Medical image segmentation method for intra-class region dynamic decoupling

Abstract

The invention relates to the technical field of digital image processing, in particular to an intra-class region dynamic decoupling medical image segmentation method, which comprises the following steps: S1, preprocessing a data set; s2, establishing a three-branch baseline segmentation network: establishing a student model and a teacher model; s3, in-class region decoupling: the blood vessel classes are decoupled into difficult sub-regions and easy sub-regions, and different constraints are applied to the difficult sub-regions and the easy sub-regions; s4, constructing a dynamic threshold module; s5, establishing a constraint optimization network: constraining the student model through the teacher model; s6, training a semi-supervised blood vessel segmentation network; and S7, target segmentation: segmenting a blood vessel region through the semi-supervised blood vessel segmentation network. According to the invention, through intra-class region decoupling, a blood vessel image is divided into a difficult sub-region and an easy sub-region; different optimization modes are adopted for different areas to guide the model to learn different intra-class features, and the intra-class imbalance problem is relieved. Different regions in the class are accurately decoupled through a dynamic threshold module, and a learnable threshold is generated for the image.

Description

Technical Field

[0001] The invention relates to the technical field of digital image processing, and in particular to a medical image segmentation method for dynamic decoupling of intra-class regions. Background Art

[0002] Accurately segmenting blood vessels from computed tomography (CT) is an important prerequisite for diagnosing and treating various diseases. Unlike other organs with a single shape and simple structure, blood vessels have a complex tree-like structure, so complete and accurate blood vessel labels are difficult to obtain. Semi-supervised medical image segmentation, as a method with low labeled data requirements, greatly alleviates the dependence of neural networks on labeled data. However, there are huge intra-class differences in the number of voxels, the number of branches, and the morphological structure between the main blood vessels and the peripheral microvessels. This serious intra-class distribution imbalance problem hinders the practical application of semi-supervised methods in blood vessel segmentation. Therefore, it is worth studying and designing a general semi-supervised medical image segmentation method.

[0003] The existing partial intra-class region dynamic decoupling methods use traditional image processing technology to perform simple image segmentation based on the grayscale value of the image. However, when processing complex vascular structures, since the structure of blood vessels is often complex and changeable and there are individual differences, traditional image processing technology may have difficulty in accurately identifying and distinguishing different regions.

[0004] In summary, the existing methods for dynamic decoupling of some intra-class regions use traditional image processing technology, which may make it difficult to accurately identify and distinguish different regions. Therefore, it is necessary to propose a medical image segmentation method for dynamic decoupling of intra-class regions. Summary of the invention

[0005] To solve the above problems, the present invention provides a medical image segmentation method with dynamic decoupling of intra-class regions. Through the decoupling of intra-class regions, the vascular image is divided into difficult sub-regions and easy sub-regions. Different optimization methods are adopted for different regions to guide the model to learn different intra-class features, thereby alleviating the problem of intra-class imbalance. The dynamic threshold module is used to accurately decouple different regions within the class, and a learnable threshold is generated for the image during training.

[0006] In order to achieve the above object, the technical solution of the present invention is as follows: a medical image segmentation method with dynamic decoupling of intra-class regions, comprising the following steps: S1, dataset preprocessing: dataset images are collected through the Internet, the voxel intensities of all dataset images are truncated and normalized, and data enhancement is performed on them to obtain enhanced data. At the same time, the images containing the blood vessels to be segmented are divided into bounding boxes.

[0007] S2, establish a three-branch baseline segmentation network: use the U-net as the baseline network for segmentation, establish several student models and teacher models, and combine the student model and the teacher model into a student-student model and a teacher-student model. All student models and teacher models are 3D U-Net networks with the same architecture but different initialization perturbations.

[0008] S3, decoupling intra-class regions: the enhanced data is divided into labeled data and unlabeled data, and the blood vessel category is decoupled into difficult sub-regions and easy sub-regions, and different constraints are imposed on the difficult sub-regions and easy sub-regions.

[0009] S4, build a dynamic threshold module: build a dynamic threshold module and use it to accurately decouple different difficult sub-areas and easy sub-areas.

