Tumor tissue segmentation method and system based on MR image, and storage medium

Abstract

The invention relates to the technical field of medical image processing, in particular to a tumor tissue segmentation method and system based on an MR image and a storage medium, and the method comprises the following steps: constructing a tumor edge segmentation module and a tumor non-edge region segmentation module through a neural network; a mutual learning module is arranged between the tumor edge segmentation module and the tumor non-edge region segmentation module; and training a model structure formed by combining the tumor edge segmentation module, the tumor non-edge region segmentation module and the mutual learning module to obtain a tumor tissue segmentation model. The tumor edge segmentation module, the tumor non-edge region segmentation module and the mutual learning module are constructed, so that edge segmentation and non-edge region segmentation are subjected to mutual learning and mutual guidance, the tumor tissue main body region can be accurately segmented, and the complex boundary of the tumor tissue can be accurately segmented; and the precise segmentation effect of the tumor tissue is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly relates to a tumor tissue segmentation method, system and storage medium based on MR images. Background Art

[0002] Tumor segmentation aims to automatically segment tumor regions from multimodal magnetic resonance (MR) images taken by advanced medical imaging devices. By segmenting tumors, tumor volume, shape and location can be provided, which play a crucial role in tumor diagnosis and monitoring.

[0003] In the prior art, deep learning algorithms are usually used to segment tumor tissues in MR images. Such tumor tissue segmentation models established based on deep learning tend to focus on the segmentation and recognition of the overall region of tumor tissues, and most of them can accurately segment the main region of tumor tissues. However, it is difficult to accurately segment the complex boundaries of tumor tissues. Therefore, the accurate segmentation effect of tumor tissues is limited. Summary of the Invention

[0004] The purpose of the present invention is to provide a tumor tissue segmentation method, system and storage medium based on MR images to solve the technical problem that it is difficult to accurately segment the complex boundaries of tumor tissues in the prior art.

[0005] To solve the above technical problem, the present invention specifically provides the following technical solutions: A tumor tissue segmentation method based on MR images, comprising the following steps: Obtain MR images containing tumor tissues; Use a neural network to respectively construct a tumor edge segmentation module for segmenting the edges of tumor tissues in MR images and a tumor non-edge region segmentation module for segmenting non-edge regions of tumor tissues in MR images; Between the tumor edge segmentation module and the tumor non-edge region segmentation module, set a mutual learning module for controlling the mutual learning between the tumor tissue edge segmentation and the tumor tissue non-edge region segmentation; Train the model structure composed of the tumor edge segmentation module, the tumor non-edge region segmentation module and the mutual learning module to obtain a tumor tissue segmentation model for segmenting tumor tissues in MR images.

[0006] As a preferred solution of the present invention, the construction method of the tumor edge segmentation module includes: Use the Mask F-CNN network as the model structure of the tumor edge segmentation module; Taking the MR image as the input of the Mask F-CNN network and the edge mask of the tumor tissue in the MR image as the output of the Mask F-CNN network, the tumor edge segmentation module is obtained as follows: G rough = Mask F-CNN(MR); where G rough is the edge mask, MR is the MR image, and Mask F-CNN is the Mask F-CNN network.

[0007] As a preferred embodiment of the present invention, the construction method of the tumor non-edge region segmentation module includes: Taking the Mask F-CNN network as the model structure of the tumor non-edge segmentation module; Taking the MR image as the input of the Mask F-CNN network and the non-edge region mask of the tumor tissue in the MR image as the output of the Mask F-CNN network, the tumor non-edge region segmentation module is obtained as follows: G body = Mask F-CNN(MR); where G body is the non-edge region mask, MR is the MR image, and Mask F-CNN is the Mask F-CNN network.

