• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The invention provides a data processing system for acquiring tumor position and edge contour in a CT image and electronic equipment. A CT image is acquired; the CT image is segmented by using a particle swarm optimization algorithm to obtain a segmented CT image data matrix; a QSOFM classifier model is built; the model is trained by using the segmented CT image data matrix; the cancerous type of tumors in the CT image is recognized by using the QSOFM classifier model; position and contour information of the cancerous tumors in the CT image are acquired based on a multi-connected region segmentation method. The image is segmented further to reduce redundant data; whereby an automatic puncture device has higher recognition rate of tumors in the CT image; the tumors are positioned faster and accurately; improving the degree of automation of the automatic puncture device; a foundation is laid for computer-assisted therapy.
    公开号:NL2025775A
    申请号:NL2025775
    申请日:2020-06-08
    公开日:2021-05-17
    发明作者:Yuan Shuanghu；Li Wei；Liu Ning；Meng Xiangwei；Yu Qingxi；Wei Yuchun；Yu Jinming；Li Li；Li Xiaoxiao；Liu Changmin
    申请人:Shandong Cancer Hospital And Inst；Univ Shandong；
    IPC主号:
    专利说明:

    DATA PROCESSING SYSTEM FOR ACQUIRING TUMOR POSITION AND
    CONTOUR IN CT IMAGE AND ELECTRONIC EQUIPMENT Field of the Invention The present invention relates to the field of image processing technology, and in particular, to a data processing system for acquiring a tumor position and contour in a CT image and electronic equipment. Background of the Invention The statement of this section merely provides the background art related to the present invention, and does not necessarily constitute the prior art. At present, there are a variety of image-based tumor recognition methods, and most of them have outstanding recognition results. The most commonly used are a convolutional neural network-based recognition method, a traditional sliding window method, an eigenvalue method, etc. The inventors of the present invention have found in research that, with the further development of computer technology and robot technology, some basic medical operations are slowly being replaced by mechanical devices, for example, during radiotherapy of tumors in radiotherapy rooms of hospitals, the tumors need to be punctured. However, these devices currently cannot perform effective data processing on the acquired CT images to obtain tumor position and contour information in the CT images, thereby limiting the development of automatic puncture devices. Therefore, how to quickly and accurately perform data processing and analysis on CT images to implement positioning, recognition and contour segmentation on tumors in the CT images so as to improve the automation level of medical automation equipment is a technical problem that needs to be urgently solved. Summary of the Invention In order to solve the shortcomings of the prior art, the present invention provides a data processing system for acquiring a tumor position and contour in a CT image and electronic equipment, where an image is segmented more delicately to reduce redundant data by using a particle swarm optimization algorithm, a quantum-based self-organization feature mapping neural network classifier model and a multi-connected region segmentation method, so that the tumor position and contour in the CT image can be quickly obtained, and the automation level of a medical automation device is improved.
    In order to achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a data processing system for acquiring a tumor position and edge contour in a CT image; A data processing system for acquiring a tumor position and edge contour in a CT image, including: a preprocessing module, configured to acquire a CT image, and segment the acquired CT image by using a particle swarm optimization algorithm to obtain a segmented CT image data matrix; a model building module, configured to build a QSOFM classifier model, and train the model by using the segmented CT image data matrix; and a data processing module, configured to recognize the cancerous type of tumors in the CT image by using the QSOFM classifier model, and acquire position and contour information of the cancerous tumors in the CT image based on a multi-connected region segmentation method.