[0010] S5, establish a constrained optimization network: all student models learn to represent different difficult sub-regions and easy sub-regions, and the teacher model constrains the consistency of all student models.

[0011] S6, training semi-supervised blood vessel segmentation network: using the loss function to utilize labeled data and unlabeled data to train the semi-supervised blood vessel segmentation network.

[0012] S7, segmentation target: input the image to be tested, and the semi-supervised blood vessel segmentation network tests the image to be tested to complete the segmentation of the blood vessel area.

[0013] Furthermore, in S1, the minimum bounding box containing the blood vessel to be segmented is used as the range of the region of interest of the student model.

[0014] Furthermore, in S1, the data enhancement method uses one or more of random horizontal flipping, vertical flipping, and random -90 degree to 90 degree rotation.

[0015] Furthermore, in S2, the teacher model is the self-integration of the average of the student models, obtained by exponential moving average.

[0016] Furthermore, in S2, the expression of the student model is as follows: (1) (2) The expression of the teacher model is as follows: (3) in, are model parameters, and All are student models. Model for teachers.

[0017] Furthermore, in S3, the unlabeled data includes probability maps generated by the student model and noisy pseudo labels.

[0018] When performing the decoupling operation, the predicted probability map of one of the student models and the noisy pseudo-label of the other student model are sent to the dynamic threshold module to obtain the first dynamic threshold; when the consistency degree of the voxels of the probability map generated by the other student model is greater than the first dynamic threshold, it is considered that the student model has reached consistency in this area, and the area where the student model is located is the easy sub-area.

[0019] The prediction consistency of the student-student model is used to obtain the easy sub-region of the vascular image, and the information of the easy sub-region is stored in the confidence map. The specific formula of the process is as follows: (4) (5) in, is the indicator function, is the dynamic threshold module, and is a probability map and noisy pseudo-label generated by one of the student models, and is a probability map and noisy pseudo-label generated by another student model, is the first dynamic threshold for each category, is the regional confidence map, is the degree of consistency between student models.

[0020] Furthermore, in S3, when the decoupling operation is performed, the probability map of the teacher model and the average value of the noisy pseudo labels of all student models are sent to the dynamic threshold module to obtain a second dynamic threshold. If the consistency degree of the voxels of the teacher model is less than the second dynamic threshold, it is considered that there is ambiguity here and it is a difficult sub-region.

[0021] The prediction uncertainty of the teacher-student model is used to obtain the difficult sub-regions of the vascular image, and the information of the difficult sub-regions is stored in the uncertainty map. The specific formula of the process is as follows: (6) (7) (8) in, is the probability graph of the teacher model, for and The average value of is the second dynamic threshold for each category, is the uncertainty diagram, is the consistency of the teacher model.

[0022] Furthermore, in S4, when performing the precise decoupling operation, the pseudo-label is first overlaid on the probability map and the voxels of the corresponding categories in the pseudo-label of each category in the probability map are found. Finally, the threshold of each category is calculated separately. The specific formula is as follows: (9) (10) in, is the indicator function, M represents mask, represents a noisy pseudo label, represents a probability graph, Indicates the threshold value.

[0023] Furthermore, in S5, for the easy sub-regions, pseudo-label cross-supervision is performed between different student models, and the pseudo-label cross-supervision loss is defined as follows: (11) (12) (13) in, is the dice loss, and They are and The pseudo labels generated are and They are and The generated probability map, is the total number of data in the dataset, is the number of labeled data, Model for students Easy region confidence map with noisy pseudo labels.

[0024] For difficult sub-regions, a teacher model is introduced to calculate the consistency loss. The parameters of the teacher model are updated by exponential moving average. The specific formula is as follows: (14) in, is the number of iterations of the current training, and Hyperparameters that control the exponential moving average updates.

[0025] The formula for consistency loss is as follows: (15) in, is the squared loss function, For unlabeled data, and are the model parameters of the student model, are the model parameters of the teacher model.