[0008] As a preferred embodiment of the present invention, the construction method of the mutual learning module includes: Constructing a first path for the tumor edge segmentation module to learn from the tumor non-edge region segmentation module and quantifying the learning rate of the first path; The learning rate of the first path is: ; where is the learning rate of the first path, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used for training the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module on the i-th sample in the dataset used for training the model structure, m is the total number of samples in the dataset used for training the model structure, is and the KL divergence between; Constructing a second path for the tumor non-edge region segmentation module to learn from the tumor edge segmentation module and quantifying the learning rate of the second path; The learning rate of the second path is: ; where is the learning rate of the second path, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used for training the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module on the i-th sample in the dataset used for training the model structure, and m is the total number of samples in the dataset used for training the model structure, is and the KL divergence between; where ; .

[0009] As a preferred solution of the present invention, the training loss function of the model structure is: ; where ; ; ; In the formula, is the total loss, is the prediction loss, is the mutual learning loss, is the reconstruction loss, , and are the balanced , and hyperparameters, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used for training the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module on the i-th sample in the dataset used for training the model structure, and m is the total number of samples in the dataset used for training the model structure, is the ground truth of the non-edge region mask on the i-th sample in the dataset used for training the model structure, is the ground truth of the edge mask on the i-th sample in the dataset used for training the model structure, is the ground truth of the overall mask of the tumor tissue on the i-th sample in the dataset used for training the model structure, is the learning rate of the first path, is the learning rate of the second path, , and are all L2 norm expressions.

[0010] As a preferred embodiment of the present invention, the method for training the model structure to obtain the tumor tissue segmentation model includes: Dividing the data set into a training set and a test set; Based on the loss function, training the model structure on the training set to obtain a tumor tissue segmentation model; Based on model evaluation metrics, evaluating the performance of the tumor tissue segmentation model on the test set.

[0011] As a preferred embodiment of the present invention, the method for the tumor tissue segmentation model to obtain the tumor tissue segmentation result of the MR image includes: Inputting the MR image into the tumor tissue segmentation model to obtain an edge mask and a non-edge region mask; Superimposing and fusing the edge mask and the non-edge region mask to obtain an overall mask of the tumor tissue, which is used as the tumor tissue segmentation result of the MR image.

[0012] As a preferred embodiment of the present invention, the hyperparameter , where ; ; In the formula, is the balance hyperparameter of , is the balance hyperparameter of , is the edge mask output by the tumor edge segmentation module on the i-th sample in the data set used for training the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module on the i-th sample in the data set used for training the model structure, m is the total number of samples in the data set used for training the model structure, is the ground truth of the non-edge region mask on the i-th sample in the data set used for training the model structure, is the ground truth of the edge mask on the i-th sample in the data set used for training the model structure, is the learning rate of the first path, is the learning rate of the second path, , are both L2 norm formulas.

[0013] As a preferred embodiment of the present invention, the present invention provides a tumor tissue segmentation system based on MR images, which is applied to a tumor tissue segmentation method based on MR images. The system includes: A data acquisition unit for acquiring MR images containing tumor tissues; A model building unit is used to respectively build a tumor edge segmentation module for segmenting the edge of tumor tissue in MR images and a non-edge area segmentation module of tumor tissue for segmenting the non-edge area of tumor tissue in MR images; between the tumor edge segmentation module and the non-edge area segmentation module of tumor tissue, a mutual learning module is set up to control the mutual learning between the tumor tissue edge segmentation and the non-edge area segmentation of tumor tissue; the model structure composed of the tumor edge segmentation module, the non-edge area segmentation module of tumor tissue and the mutual learning module is trained to obtain a tumor tissue segmentation model for segmenting tumor tissue in MR images. A segmentation output unit is used to obtain the tumor tissue segmentation result of the MR image by using the tumor tissue segmentation model.

[0014] As a preferred solution of the present invention, the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the tumor tissue segmentation method based on MR images is implemented.

[0015] The present invention has the following beneficial effects compared with the prior art: By constructing a tumor edge segmentation module, a non-edge area segmentation module of tumor tissue and a mutual learning module, the present invention enables interactive learning and mutual guidance between the edge segmentation and the non-edge area segmentation, and can accurately segment the main area of tumor tissue and accurately segment the complex boundary of tumor tissue, improving the accurate segmentation effect of tumor tissue. Description of the Drawings

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can also be obtained according to the provided drawings.

