Semi-supervised medical image segmentation data enhancement method and system based on cutting and splicing

Abstract

The invention discloses a semi-supervised medical image segmentation data enhancement method and system based on cutting and splicing, and relates to the technical field of medical image processing. Comprising the steps of obtaining medical image data, and performing horizontal cutting and vertical cutting on the medical image data to obtain a mixed picture data set; inputting the pre-training set into a Mean Teamer model, carrying out pre-training to obtain a pre-trained Mean Teamer model, inputting the training set into the pre-trained Mean Teamer model, and carrying out iteration for a plurality of times to obtain a trained Mean Teamer model; and inputting the test set into the trained Mean Teacher model, and evaluating a medical image segmentation result. According to the method, through a consistent learning strategy, complementarity of marked and unmarked data is fully utilized, and the segmentation effect is further improved, so that the model can show higher precision and reliability when facing a complex medical image.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a semi-supervised medical image segmentation data augmentation method and system based on cropping and stitching. Background Art

[0002] In the field of medical image segmentation, semi-supervised learning has been widely used to solve the problem of insufficient labeled data. Although this method can effectively combine labeled and unlabeled data, existing semi-supervised medical image segmentation methods still have some deficiencies, especially in how to effectively balance the use of these two types of data. Many existing methods usually use labeled data for supervised learning, while unlabeled data is processed through pseudo-labels or self-supervised learning. Although these methods have improved the performance of the model to a certain extent, there are still some key problems. First, such methods often process labeled and unlabeled data separately and fail to fully combine their potential synergistic effects. Due to the failure to effectively fuse these two types of data, the utilization efficiency of labeled data is low, resulting in limited generalization ability of the model. In practical applications, the model may not be able to extract sufficient semantic information from limited labeled data, and unlabeled data also fails to fully play its role during training, ultimately leading to unsatisfactory segmentation results.

[0003] Although there are some strategies for fusing labeled and unlabeled data in existing methods, such as the bidirectional copy-paste strategy based on CutMix, which alleviates the problem of mismatched distribution of labeled and unlabeled data to a certain extent, its efficiency is not high and there are still some significant defects. Specifically, the bidirectional copy-paste strategy is random when cutting and splicing, which may cause the key parts of the image to be accidentally lost, which in turn has a negative impact on the accuracy of the segmentation results. At the same time, the number of labeled and unlabeled pixels in the fused mixed image is often not equal, which makes it impossible for the model to effectively learn equivalent labeled and unlabeled knowledge. In this case, the model's ability to learn for unlabeled data is still limited and fails to fully improve the overall performance. In addition, the lack of knowledge transfer is particularly prominent when labeled data is scarce, which makes it difficult for the model to effectively balance the advantages of labeled and unlabeled data, ultimately reducing the segmentation accuracy and consistency of the model. In existing semi-supervised methods, the quality of pseudo-labels cannot be ignored. In the training process of unlabeled data, the model has a strong dependence on pseudo-labels, but the quality of pseudo-labels is often not guaranteed. Especially when labeled data is scarce, erroneous pseudo-labels may be amplified by the model during training, further increasing the noise, thus affecting the convergence and segmentation performance of the model. As training progresses, the accumulation of errors will further weaken the stability and accuracy of the model. In general, although the existing semi-supervised medical image segmentation methods have improved the utilization efficiency of labeled and unlabeled data to a certain extent, the existing fusion strategies have not yet achieved the full transfer of knowledge from labeled data to unlabeled data, and still face problems such as insufficient generalization ability, poor pseudo-label quality, and low data utilization efficiency. Therefore, how to design a more effective fusion strategy to maximize the use of labeled and unlabeled data and alleviate the distribution differences between the two is still a key problem that needs to be solved in the current field of semi-supervised medical image segmentation.