[0026] Furthermore, in S6, the loss function includes supervised loss and unsupervised loss. The specific formula of the loss function is as follows: (16) in, represents the supervision loss, represents the unsupervised loss, is a balance parameter used to balance supervised loss and unsupervised loss.

[0027] The specific formula of supervision loss is as follows: (17) in, is a linear combination of cross entropy loss and dice loss, For labeled data, is the label of the labeled data.

[0028] The unsupervised loss includes the pseudo-label cross-supervision loss between student models and the consistency loss between teacher and student models. The specific formula is as follows: (18) in, represents the pseudo-label cross-supervision loss, Indicates consistency loss.

[0029] The above scheme has the following beneficial effects: 1. As a general medical image segmentation method, this scheme can achieve accurate segmentation of various medical images. Through intra-class region decoupling, the vascular image is separated into difficult sub-regions and easy sub-regions. Different optimization methods are adopted for different regions to guide the model to learn different intra-class features, thereby alleviating the serious intra-class imbalance problem. Through the dynamic threshold module based on confidence learning, different regions within the class are accurately decoupled. The dynamic threshold module generates adaptive and learnable thresholds for images during training.

[0030] 2. This scheme, by limiting the scope of the student model's area of ​​interest to the minimum bounding box containing the blood vessels to be segmented, enables the student model to focus on processing this part of the image data, thereby significantly reducing the amount of data that needs to be processed, thereby reducing the computational complexity and speeding up the image processing.

[0031] 3. This solution can simulate observations from different angles and directions by randomly rotating and flipping images, thereby increasing the diversity of the data. This diversity helps the teacher model and the student model learn more comprehensive features and improve the generalization ability of the teacher model and the student model. At the same time, operations such as random rotation and flipping can force the teacher model and the student model to learn more robust feature representations that remain consistent under different transformations, which helps improve the performance of the teacher model and the student model on the test set.

[0032] 4. In this scheme, by integrating the student model parameters in multiple training steps, the teacher model can capture a wider range of feature representations. Since these feature representations have good generalization capabilities on different data sets and test sets, the teacher model can better cope with unseen data and improve the generalization performance of the teacher model.

[0033] 5. This scheme combines the student model with the teacher model and uses the dynamic threshold module to decouple the difficult and easy sub-regions of the vascular category. This method can more accurately identify and process complex areas in the image. This fine-grained processing strategy makes the segmentation results more accurate.

[0034] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The figure is a flow chart of the medical image segmentation method with dynamic decoupling of intra-class regions according to the present invention.

[0036] Figure 2 A schematic diagram of the process of establishing a student model and a teacher model in the medical image segmentation method with dynamic decoupling of intra-class regions of the present invention.

[0037] Figure 3 A schematic diagram of the classification of blood vessel categories in the medical image segmentation method with dynamic decoupling of intra-class regions of the present invention. DETAILED DESCRIPTION

[0038] The following is further described in detail through specific implementation methods: Embodiment 1:

[0039] As attached Figure 1 , Figure 2 and Figure 3 As shown: A medical image segmentation method with dynamic decoupling of intra-class regions, comprising the following steps: S1, dataset preprocessing: dataset images are collected through the Internet, the voxel intensities of all dataset images are truncated and normalized, and data enhancement is performed on them to obtain enhanced data. At the same time, the images containing the blood vessels to be segmented are divided into bounding boxes.

[0040] S2, establish a three-branch baseline segmentation network: use the U-net as the baseline network for segmentation, establish several student models and teacher models, and combine the student model and the teacher model into a student-student model and a teacher-student model. All student models and teacher models are 3D U-Net networks with the same architecture but different initialization perturbations.

[0041] The expression of the student model is as follows: (1) (2) The expression of the teacher model is as follows:

[0042] (3) in, are model parameters, and All are student models. Model for teachers.

[0043] S3, decouple intra-class regions: The enhanced data is divided into labeled data and unlabeled data, and the blood vessel category is decoupled into difficult sub-regions and easy sub-regions, and different constraints are imposed on the difficult sub-regions and easy sub-regions. Among them, the unlabeled data includes the probability map generated by the student model and the noisy pseudo-label.