    As some possible implementation manners, the preprocessing module segments the acquired CT image by using a particle swarm optimization algorithm, specifically: (1-1) establishing an energy minimization control point generalization function for image segmentation, and generating a particle swarm; (1-2) calculating a fitness function of the particle swarm algorithm to generate a fitness value of each particle; (1-3) comparing the fitness value of each particle with the individual optimal value of the particle and the individual optimal value of the particle swarm, and replacing, if the fitness value of the particle is better, the individual optimal value of the particle and the individual optimal value of the particle swarm; (1-4) generating an optimal assistance point according to an orthogonal test mechanism; (1-5) calculating a linear decreasing weight; (1-6) calculating a new position and velocity of the particle swarm, and calculating a fitness value thereof, (1-7) terminating, if the number of iterations of the particle swarm exceeds the maximum number of iterations, the search process of the particle swarm; (1-8) obtaining energy-minimum control points of a question according to the individual optimal value of the particle swarm; and (1-9) segmenting the image according to the energy-minimum control points.
    As some possible implementation manners, the method of training the QSOFM classifier model is specifically: (2-1) initializing system parameters of the QSOFM classifier model to set the parameters; (2-2) performing cluster training on the QSOFM classifier model by using the CT image data matrix processed by the particle swarm optimization algorithm to obtain a QSOFM classifier model that can be used for classification and recognition; and (2-3) randomly selecting a plurality of tumor CT images, and judging by using the trained QSOFM classifier model that the tumors on the CT images are normal, cancerous, or unknown.
    As a further limitation, in step (2-1), the parameters are set according to 255 competing nodes, a neighborhood radius 5, a learning rate 1.0, a threshold 2.0, 100 unsupervised training steps and 100 supervised training steps.
    As a further limitation, in step (2-3), a plurality of CT images are randomly selected to construct a CT image sample matrix to be tested, a feature data matrix is generated after calculation using the particle swarm optimization algorithm, and the feature data matrix is input to the QSOFM classifier model.
    As some possible implementation manners, the data processing module is also configured to segment and number the connected regions of high gray value regions of the CT image after segmentation using the particle swarm optimization algorithm. As a further limitation, when the connected regions are segmented, the centroid of each connected region is calculated, and the connected regions are segmented according to the positions of the centroids; and at the same time, the connected regions are selected by using an external rectangular box and numbered sequentially.
    As a further limitation, the numbered images are copied by the same number as the number of connected regions and numbered sequentially, and the connected regions with the same numbers as the images are filled; wherein the images and the connected regions are numbered by the same rule.
    Further, when the numbered connected regions are sequentially filled, the selected connected regions are filled without affecting other connected regions; Further, after the images are filled, the filled images are input to the QSOFM classifier model, and the numbers of images whose processing results are normal are recorded, and the connected regions corresponding to the numbers represent the positions of tumors in the CT images. As a further limitation, the connected regions with recorded numbers are retained, while other connected regions are filled and processed to obtain the numbers of all images whose processing results are normal, and the filled images are compared with the original CT image to obtain position and contour information of the tumors in the original CT image. In a second aspect, the present invention provides a readable storage medium, storing a program that, when executed by a processor, implements a data processing method of the data processing system for acquiring a tumor position and contour in a CT image according to the present invention.
    According to a third aspect, the present invention provides electronic equipment, including a memory, a processor, and a program stored on the memory and executable on the processor, wherein the processor, when executing the program, implements a data processing method of the data processing system for acquiring a tumor position and contour in a CT image according to the present invention.
    Compared with the prior art, the beneficial effects of the present invention are:
    An image is segmented more delicately to reduce redundant data by using a particle swarm optimization algorithm, a quantum-based self-organization feature mapping neural network classifier model (QSOFM classifier model) and a multi-connected region segmentation method, so that the medical automation equipment has higher 5 recognition rate of tumors in the CT image, and the tumors are positioned more quickly and accurately. The QSOFM classifier model is built based on a quantum neural network, so that faster processing speed and larger storage space are achieved when the CT image is processed and recognized, the retrieval time is reduced, and the retrieval results are optimized. The CT image is further segmented by a multi-connected region segmentation method in the presence of tumors to acquire tumor position information and edge contour information in the CT image, thereby improving the degree of automation of an automatic puncture device, and laying a foundation for computer-assisted therapy.