[0004] Therefore, it is an urgent problem for those skilled in the art to propose a semi-supervised medical image segmentation data enhancement method and system based on cropping and splicing to solve the difficulties existing in the prior art. Summary of the invention

[0005] In view of this, the present invention provides a semi-supervised medical image segmentation data enhancement method and system based on cropping and splicing. Through a consistent learning strategy, it fully utilizes the complementarity of labeled and unlabeled data, further improves the segmentation effect, and enables the model to show higher accuracy and reliability when facing complex medical images.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A semi-supervised medical image segmentation data enhancement method based on cropping and splicing, comprising the following steps: S1. Obtain data: Obtain medical image data, and perform horizontal and vertical cropping on the medical image data to obtain a mixed picture dataset; S2. Data preprocessing: Preprocess the obtained mixed picture dataset; S3. Data partitioning: Partition the preprocessed mixed picture dataset into a pre-training set, a training set, and a test set; S4. Model training: Input the pre-training set into the Mean Teacher model for pre-training to obtain the pre-trained Mean Teacher model. Input the training set into the pre-trained Mean Teacher model, and after several iterations, obtain the trained Mean Teacher model; S5. Segmentation result: Input the test set into the trained Mean Teacher model to evaluate the medical image segmentation result.

[0007] Optionally, in S1, 50% of the obtained medical image data is labeled, and the medical image data is horizontally and vertically cropped, and a mixed picture dataset is generated through component splicing. The mixed picture dataset includes mixed labeled picture data, mixed unlabeled picture data, mixed picture data, and mixed label picture data.

[0008] Optionally, the pre-training set in S3 is composed of mixed labeled picture data, and the training set and the test set are composed of mixed picture data.

[0009] Optionally, the Mean Teacher model in S4 consists of a teacher network and a student network where are the parameters of the teacher network,

[0010] Optionally, in the pre-training stage of S4, the data in the pre-training set is used, and the corresponding real mixed labeled picture data is used to generate mixed label picture data for supervision; in the training stage, the mixed picture data in the training set is used, and the mixed label picture data is used for supervision; the pre-training stage and the training stage are calculated through the following methods: where is the training set data, is the horizontally cropped mixed image sent to the teacher network for training, tb is top and bottom, t is the teacher network teacher, and The first and second mixed images sent for student network training The mixed image after vertical cropping sent for teacher network training lr For left and right The labeled image The unlabeled image For element-wise multiplication And Are four zero-tensor masks, which are square, and the values of the four half-regions of up, down, left, and right are set to 1 respectively; Use the following formula to calculate the generation of mixed label image data for supervision in the pre-training stage corresponding to the real mixed labeled picture data: Use the following formula to calculate the supervision using mixed label image data in the training stage: Among them, Is the pseudo-label generated by the pre-trained teacher network For And The corresponding real label And The mixed label after horizontal cropping and mixing For And The corresponding real label And The mixed label after vertical cropping and mixing For And The corresponding real label And the pseudo-label The mixed label after horizontal cropping and mixing For And The corresponding pseudo-label And the real label The mixed label after horizontal cropping and mixing For The corresponding real label For The corresponding real label For The corresponding pseudo-label For The corresponding real label For The corresponding pseudo-label For The corresponding real label.

[0011] Optionally, in each iteration of S4, first use the stochastic gradient descent method to optimize the student network parameters, and use the exponential moving average model of the student network parameters to update the teacher network parameters. In the pre-training stage of the teacher network, the loss function is defined as follows: The prediction result of the teacher network is calculated by the following formula: The total loss function in the pre-training stage is defined as: The total loss function in the training stage is defined as: Among them, is a linear combination of the loss and the cross-entropy loss, and is the prediction result of the teacher network, is the horizontally cropped mixed image fed into the teacher network for training is the prediction result obtained through model prediction and the mixed label is the value of the loss function calculated is the vertically cropped mixed image fed into the teacher network for training is the prediction result obtained through model prediction and the mixed label is the value of the loss function calculated and is and corresponding true labels and the mixed label after horizontal cropping and mixing, and and corresponding true labels and the mixed label after vertical cropping and mixing, is the value of the loss function calculated from the prediction result of the mixed image and the corresponding mixed label, is the value of the loss function of the mixed image is the value of the loss function of the mixed image is the value of the loss function of the mixed image