[0044] S4, build a dynamic threshold module: build a dynamic threshold module and use it to accurately decouple different difficult sub-areas and easy sub-areas.

[0045] When performing the decoupling operation, the predicted probability map of one of the student models and the noisy pseudo-label of the other student model are sent to the dynamic threshold module to obtain the first dynamic threshold; when the consistency degree of the voxels of the probability map generated by the other student model is greater than the first dynamic threshold, it is considered that the student model has reached consistency in this area, and the area where the student model is located is the easy sub-area.

[0046] The prediction consistency of the student-student model is used to obtain the easy sub-region of the vascular image, and the information of the easy sub-region is stored in the confidence map. The specific formula of the process is as follows: (4) (5) in, is the indicator function, is the dynamic threshold module, and is a probability map and noisy pseudo-label generated by one of the student models, and is a probability map and noisy pseudo-label generated by another student model, is the first dynamic threshold for each category, is the regional confidence map, is the degree of consistency between student models.

[0047] The probability map of the teacher model and the average of the noisy pseudo labels of all student models are sent to the dynamic threshold module to obtain the second dynamic threshold. If the consistency of the voxels of the teacher model is less than the second dynamic threshold, it is considered that there is ambiguity here and it is a difficult sub-region.

[0048] The prediction uncertainty of the teacher-student model is used to obtain the difficult sub-regions of the vascular image, and the information of the difficult sub-regions is stored in the uncertainty map. The specific formula of the process is as follows: (6) (7) (8) in, is the probability graph of the teacher model, for and The average value of is the second dynamic threshold for each category, is the uncertainty diagram, is the consistency of the teacher model.

[0049] When performing precise decoupling operations, first overlay the pseudo labels onto the probability map and find the voxels of the corresponding categories in the pseudo labels for each category in the probability map. Finally, calculate the threshold for each category separately. The specific formula is as follows: (9) (10) in, is the indicator function, M represents mask, represents a noisy pseudo label, represents a probability graph, Indicates the threshold value.

[0050] S5, establish a constrained optimization network: all student models learn to represent different difficult sub-regions and easy sub-regions, and the teacher model constrains the consistency of all student models.

[0051] For easy sub-regions, pseudo-label cross-supervision is performed between different student models, and the pseudo-label cross-supervision loss is defined as follows: (11) (12) (13) in, is the dice loss, and They are and The pseudo labels generated are and They are and The generated probability map, is the total number of data in the dataset, is the number of labeled data, Model for students Easy region confidence map with noisy pseudo labels.

[0052] For difficult sub-regions, a teacher model is introduced to calculate the consistency loss. The parameters of the teacher model are updated by exponential moving average. The specific formula is as follows: (14) in, is the number of iterations of the current training, and Hyperparameters that control the exponential moving average updates.

[0053] The formula for consistency loss is as follows: (15) in, is the squared loss function, For unlabeled data, and are the model parameters of the student model, are the model parameters of the teacher model.

[0054] S6, training semi-supervised blood vessel segmentation network: using the loss function to utilize labeled data and unlabeled data to train the semi-supervised blood vessel segmentation network.

[0055] The loss function includes supervised loss and unsupervised loss. The specific formula of the loss function is as follows: (16) in, represents the supervision loss, represents the unsupervised loss, is a balance parameter used to balance supervised loss and unsupervised loss.

[0056] The specific formula of supervision loss is as follows: (17) in, is a linear combination of cross entropy loss and dice loss, For labeled data, is the label of the labeled data; The unsupervised loss includes the pseudo-label cross-supervision loss between student models and the consistency loss between teacher and student models. The specific formula is as follows: (18) in, represents the pseudo-label cross-supervision loss, Indicates consistency loss.

[0057] S7, segmentation target: input the image to be tested, and the semi-supervised blood vessel segmentation network tests the image to be tested to complete the segmentation of the blood vessel area.

[0058] The specific implementation process is as follows: In this embodiment, the axial size of all slices of the image is 512×512, and the spatial resolution is 0.5mm-1.0mm. The image labels are obtained by carefully annotating the images by 3 radiologists with more than 5 years of clinical experience.