    Brief Description of the Drawings Fig. 1 is a schematic structural diagram of a data processing system for acquiring tumor position and contour information in a CT image according to Embodiment | of the present invention. Fig. 2 is a flowchart of segmenting the CT image by a particle swarm optimization algorithm according to Embodiment 1 of the present invention. Fig. 3 is a flowchart of training and recognizing a QSOFM classifier model according to Embodiment 1 of the present invention. Fig. 4 is a flowchart of a data processing method of the data processing system for acquiring tumor position and contour information in a CT image according to Embodiment 1 of the present invention. Detailed Description of Embodiments It should be pointed out that the following detailed descriptions are all exemplary and aim to further illustrate the present invention. Unless otherwise specified, all technical and scientific terms used in the descriptions have the same meanings generally understood by those of ordinary skill in the art of the present invention.
    It should be noted that the terms used herein are merely for describing specific embodiments, but are not intended to limit exemplary embodiments according to the present invention.
    As used herein, unless otherwise explicitly pointed out by the context, the singular form is also intended to include the plural form.
    In addition, it should also be understood that when the terms “include” and/or “comprise” are used in the specification, they indicate features, steps, operations, devices, components and/or their combination.
    Embodiment 1 As shown in Figs. 1-4, Embodiment 1 of the present invention provides a data processing system for acquiring tumor positions and edge contours in a CT image, including: a preprocessing module, configured to acquire a CT image, and segment the acquired CT image by using a particle swarm optimization algorithm to obtain a segmented CT image data matrix; a model building module, configured to build a QSOFM classifier model, and train the model by using the segmented CT image data matrix; and a data processing module, configured to recognize the cancerous type of tumors in the CT image by using the QSOFM classifier model, and acquire position and contour information of the cancerous tumors in the CT image based on a multi-connected region segmentation method.
    The QSOFM classifier model is a two-layer network model, including an input layer and a competition layer.
    The input layer is a sample space for receiving external signals, and the competition layer may also be referred to as an output layer, which is a one-dimensional or two-dimensional planar array composed of multiple neurons.
    The neurons in the output layer are in an interaction mode of “near neighborhood excitation, far neighbor suppression”. The preprocessing module segments the acquired tumor CT image by calculation using the particle swarm optimization algorithm, specifically:
    (1-1) establishing an energy minimization control point generalization function for Image segmentation, and generating a particle swarm; (1-2) calculating a fitness function of the particle swarm algorithm to generate a fitness value of each particle;
    (1-3) comparing the fitness value of each particle with the individual optimal value (Pwest) of the particle and the individual optimal value (gpes) of the particle swarm, and replacing, if the fitness value of the particle is better, the individual optimal value of the particle and the individual optimal value of the particle swarm;
    A D-dimensional space is assumed to have m particles that form a particle swarm.
    At the &-th iteration, the position vector and velocity vector of the i-th particle are respectively Xi[k] and Vi[k], wherein the position vector corresponds to a potential solution of a question.
    All particles are searched in the space of the solution according tO Pivest and Zes: guidance, and the calculation formulas are:
    Vk+1]=oV, [k]+en(Ph, —X, [D+ cn Sese — X, [ED 0)
    X, [k+1]=X [k]+F,[k +1] ©)
    Where ¢; and ¢; are individual and social cognition coefficients, ry and #2 are random numbers, and © is an inertia weight and is changed as follows:
    © = EC —@ LE 7 3)
    (1-4) generating an optimal assistance point Xpes according to an orthogonal test mechanism, and replacing Spes With Xe;
    Pi is assumed to denote the i-th energy-minimum control point.