[0017] Figure 1 It is a flowchart of the tumor tissue segmentation method based on MR images provided by the embodiments of the present invention; Figure 2 It is a block diagram of the tumor tissue segmentation system provided by the embodiments of the present invention; Figure 3 It is a schematic diagram of the tumor tissue segmentation model structure provided by the embodiments of the present invention. Detailed Embodiments

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] like Figure 1 and Figure 3 As shown, the present invention provides a tumor tissue segmentation method based on MR images, comprising the following steps: Acquiring MR images containing tumor tissue; A tumor edge segmentation module for segmenting the edge of tumor tissue in MR images and a tumor non-edge region segmentation module for segmenting the non-edge region of tumor tissue in MR images are constructed by using a neural network; Between the tumor edge segmentation module and the tumor non-edge area segmentation module, a mutual learning module is provided for controlling the mutual learning between the tumor tissue edge segmentation and the tumor tissue non-edge area segmentation; The model structure composed of a tumor edge segmentation module, a tumor non-edge area segmentation module and a mutual learning module is trained to obtain a tumor tissue segmentation model for segmenting tumor tissue in MR images.

[0020] The present invention divides the overall segmentation of tumor tissue in MR images into two local segmentations, namely, tumor tissue edge segmentation and tumor tissue non-edge area segmentation, and finally superimposes and fuses the two local segmentation results to obtain the overall segmentation result of the tumor tissue.

[0021] After the overall segmentation is divided into two local segmentations, the overall segmentation process needs to balance the two focus points of edge segmentation and main segmentation, that is, in the overall segmentation process, edge segmentation and main segmentation have mutual compromise, so that it is difficult for the overall segmentation to achieve the most accurate performance in edge segmentation and main segmentation. The present invention changes the overall segmentation into two segmentation processes with independent focus points, establishes a tumor edge segmentation module and a tumor non-edge area segmentation module, which respectively correspond to being able to achieve the most accurate performance in edge segmentation and the most accurate performance in the segmentation of the main area, so that the overall segmentation effect after superposition also has the most accurate performance of edge segmentation and main segmentation, and realizes high-precision segmentation of tumor tissue in MR images.

[0022] Since the tumor edge and the non-edge region of the tumor are adjacent and blend with each other, rather than being completely separated, there will also be a certain degree of correlation and constraint in the segmentation of the tumor edge and the non-edge region of the tumor. That is to say, the edge segmentation can assist or constrain the segmentation of the outer perimeter of the non-edge region of the tumor. Correspondingly, the outer perimeter of the non-edge region of the tumor can also assist or constrain the edge segmentation. The results of these two segmentation processes, namely the edge segmentation and the segmentation of the outer perimeter of the non-edge region of the tumor, are similar and can learn from each other. Therefore, the present invention sets a mutual learning unit between the tumor edge segmentation module and the non-edge region segmentation module of the tumor to enable the mutual learning of the similarity process between the tumor edge segmentation module and the non-edge region segmentation module of the tumor, corresponding to the first path and the second path, and quantifies the learning rate, that is, the degree of difference between the results of the tumor edge segmentation module and the non-edge region segmentation module of the tumor. The lower the degree of difference, the higher the learning rate of the first path or the second path, and the learning effect of the first path or the second path reaches the expectation. The higher the degree of difference, the lower the learning rate of the first path or the second path, and the learning effect of the first path or the second path does not reach the expectation.

[0023] When training a tumor tissue segmentation model that includes a tumor edge segmentation module, a non-edge region segmentation module of the tumor, and a mutual learning module, the present invention uses a loss function that includes three parts. The first part is the independent prediction loss of the tumor edge segmentation module and the non-edge region segmentation module of the tumor, that is, the loss between the output result and the edge mask ground truth and the main region mask. Training the model with this as the loss can achieve the most accurate performance in edge segmentation and the most accurate performance in the segmentation of the main region.

[0024] The second part is the reconstruction loss of the mutual fusion of the tumor edge segmentation module and the non-edge region segmentation module of the tumor, that is, the loss between the fusion mask of the output results of the tumor edge segmentation module and the non-edge region segmentation module of the tumor and the tumor tissue overall mask ground truth. Training the model with this as the loss can ensure that the overall segmentation effect after superposition also reaches the most accurate performance of the overall segmentation.

