-11-19|The General Hospital Corporation|System and methods for automated image analysis|

---

CN110346743B|2019-07-22|2021-09-14|上海东软医疗科技有限公司|Magnetic resonance diffusion weighted imaging method and device|

---

CN110472670B|2019-07-24|2022-03-01|上海联影智能医疗科技有限公司|Image centerline detection method, computer device, and storage medium|

---

CN110956636A|2019-11-28|2020-04-03|北京推想科技有限公司|Image processing method and device|

---

USD912822S1|2019-12-10|2021-03-09|Hyperfine Research, Inc.|Frame for magnets in magnetic resonance imaging|

---

USD932014S1|2019-12-10|2021-09-28|Hyperfine, Inc.|Frame for magnets in magnetic resonance imaging|

---

WO2021167124A1|2020-02-19|2021-08-26|서울대학교 산학협력단|Mri data conversion device including conversion network unit using mri data and additional information, and mri data conversion method using same|

---

DE102020202505A1|2020-02-27|2021-09-02|Robert Bosch Gesellschaft mit beschränkter Haftung|Method for operating a magnetic resonance measuring device|

---

CN111611218A|2020-04-24|2020-09-01|武汉大学|Distributed abnormal log automatic identification method based on deep learning|

---

CN111563903B|2020-04-26|2021-05-28|北京航空航天大学|MRI whole brain tissue segmentation method and system based on deep learning|

---

CN111583212B|2020-04-29|2021-11-30|上海杏脉信息科技有限公司|Method and device for determining brain midline shift|

法律状态:
2021-10-05| B350| Update of information on the portal [chapter 15.35 patent gazette]|

---

2021-12-28| B25D| Requested change of name of applicant approved|Owner name: HYPERFINE, INC. (US) |

优先权:

---

申请号 | 申请日 | 专利标题

---

US201662425569P| true| 2016-11-22|2016-11-22|

---

PCT/US2017/062763|WO2018098141A1|2016-11-22|2017-11-21|Systems and methods for automated detection in magnetic resonance images|

---