    In the i-th search window SW, gq; is the j-th candidate energy-minimum control point, and the following generalization function of energy-minimum control points is established:
    E,= : ipa, HBP, 29, +P J= reg, Vr VG 1a, T+ Vim ET 4)
    Where & B Sine Vege and Vien are all weight coefficients; C represents an edge function of a target image; (xy represents a partial derivative of the C' function in the x direction and then a partial derivative of y; Cx represents a partial derivative of the C° function in the x direction; and Cy represents a partial derivative of the C function in the y direction. (1-5) calculating a linear decreasing weight; (1-6) calculating a new position and velocity of the particle swarm by formulas (3) and (4), and calculating a fitness value thereof; (1-7) terminating, if the number of iterations of the particle swarm exceeds the maximum number of iterations, the search process of the particle swarm; (1-8) obtaining energy-minimum control points of a question according to the individual optimal value of the particle swarm; and (1-9) segmenting the image according to the energy-minimum control points.
    The method of building the QSOFM classifier model is specifically: (2-1) initializing system parameters of the QSOFM classifier model, and setting the parameters according to 255 competing nodes, a neighborhood radius 5, a learning rate 1.0, a threshold 2.0, 100 unsupervised training steps and 100 supervised training steps; (2-2) performing cluster training on the QSOFM classifier model by using the CT image data matrix processed by the particle swarm optimization algorithm to obtain a QSOFM classifier model that can be used for classification and recognition; and (2-3) randomly selecting a plurality of tumor CT images to construct a CT image sample matrix to be tested, generating a feature data matrix after calculation using the particle swarm optimization algorithm, inputting the feature data matrix to the QSOFM classifier model, and judging by using the trained QSOFM classifier model that the tumors on the CT images are normal, cancerous, or unknown.
    The data processing module is also configured to segment and number the connected regions of high gray value regions of the CT image after segmentation using the particle swarm optimization algorithm.
    When the connected regions are segmented, the centroid of each connected region is calculated, and the connected regions are segmented according to the positions of the centroids; and at the same time, the connected regions are selected by using an external rectangular box and numbered sequentially as a=1, 2,3, 4 …; The numbered images are copied by the same number as the number of connected regions and numbered sequentially as b=1, 2, 3, 4 ..., and the connected regions with the same numbers as the images are filled; When the numbered connected regions are sequentially filled, the selected connected regions are filled without affecting other connected regions; After the images are filled, the filled images are input to the QSOFM classifier model, and the numbers of images whose processing results are normal are recorded, and the connected regions corresponding to the numbers represent the positions of tumors in the CT images.
    The connected regions with recorded numbers are retained, while other connected regions are filled and processed to obtain the numbers of all images whose processing results are normal, and the filled images are compared with the original CT image to obtain position and contour information of the tumors in the original CT image.
    Embodiment 2: Embodiment 2 of the present invention provides a readable storage medium, storing a program that, when executed by a processor, implements a data processing method of the data processing system for acquiring a tumor position and edge contour in a CT image according to Embodiment 1 of the present invention.
    Embodiment 3: Embodiment 3 of the present invention provides electronic equipment, including a memory, a processor, and a program stored on the memory and executable on the processor, wherein the processor, when executing the program, implements a data processing method of the data processing system for acquiring a tumor position and contour in a CT image according to Embodiment 1 of the present invention.
    Described above are merely preferred embodiments of the present invention, and the present invention is not limited thereto.
    Various modifications and variations may be made to the present invention for those skilled in the art.
    Any modification, equivalent substitution or improvement made within the spirit and principle of the present invention shall fall into the protection scope of the present invention.
    权利要求:
    Claims (10)
    [1]
    A data processing system for obtaining the position and edge contour of a tumor in a CT image, comprising: a preprocessing module configured to acquire a CT image and to segment the acquired CT image by means of a particle swarm optimization algorithm to obtain a data matrix of segmented CT images; a model building module configured to build a QSOFM layout model and to train the model using the data matrix of segmented CT images; and a data processing module configured to recognize the cancerous type of tumors in the CT image by means of the QSOFM format model, and to obtain the position and the edge contour information of cancerous tumors in the CT image based on of a segmentation method for a multi-connected region.