[0025] The third part is the mutual learning loss, which consists of the learning rates of the first path and the second path. Using this as the loss can ensure that the learning rates of the first path and the second path reach the highest, and the learning effects of the first path and the second path reach the expectation, making full use of the correlation and constraint that also exist in the segmentation of the tumor edge and the non-edge region of the tumor, and more quickly completing the segmentation of the similar parts of the tumor edge segmentation module and the non-edge region segmentation module of the tumor, improving the segmentation efficiency.

[0026] Furthermore, in order to balance the learning progress of the two paths, the present invention sets hyperparameters , is the hyperparameter of the first path, is the hyperparameter of the second path. It is quantified by the training progress of the tumor edge segmentation module (characterized by prediction loss). The smaller the prediction loss, the faster the training progress is completed. Similarly, It is quantified by the training progress of the tumor non-edge area segmentation module (characterized by prediction loss). The smaller the prediction loss, the faster the training progress is completed. The present invention expects that the learning progress of the two paths can be kept consistent as much as possible, so as to utilize the correlation and constraints existing in the segmentation of the tumor edge and the tumor non-edge area. Therefore, the training progress of the tumor non-edge area segmentation module and the tumor edge segmentation module needs to be relatively consistent. If it is inconsistent, it will cause deviations in correlation and constraints, resulting in invalid application.

[0027] In this regard, the present invention and It can be achieved that when the training progress of the tumor edge segmentation module or the tumor non-edge area segmentation module is fast, the first path or the second path is assigned a low learning rate weight, or when the training progress of the tumor edge segmentation module or the tumor non-edge area segmentation module is slow, the first path or the second path is assigned a high learning rate weight. For example, when the training progress of the tumor edge segmentation module is fast, the first path is assigned a low learning rate weight, and when the training progress of the tumor non-edge area segmentation module is slow, the second path is assigned a high learning rate weight, thereby achieving the balance of the learning rate of the first path and the learning rate of the second path in the mutual learning loss, achieving the effect of balancing the learning progress of the two paths, and ensuring the effectiveness of mutual learning.

[0028] The present invention transforms the overall segmentation into two segmentation processes with independent focus points, establishes a tumor edge segmentation module and a tumor non-edge area segmentation module, which can respectively achieve the most accurate performance in edge segmentation and the most accurate performance in the segmentation of the main area, so that the overall segmentation effect after superposition also has the most accurate performance of edge segmentation and main area segmentation, and realizes high-precision segmentation of tumor tissue in MR images, as follows: The construction method of the tumor edge segmentation module includes: The Mask F-CNN network is used as the model structure of the tumor edge segmentation module; The MR image is used as the input of the Mask F-CNN network, and the edge mask of the tumor tissue in the MR image is used as the output of the Mask F-CNN network. The tumor edge segmentation module is obtained as follows: G rough =Mask F-CNN(MR); In the formula, G rough is the edge mask, MR is the MR image, and Mask F-CNN is the Mask F-CNN network.

[0029] The construction method of the tumor non-edge region segmentation module includes: Using the Mask F-CNN network as the model structure of the tumor non-edge segmentation module; Taking the MR image as the input of the Mask F-CNN network, and taking the non-edge region mask of the tumor tissue in the MR image as the output of the Mask F-CNN network, the tumor non-edge region segmentation module obtained is: G body =Mask F-CNN(MR); In the formula, G body is the non-edge region mask, MR is the MR image, and Mask F-CNN is the Mask F-CNN network.