[0012] ​​​A semi-supervised medical image segmentation data augmentation system based on cropping and stitching, applying a semi-supervised medical image segmentation data augmentation method according to any one of the above, including: a data acquisition module, a data preprocessing module, a data partitioning module, a model training module, and a segmentation result module; The data acquisition module, connected to the input end of the data preprocessing module, is used to acquire medical image data, and perform horizontal cropping and vertical cropping on the medical image data to obtain a mixed picture data set; The data preprocessing module, connected to the input end of the data partitioning module, is used to preprocess the obtained mixed picture data set; The data partitioning module, connected to the input end of the model training module, is used to partition the preprocessed mixed picture data set into a pre-training set, a training set, and a test set; The model training module, connected to the input end of the segmentation result module, is used to input the pre-training set into the MeanTeacher model for pre-training to obtain a pre-trained Mean Teacher model, input the training set into the pre-trained Mean Teacher model, and obtain a trained Mean Teacher model after several iterations; The segmentation result module, connected to the output end of the model training module, is used to input the test set into the trained MeanTeacher model to evaluate the medical image segmentation result.

[0013] It can be seen from the above technical solutions that compared with the prior art, the present invention provides a semi-supervised medical image segmentation data augmentation method and system based on cropping and stitching, having the following beneficial effects: (1) In the present invention, the labeled image and the unlabeled image are cropped into upper and lower and left and right components of the same size along the horizontal and vertical directions, and then pixel-based bidirectional hybrid stitching is performed, effectively integrating the labeled and unlabeled data, significantly increasing the diversity of the input data, solving the problem of empirical mismatch between the labeled data and the unlabeled data, and thus improving the overall performance of the model; (2) The present invention deploys the data augmentation method based on cropping and stitching on the Mean Teacher architecture with UNet or VNet as the backbone network. Through the consistency learning strategy, the complementarity of the labeled and unlabeled data is fully utilized, further improving the segmentation effect, so that the model can show higher accuracy and reliability when facing complex medical images; (3)The proposed cropping and stitching method ensures that the amount of labeled and unlabeled pixels in the mixed image is equal, enabling the model to simultaneously learn the balanced semantic knowledge of labeled and unlabeled data. In addition, by converting each image into such a mixed image with equal pixel amounts, it plays an important role in constructing a more unified dataset, significantly alleviating the distribution difference between labeled and unlabeled data; (4)By effectively integrating labeled and unlabeled data, the present invention significantly enhances the adaptability of the model to medical images, successfully alleviates the distribution mismatch problem between labeled and unlabeled data, improves the robustness of the model in dealing with diverse and complex scenarios, enabling it to more effectively cope with the diversity and challenges in medical image segmentation. This innovation provides crucial support for achieving high-performance optimization in the case of scarce labeled data and lays a more reliable technical foundation for clinical applications; (5)The present invention is widely applied to the image segmentation of important organs such as the prostate and heart and the diagnosis of related diseases. Specifically for prostate-related diseases (such as benign prostatic hyperplasia, prostatitis, and prostate cancer) and heart-related diseases, etc., through data augmentation techniques such as cropping and stitching, the problem of unbalanced data distribution is effectively alleviated, thus significantly improving the segmentation performance, which makes the identification of lesions more accurate; this accurate lesion segmentation provides doctors with clearer and more accurate imaging analysis support, thereby enhancing the ability of early diagnosis and disease screening and improving the accuracy of risk assessment for prostate diseases and heart diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or 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 the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the provided drawings.

[0015] Figure 1 It is a flowchart of a semi-supervised medical image segmentation data augmentation method based on cropping and stitching provided by the present invention; Figure 2 It is a schematic diagram of medical image cropping and stitching provided by the present invention; Figure 3 It is a structure diagram of a pre-trained network provided by the present invention; Figure 4 It is a structure diagram of a training network provided by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Referring to Figure 1 as shown, the present invention discloses a semi-supervised medical image segmentation data augmentation method based on cropping and splicing, including the following steps: S1. Obtain data: Obtain medical image data, and perform horizontal cropping and vertical cropping on the medical image data to obtain a mixed picture dataset; S2. Data preprocessing: Perform preprocessing on the obtained mixed picture dataset; S3. Data division: Divide the preprocessed mixed picture dataset into a pre-training set, a training set, and a test set; S4. Model training: Input the pre-training set into the Mean Teacher model for pre-training to obtain the pre-trained Mean Teacher model. Input the training set into the pre-trained Mean Teacher model, and after several iterations, obtain the trained Mean Teacher model; S5. Segmentation result: Input the test set into the trained Mean Teacher model to evaluate the medical image segmentation result.