    [2]
    A data processing system for obtaining the position and edge contour of a tumor in a CT image according to claim 1, wherein the pre-processing module segments the obtained CT image by means of a particle swarm optimization algorithm, in particular: (1-1) generating of a generalization function of setpoints with energy minimization for the image segmentation, and generating a particle swarm; (1-2) calculation of a fitness function of the particle swarm optimization algorithm to generate a fitness value of each particle; (1-3) comparison of the fitness value of each particle with the individual optimal value of the particle and with the individual optimal value of the swarm, and substitution, if the fitness value of the particle is better, of the individual optimal value of the particle and of the individual optimum value of the particle swarm; (1-4) generating an optimal auxiliary point according to an orthogonal testing mechanism; (1-5) calculation of a linearly decreasing weight; (1-6) calculation of a new position and velocity of the particle swarm, and calculation of a fitness value thereof; (1-7) termination of the particle swarm search process if the number of iterations of the particle swarm exceeds the maximum number of iterations; (1-8) Acquiring energy minimum set points of a demand according to the individual optimal value of the particle swarm; and (1-9) segmentation of the image according to the energy minimum set points.
    [3]
    The data processing system for obtaining the position and the edge contour of a tumor in a CT image according to claim 1, wherein the training of the QSOFM layout model comprises in particular the following steps: (2-1) imitation of system parameters of the QSOFM layout model to set the parameters; (2-2) performing cluster training on the QSOFM mapping model using the CT image data matrix processed by the particle swarm optimization algorithm to obtain a trained QSOFM mapping model that can be used for mapping and recognition , and (2-3) randomly selecting a plurality of CT tumor images, and assessing, using the QSOFM grading model, whether the tumors on the CT images are normal, cancerous, or unknown.
    [4]
    The data processing system for obtaining the position and the edge contour of a tumor in a CT image according to claim 3, wherein in step (2-1) the parameters are set according to 255 competing nodes, a learning speed of 1.0, a neighborhood radius of 5, a threshold of 2.0, 100 unsupervised training steps and 100 supervised training steps; also in step (2-3) a plurality of CT images are randomly selected to build a CT image sample matrix to be tested, a feature data matrix is generated after calculation using the particle swarm optimization algorithm, and the feature data matrix in the QSOFM format model is introduced.
    [5]
    The data processing system for obtaining the position and edge contour of a tumor in a CT image according to claim 1, wherein the data processing module also for segmenting and numbering the connected regions of the high grayscale regions of the CT image after segmentation configured using the particle swarm optimization algorithm.
    [6]
    The data processing system for obtaining the position and edge contour of a tumor in a CT image according to claim 5, wherein, when the connected regions are segmented, the centroid of each connected region is calculated, and the connected regions are calculated according to the locations of the centers of gravity are segmented; and at the same time the connected areas are selected by means of an outer rectangular box and numbered in sequence.
    [7]
    The data processing system for obtaining the position and edge contour of a tumor in a CT image according to claim 6, wherein the numbered images are copied in the same number as the number of connected areas, and numbered in the order, and the connected areas are the same numbers as the images show are filled in; further, when the numbered connected areas in the order are padded, the selected connected areas are filled without affecting other connected areas; further, after the images are filled, the filled images are entered into the QSOFM layout model, and the numbers of the images whose processing results are normal are recorded, and the connected areas corresponding to the numbers, the positions of the tumors on represent the CT images.
    [8]
    The data processing system for obtaining the position and edge contour of a tumor in a CT image according to claim 7, wherein the connected regions with recorded numbers are kept while other connected regions are filled and processed to obtain the numbers of all images whose processing results are normal, and the filled images are compared with the original CT image to obtain position and contour information of the tumors in the original CT image.
    [9]
    A readable storage medium that stores a program which, when executed by a processor, applies a data processing method of the data processing system to obtain the position and edge contour of a tumor in a CT image according to any one of claims 1 to 8.
    [10]
    An electronic equipment comprising a memory, a processor, and a program stored in the memory and executable by the processor, wherein the processor, when executing the program, uses a data processing method of the data processing system to obtain the position and edge contour of a tumor in a CT image according to one of the claims | - 8 applies.