[0030] In the present invention, a mutual learning unit is set between the tumor edge segmentation module and the tumor non-edge region segmentation module for the mutual learning similarity process between the tumor edge segmentation module and the tumor non-edge region segmentation module, corresponding to the first path and the second path, and the learning rate is quantified as follows: The construction method of the mutual learning module includes: Constructing the first path for the tumor edge segmentation module to learn from the tumor non-edge region segmentation module, and quantifying the learning rate of the first path; The learning rate of the first path is: ; In the formula, is the learning rate of the first path, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used for training the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module on the i-th sample in the dataset used for training the model structure, m is the total number of samples in the dataset used for training the model structure, is and the KL divergence between; Constructing the second path for the tumor non-edge region segmentation module to learn from the tumor edge segmentation module, and quantifying the learning rate of the second path; The learning rate of the second path is: ; In the formula, is the learning rate of the second path, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used for training the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module on the i-th sample in the dataset used for training the model structure, m is the total number of samples in the dataset used for training the model structure, The KL divergence between and is ; .

[0031] The training loss function of the model structure is: ; where ; ; ; In the formula, is the total loss, is the prediction loss, is the mutual learning loss, is the reconstruction loss, , and are the hyperparameters for balancing , and respectively, is the edge mask output by the tumor edge segmentation module for the i-th sample in the dataset used to train the model structure, is the non-edge region mask output by the tumor non-edge region segmentation module for the i-th sample in the dataset used to train the model structure, m is the total number of samples in the dataset used to train the model structure, is the ground truth of the non-edge region mask for the i-th sample in the dataset used to train the model structure, is the ground truth of the edge mask for the i-th sample in the dataset used to train the model structure, is the ground truth of the overall tumor tissue mask for the i-th sample in the dataset used to train the model structure, is the learning rate of the first path, is the learning rate of the second path, , and are all L2 norm expressions, and here is set, , and it can also be reset according to actual needs in actual use.

[0032] When training a tumor tissue segmentation model that includes a tumor edge segmentation module, a non-edge region segmentation module of the tumor, and a mutual learning module, a loss function consisting of three parts is adopted. The first part is the independent prediction loss of the tumor edge segmentation module and the non-edge region segmentation module of the tumor, that is, the loss between the output result and the edge mask ground truth and the main region mask. Using this as the loss to train the model can achieve the most accurate performance in edge segmentation and the most accurate performance in the segmentation of the main region.

[0033] The second part is the reconstruction loss of the mutual fusion of the tumor edge segmentation module and the non-edge region segmentation module of the tumor, that is, the loss between the fusion mask of the output results of the tumor edge segmentation module and the non-edge region segmentation module of the tumor and the tumor tissue overall mask ground truth. Using this as the loss to train the model can ensure that the overall segmentation effect after superposition also reaches the most accurate performance of the overall segmentation.

[0034] The third part is the mutual learning loss, which consists of the learning rates of the first path and the second path. Using this as the loss can ensure that the learning rates of the first path and the second path reach the highest, and the learning effects of the first path and the second path meet the expectations. It can make full use of the correlation and constraint existing in the segmentation of the tumor edge and the non-edge region of the tumor, and more quickly complete the segmentation of the similar parts of the tumor edge segmentation module and the non-edge region segmentation module of the tumor, improving the segmentation efficiency.

[0035] The method for obtaining a tumor tissue segmentation model by training the model structure includes: Dividing the data set into a training set and a test set; Based on the loss function, training the model structure on the training set to obtain a tumor tissue segmentation model; Based on the model evaluation index, evaluating the performance of the tumor tissue segmentation model on the test set.

[0036] The method for the tumor tissue segmentation model to obtain the tumor tissue segmentation result of the MR image includes: Inputting the MR image into the tumor tissue segmentation model to obtain an edge mask and a non-edge region mask; Overlaying and fusing the edge mask and the non-edge region mask to obtain an overall mask of the tumor tissue, which is used as the tumor tissue segmentation result of the MR image.

[0037] Hyperparameter , where ; ; In the formula, is the balancing hyperparameter, is the balancing hyperparameter, is the edge mask output by the tumor edge segmentation module for the i-th sample in the dataset used for training the model structure. is the non-edge region mask output by the tumor non-edge region segmentation module for the i-th sample in the dataset used for training the model structure, and m is the total number of samples in the dataset used for training the model structure. is the ground truth of the non-edge region mask for the i-th sample in the dataset used for training the model structure. is the ground truth of the edge mask for the i-th sample in the dataset used for training the model structure. is the learning rate of the first path. is the learning rate of the second path. 、 are both L2 norm expressions.