[0059] First, we collected dataset images from the Internet, truncated the voxel intensity of all dataset images within the window of [-1000,600] Hounsfield units, normalized them to [0,1], and performed data enhancement processing on them to obtain enhanced data. At the same time, we divided the images containing the blood vessels to be segmented into bounding boxes. Using the U-net as the baseline network for segmentation, we established several student models and teacher models, and combined the student models and teacher models into student-student models and teacher-student models. All the student models and teacher models established were 3D U-Net networks with the same architecture but different initialization perturbations.

[0060] In this embodiment, the number of student models is two, and the expression of the student model is as follows: (1) (2) The expression of the teacher model is as follows: (3) in, are model parameters, and All are student models. Model for teachers.

[0061] After the student model and the teacher model are established, the enhanced data are divided into labeled data and unlabeled data. At the same time, the blood vessel categories are decoupled into difficult sub-regions and easy sub-regions, and different constraints are imposed on the difficult sub-regions and the easy sub-regions.

[0062] After completing the decoupling of the intra-class regions, a dynamic threshold module is constructed, and the dynamic threshold module is used to accurately decouple different difficult sub-regions and easy sub-regions.

[0063] When decoupling the vascular category into difficult sub-regions and easy sub-regions, the prediction consistency of the student-student model is first used to obtain the easy sub-region of the vascular image, and the information of the easy sub-region is stored in the confidence map. The predicted probability map of one student model and the noisy pseudo-label of the other student model are sent to the dynamic threshold module to obtain the first dynamic threshold; when the consistency of the voxels of the probability map generated by the other student model is greater than the first dynamic threshold, it is considered that the student model has reached consistency in this area, and the area where the student model is located is the easy sub-region.

[0064] The specific formula of this process is as follows: (4) (5) in, is the indicator function, is the dynamic threshold module, and is a probability map and noisy pseudo-label generated by one of the student models, and is a probability map and noisy pseudo-label generated by another student model, is the first dynamic threshold for each category, is the regional confidence map, is the degree of consistency between student models.

[0065] Subsequently, the prediction uncertainty of the teacher-student model is used to obtain the difficult sub-region of the vascular image, and the information of the difficult sub-region is stored in the uncertainty map. The probability map of the teacher model and the average of the noisy pseudo-labels of all student models are sent to the dynamic threshold module to obtain the second dynamic threshold. If the consistency of the voxels of the teacher model is less than the second dynamic threshold, it is considered that there is ambiguity here and it is a difficult sub-region.

[0066] The specific formula of this process is as follows: (6) (7) (8) in, is the probability graph of the teacher model, for and The average value of is the second dynamic threshold for each category, is the uncertainty diagram, is the consistency of the teacher model.

[0067] When performing precise decoupling operations, first overlay the pseudo labels onto the probability map and find the voxels of the corresponding categories in the pseudo labels for each category in the probability map. Finally, calculate the threshold for each category separately. The specific formula is as follows: (9) (10) in, is the indicator function, M represents mask, represents a noisy pseudo label, represents a probability graph, Indicates the threshold value.

[0068] Subsequently, a constrained optimization network is established to enable all student models to further learn the representation of different difficult and easy sub-regions, and the consistency constraints are imposed on all student models through the teacher model.

[0069] For easy sub-regions, pseudo-label cross-supervision is performed between different student models, and the pseudo-label cross-supervision loss is defined as follows: (11) (12) (13) in, is the dice loss, and They are and The pseudo labels generated are and They are and The generated probability map, is the total number of data in the dataset, is the number of labeled data, Model for students Easy region confidence map with noisy pseudo labels.

[0070] For difficult sub-regions, a teacher model is introduced to calculate the consistency loss. The parameters of the teacher model are updated by exponential moving average. The specific formula is as follows: (14) in, is the number of iterations of the current training, and Hyperparameters that control the exponential moving average updates.

[0071] The formula for consistency loss is as follows: (15) in, is the squared loss function, For unlabeled data, and are the model parameters of the student model, are the model parameters of the teacher model.