    类似技术:
    公开号 | 公开日 | 专利标题
    Li et al.2016|Deepsaliency: Multi-task deep neural network model for salient object detection
    Chen et al.2020|Detection of rice plant diseases based on deep transfer learning
    Srivastava et al.2021|Comparative analysis of deep learning image detection algorithms
    Lingappa et al.2018|Medical image segmentation based on extreme learning machine algorithm in kernel fuzzy c-means using artificial bee colony method
    CN107577983A|2018-01-12|It is a kind of to circulate the method for finding region-of-interest identification multi-tag image
    NL2025775B1|2021-12-14|Data processing system for acquiring tumor position and contour in ct image and electronic equipment
    Wang et al.2017|Point linking network for object detection
    Basly et al.2020|CNN-SVM learning approach based human activity recognition
    Teimouri et al.2019|A real-time ball detection approach using convolutional neural networks
    Yu et al.2019|Exemplar-based recursive instance segmentation with application to plant image analysis
    JP2020038660A|2020-03-12|Learning method and learning device for detecting lane by using cnn, and test method and test device using the same
    Zrira et al.2018|Discriminative deep belief network for indoor environment classification using global visual features
    Jia et al.2020|Fruit recognition based on pulse coupled neural network and genetic Elman algorithm application in apple harvesting robot
    Gojcic et al.2020|F2S3: Robustified determination of 3D displacement vector fields using deep learning
    Suhail et al.2020|Convolutional neural network based object detection: A review
    Wang et al.2018|Fast classification and detection of fish images with YOLOv2
    Basak et al.2021|Cervical Cytology Classification Using PCA and GWO Enhanced Deep Features Selection
    Bi et al.2020|Improving image-based plant disease classification with generative adversarial network under limited training set
    Deshmukh et al.2018|Fine object detection in automated solar panel layout generation
    Gorbatsevich et al.2018|SINGLE-SHOT SEMANTIC MATCHER FOR UNSEEN OBJECT DETECTION.
    Fung et al.2020|Using deep learning to find victims in unknown cluttered urban search and rescue environments
    Zheng et al.2014|Marginal space learning
    Pandian et al.2016|Analysis on Shape Image Retrieval Using DNN and ELM Classifiers for MRI Brain Tumor Images.
    Shi et al.2019|Underwater dense targets detection and classification based on YOLOv3
    Dorrer et al.2020|Building an artificial vision system of an agricultural robot based on the DarkNet system
    同族专利:
    公开号 | 公开日
    NL2025775B1|2021-12-14|
    CN110619644B|2022-01-28|
    CN110619644A|2019-12-27|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    CN104021552B|2014-05-28|2017-02-22|华南理工大学|Multi-objective particle swarm parameter optimization method based on graph segmentation process|
    CN104732213B|2015-03-23|2018-04-20|中山大学|A kind of area of computer aided Mass detection method based on mammary gland magnetic resonance image|
    CN105405136A|2015-11-04|2016-03-16|南方医科大学|Self-adaptive spinal CT image segmentation method based on particle swarm optimization|
    CN106952275A|2017-03-16|2017-07-14|南京师范大学|A kind of cell image segmentation method based on PSO Neural Network|
    CN107845098A|2017-11-14|2018-03-27|南京理工大学|Liver cancer image full-automatic partition method based on random forest and fuzzy clustering|
    CN108921049B|2018-06-14|2021-08-03|华东交通大学|Tumor cell image recognition device and equipment based on quantum gate line neural network|CN111419194A|2020-04-30|2020-07-17|山东大学|Fluorescent laser and OCT-based combined imaging device and method|
    法律状态:
    优先权:
    申请号 | 申请日 | 专利标题
    CN201910882016.2A|CN110619644B|2019-09-18|2019-09-18|Data processing system and electronic equipment for acquiring tumor position and contour in CT image|
    [返回顶部]