[0038] In the present invention and can achieve that when the training progress of the tumor edge segmentation module or the tumor non-edge region segmentation module is fast, a low weight is assigned to the learning rate of the first path or the second path, or when the training progress of the tumor edge segmentation module or the tumor non-edge region segmentation module is slow, a high weight is assigned to the learning rate of the first path or the second path. For example, when the training progress of the tumor edge segmentation module is fast, a low weight is assigned to the learning rate of the first path, and when the training progress of the tumor non-edge region segmentation module is slow, a high weight is assigned to the learning rate of the second path, so as to achieve the equalization of the learning rates of the first path and the second path in the mutual learning loss, achieve the effect of balancing the learning progress of the two paths, and ensure the effectiveness of mutual learning.

[0039] Such as Figure 2 shown, the present invention provides a tumor tissue segmentation system based on MR images, which is applied to a tumor tissue segmentation method based on MR images. The system includes: A data acquisition unit for acquiring MR images containing tumor tissue. A model establishment unit for respectively constructing, using a neural network, a tumor edge segmentation module for segmenting the edge of tumor tissue in MR images, and a tumor non-edge region segmentation module for segmenting the non-edge region of tumor tissue in MR images; between the tumor edge segmentation module and the tumor non-edge region segmentation module, a mutual learning module for controlling the mutual learning between the tumor tissue edge segmentation and the tumor non-edge region segmentation is set; the model structure composed of the tumor edge segmentation module, the tumor non-edge region segmentation module, and the mutual learning module is trained to obtain a tumor tissue segmentation model for segmenting tumor tissue in MR images. A segmentation output unit for obtaining the tumor tissue segmentation result of the MR image using the tumor tissue segmentation model.

[0040] The present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement a method for segmenting tumor tissues based on MR images.

[0041] By constructing a tumor edge segmentation module, a tumor non-edge region segmentation module, and a mutual learning module, the present invention enables interactive learning and mutual guidance between edge segmentation and non-edge region segmentation, capable of accurately segmenting the main region of tumor tissues and achieving accurate segmentation of the complex boundaries of tumor tissues, thereby improving the accurate segmentation effect of tumor tissues.

[0042] The above embodiments are only exemplary embodiments of the present application and are not used to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements within the essence and protection scope of the present application, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present application.

Claims

1. A tumor tissue segmentation method based on MR images, characterized in that: The following steps are involved: Acquiring MR images containing tumor tissue; A tumor edge segmentation module for segmenting the edge of tumor tissue in MR images and a tumor non-edge region segmentation module for segmenting the non-edge region of tumor tissue in MR images are constructed by using a neural network; Between the tumor edge segmentation module and the tumor non-edge region segmentation module, a mutual learning module is provided for controlling the mutual learning between the tumor tissue edge segmentation and the tumor tissue non-edge region segmentation; The model structure composed of a tumor edge segmentation module, a tumor non-edge area segmentation module and a mutual learning module is trained to obtain a tumor tissue segmentation model for segmenting tumor tissue in MR images.

2. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The method for constructing the tumor edge segmentation module includes: The Mask F-CNN network is used as the model structure of the tumor edge segmentation module; The MR image is used as the input of the Mask F-CNN network, and the edge mask of the tumor tissue in the MR image is used as the output of the MaskF-CNN network, and the tumor edge segmentation module is obtained as follows: G rough =Mask F-CNN(MR); In the formula, G rough is the edge mask, MR is the MR image, and Mask F-CNN is the Mask F-CNN network.

3. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The method for constructing the tumor non-marginal area segmentation module comprises: The Mask F-CNN network is used as the model structure of the tumor non-edge segmentation module; The MR image is used as the input of the Mask F-CNN network, and the non-edge area mask of the tumor tissue in the MR image is used as the output of the Mask F-CNN network, and the tumor non-edge area segmentation module is obtained as follows: G body =Mask F-CNN(MR); In the formula, G body is the non-edge region mask, MR is the MR image, and Mask F-CNN is the Mask F-CNN network.

4. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The method for constructing the mutual learning module includes: Constructing a first path for the tumor edge segmentation module to learn from the tumor non-edge area segmentation module, and quantifying the learning rate of the first path; The learning rate of the first path is: ； In the formula, is the learning rate of the first path, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used to train the model structure, is the non-marginal area mask output by the tumor non-marginal area segmentation module on the i-th sample in the data set used to train the model structure, m is the total number of samples in the data set used to train the model structure, for and The KL divergence between Constructing a second path for the tumor non-margin area segmentation module to learn from the tumor margin segmentation module, and quantifying the learning rate of the second path; The learning rate of the second path is: ； In the formula, is the learning rate of the second path, is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used to train the model structure, is the non-marginal area mask output by the tumor non-marginal area segmentation module on the i-th sample in the data set used to train the model structure, m is the total number of samples in the data set used to train the model structure, for and The KL divergence between in, ; 。 5. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The training loss function of the model structure is: ； in, ; ； ； In the formula, is the total loss, To predict the loss, is the mutual learning loss, To rebuild the losses, , and Balance , and The hyperparameters of is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used to train the model structure, is the non-marginal area mask output by the tumor non-marginal area segmentation module on the i-th sample in the data set used to train the model structure, m is the total number of samples in the data set used to train the model structure, The true value of the non-edge area mask on the i-th sample in the dataset used to train the model structure, The true value of the edge mask on the i-th sample in the dataset used to train the model structure, is the true value of the overall mask of the tumor tissue on the i-th sample in the data set used to train the model structure, is the learning rate of the first path, is the learning rate of the second path, , and All are L2 norm forms.

6. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The method of training the model structure to obtain the tumor tissue segmentation model includes: Dividing the data set into a training set and a test set; Based on the loss function, the model structure is trained on a training set to obtain a tumor tissue segmentation model; Based on the model evaluation indicators, the performance of the tumor tissue segmentation model is evaluated on the test set.

7. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The method for obtaining the tumor tissue segmentation result of the MR image by the tumor tissue segmentation model comprises: Input the MR image into the tumor tissue segmentation model to obtain edge masks and non-edge area masks; The edge mask and the non-edge area mask are superimposed and fused to obtain the overall tumor tissue mask as the tumor tissue segmentation result of the MR image.

8. The method for tumor tissue segmentation based on MR images according to claim 1, characterized in that: The hyperparameters ,in, ； ； In the formula, for The balanced hyperparameters of for The balanced hyperparameters of is the edge mask output by the tumor edge segmentation module on the i-th sample in the dataset used to train the model structure, is the non-marginal area mask output by the tumor non-marginal area segmentation module on the i-th sample in the data set used to train the model structure, m is the total number of samples in the data set used to train the model structure, The true value of the non-edge area mask on the i-th sample in the dataset used to train the model structure, The true value of the edge mask on the i-th sample in the dataset used to train the model structure, is the learning rate of the first path, is the learning rate of the second path, , All are L2 norm forms.

9. A tumor tissue segmentation system based on MR images, characterized in that: A method for segmenting tumor tissue based on MR images as described in any one of claims 1 to 8, the system comprising: A data acquisition unit, used for acquiring an MR image containing tumor tissue; A model building unit is used to use a neural network to respectively construct a tumor edge segmentation module for segmenting the edge of tumor tissue in an MR image, and a tumor non-edge region segmentation module for segmenting the non-edge region of tumor tissue in an MR image; between the tumor edge segmentation module and the tumor non-edge region segmentation module, a mutual learning module is set to control the mutual learning of tumor tissue edge segmentation and tumor tissue non-edge region segmentation; a model structure composed of the tumor edge segmentation module, the tumor non-edge region segmentation module and the mutual learning module is trained to obtain a tumor tissue segmentation model for segmenting tumor tissue in MR images; The segmentation output unit is used to obtain the tumor tissue segmentation result of the MR image by using the tumor tissue segmentation model.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when a processor executes the computer-executable instructions, the method according to any one of claims 1 to 8 is implemented.

Images