[0072] Finally, the loss function is used to utilize labeled data and unlabeled data to train the semi-supervised vascular segmentation network. After the staff inputs the test image into the semi-supervised vascular segmentation network, the semi-supervised vascular segmentation network tests the image and completes the segmentation of the vascular area.

[0073] The loss function includes supervised loss and unsupervised loss. The specific formula of the loss function is as follows: (16) in, represents the supervision loss, represents the unsupervised loss, is a balance parameter used to balance supervised loss and unsupervised loss.

[0074] The specific formula of supervision loss is as follows: (17) in, is a linear combination of cross entropy loss and dice loss, For labeled data, is the label of the labeled data.

[0075] The unsupervised loss includes the pseudo-label cross-supervision loss between student models and the consistency loss between teacher and student models. The specific formula is as follows: (18) in, represents the pseudo-label cross-supervision loss, Indicates consistency loss.

[0076] Embodiment 2: As attached Figure 1 As shown, the difference from Example 1 is that in S1, the minimum bounding box containing the blood vessel to be segmented is used as the range of the region of interest of the student model.

[0077] The specific implementation process is as follows: when processing a smaller data set, limiting the scope of the student model's area of ​​interest to the vascular region can reduce the risk of overfitting of the model. At the same time, the student model can focus more on learning the features of the vascular region rather than multiple features in the entire image, so that the student model can more accurately identify vascular boundaries and improve segmentation accuracy.

[0078] Embodiment 3:

[0079] As attached Figure 1 As shown, the difference from Example 2 is that in S1, the data enhancement method adopts one or more of random horizontal flipping, vertical flipping and random -90 degree to 90 degree rotation, and this embodiment adopts a combination of horizontal flipping and vertical flipping.

[0080] The specific implementation process is as follows: operations such as random horizontal flipping, vertical flipping, and random rotation can be flexibly combined to generate more diverse training samples. This flexibility allows the data augmentation method to be customized and optimized according to the characteristics of specific tasks and datasets.

[0081] Embodiment 4:

[0082] As attached Figure 2 As shown, the difference from Example 3 is that in S2, the teacher model is a self-integration of the average value of the student model, obtained by exponential moving average.

[0083] The specific implementation process is as follows: When processing complex data, such as noisy data, the teacher model can mitigate the impact of data complexity on model performance through its stability and robustness. At the same time, the teacher model can also capture the potential patterns and regularities in the data, thereby helping the student model to better understand and process the data.

[0084] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A medical image segmentation method with dynamic decoupling of intra-class regions, characterized in that: The following steps are involved: S1, dataset preprocessing: dataset images are collected through the Internet, the voxel intensities of all dataset images are truncated and normalized, and data enhancement is performed on them to obtain enhanced data. At the same time, the images containing the blood vessels to be segmented are divided into bounding boxes; S2, establish a three-branch baseline segmentation network: use the U-net as the baseline network for segmentation, establish several student models and teacher models, and combine the student model and the teacher model into a student-student model and a teacher-student model. All student models and teacher models are 3D U-Net networks with the same architecture but different initialization perturbations. S3, decoupling intra-class regions: the enhanced data is divided into labeled data and unlabeled data, and the blood vessel category is decoupled into difficult sub-regions and easy sub-regions, and different constraints are imposed on the difficult sub-regions and easy sub-regions; S4, constructing a dynamic threshold module: constructing a dynamic threshold module, and accurately decoupling different difficult sub-regions and easy sub-regions through the dynamic threshold module; S5, establish a constrained optimization network: all student models learn to represent different difficult sub-regions and easy sub-regions, and the teacher model constrains all student models to be consistent; S6, training a semi-supervised blood vessel segmentation network: using the loss function to utilize labeled data and unlabeled data to train a semi-supervised blood vessel segmentation network; S7, segmentation target: input the image to be tested, and the semi-supervised blood vessel segmentation network tests the image to be tested to complete the segmentation of the blood vessel area.

2. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 1, characterized in that: In S1, the minimum bounding box containing the blood vessel to be segmented is used as the range of the region of interest of the student model.

3. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 1, characterized in that: In S1, the data augmentation method uses one or more of random horizontal flipping, vertical flipping, and random -90 to 90 degree rotation.

4. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 1, characterized in that: In S2, the teacher model is the self-integration of the average of the student models, obtained by exponential moving average.

5. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 1, characterized in that: In S2, the expression of the student model is as follows: （1） （2） The expression of the teacher model is as follows: （3） in, are model parameters, and All are student models. Model for teachers.

6. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 1, characterized in that: In S3, the unlabeled data includes the probability map generated by the student model and the noisy pseudo labels; When performing the decoupling operation, the predicted probability map of one student model and the noisy pseudo-label of the other student model are sent to the dynamic threshold module to obtain the first dynamic threshold; when the consistency of the voxels of the probability map generated by the other student model is greater than the first dynamic threshold, it is considered that the student model has reached consistency in this area, and the area where the student model is located is an easy sub-area; The prediction consistency of the student-student model is used to obtain the easy sub-region of the vascular image, and the information of the easy sub-region is stored in the confidence map. The specific formula of the process is as follows: （4） （5） in, is the indicator function, is the dynamic threshold module, and is a probability map and noisy pseudo-label generated by one of the student models, and is a probability map and noisy pseudo-label generated by another student model, is the first dynamic threshold for each category, is the regional confidence map, is the degree of consistency between student models.

7. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 6, characterized in that: In S3, when performing the decoupling operation, the probability map of the teacher model and the average value of the noisy pseudo labels of all student models are sent to the dynamic threshold module to obtain the second dynamic threshold. If the consistency of the voxels of the teacher model is less than the second dynamic threshold, it is considered that there is ambiguity here, which is a difficult sub-region; The prediction uncertainty of the teacher-student model is used to obtain the difficult sub-regions of the vascular image, and the information of the difficult sub-regions is stored in the uncertainty map. The specific formula of the process is as follows: （6） （7） （8） in, is the probability graph of the teacher model, for and The average value of is the second dynamic threshold for each category, is the uncertainty diagram, is the consistency of the teacher model.

8. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 7, characterized in that: In S4, when performing precise decoupling operations, firstly, the pseudo labels are overlaid on the probability map and the voxels of the corresponding categories in the pseudo labels of each category in the probability map are found. Finally, the threshold of each category is calculated separately. The specific formula is as follows: （9） （10） in, is the indicator function, M represents mask, represents a noisy pseudo label, represents a probability graph, Indicates the threshold value.

9. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 8, characterized in that: In S5, for the easy sub-region, pseudo-label cross-supervision is performed between different student models, and the pseudo-label cross-supervision loss is defined as follows: （11） （12） （13） in, is the dice loss, and They are and The pseudo labels generated are and They are and The generated probability map, is the total number of data in the dataset, is the number of labeled data, Model for students Easy region confidence map with noisy pseudo labels; For difficult sub-regions, a teacher model is introduced to calculate the consistency loss. The parameters of the teacher model are updated by exponential moving average. The specific formula is as follows: （14） in, is the number of iterations of the current training, and Hyperparameters for controlling exponential moving average updates; The formula for consistency loss is as follows: （15） in, is the squared loss function, For unlabeled data, and are the model parameters of the student model, are the model parameters of the teacher model.

10. The medical image segmentation method with dynamic decoupling of intra-class regions according to claim 9, characterized in that: In S6, the loss function includes supervised loss and unsupervised loss. The specific formula of the loss function is as follows: （16） in, represents the supervision loss, represents the unsupervised loss, is a balance parameter used to balance supervised loss and unsupervised loss; The specific formula of supervision loss is as follows: （17） in, is a linear combination of cross entropy loss and dice loss, For labeled data, is the label of the labeled data; The unsupervised loss includes the pseudo-label cross-supervision loss between student models and the consistency loss between teacher and student models. The specific formula is as follows: （18） in, represents the pseudo-label cross-supervision loss, Indicates consistency loss.

Images