[0018] Further, in S1, 50% of the obtained medical image data is marked, and the medical image data is horizontally and vertically cropped, and a mixed picture dataset is generated through component splicing. The mixed picture dataset includes mixed marked picture data, mixed unmarked picture data, mixed picture data, and mixed label picture data.

[0019] Further, the pre-training set in S3 is composed of mixed marked picture data, and the training set and the test set are composed of mixed picture data.

[0020] Further, the Mean Teacher model in S4 consists of a teacher network and a student network wherein, are the parameters of the teacher network, are the parameters of the student network.

[0021] Further, in the pre-training stage of S4, the data in the pre-training set is used, and the mixed label image data is generated using the corresponding real mixed labeled image data for supervision; in the training stage, the mixed image data in the training set is used, and the mixed label image data is used for supervision; the pre-training stage and the training stage are calculated in the following way: Among them, is the training set data, is the mixed image after horizontal cropping sent to the teacher network for training, tb is top and bottom, t is the teacher network teacher, and are the first and second mixed images sent to the student network for training, is the mixed image after vertical cropping sent to the teacher network for training, lr is left and right, is the labeled image, is the unlabeled image, is element-wise multiplication, and are four zero-tensor masks, which are square, and the values of the four half-regions of up, down, left, and right are set to 1 respectively; The mixed label image data is generated using the corresponding real mixed labeled image data for supervision in the pre-training stage by calculating with the following formula: The mixed label image data is used for supervision in the training stage by calculating with the following formula: Among them, is the pseudo-label generated by the pre-trained teacher network, is and the corresponding real label and the mixed label after horizontal cropping and mixing, is and the corresponding real label and the mixed label after vertical cropping and mixing, is and the corresponding real label and the pseudo-label the mixed label after horizontal cropping and mixing, is and The corresponding pseudo-label and the true label The mixed label after horizontal cropping and mixing, is the corresponding true label, is the corresponding true label, is the corresponding pseudo-label, is the corresponding true label, is the corresponding pseudo-label, is the corresponding true label.

[0022] Furthermore, in each iteration of S4, first, the student network parameters are optimized using the stochastic gradient descent method, and the exponential moving average model of the student network parameters is used to update the teacher network parameters. In the pre-training stage of the teacher network, the loss function is defined as follows: The prediction result of the teacher network is calculated by the following formula: The total loss function in the pre-training stage is defined as: The total loss function in the training stage is defined as: Where, is a linear combination of the loss and the cross-entropy loss, and is the prediction result of the teacher network, is the mixed image after horizontal cropping fed into the teacher network for training The prediction result obtained through model prediction and the mixed label The loss function value calculated through calculation, is the mixed image after vertical cropping fed into the teacher network for training The prediction result obtained through model prediction and the mixed label The loss function value calculated through calculation, and is and the corresponding true label and the mixed label after horizontal cropping and mixing and and The corresponding true label and the mixed label after vertical cropping and mixing, is the mixed image The loss function value obtained by calculating the prediction result obtained by model prediction and the corresponding mixed label, is the mixed image 's loss function value, is the mixed image 's loss function value, is the mixed image 's loss function value.

[0023] Specifically, by multiplying the loss value of the pseudo - supervision part by a factor to adjust the loss value of the pseudo - supervision part, and the value is set to 0.5 by default. The loss function of each mixed image is defined as follows: and are the prediction results of the student network respectively, and can be calculated by the following formula: where, is the student network, are the parameters of the student network.

[0024] In a specific embodiment, it includes the following content: Select four medical images, two labeled images and two unlabeled images , pair the labeled and unlabeled images to form combinations and , through Figure 2 the horizontal cropping and vertical cropping methods in , mix and splice to form four mixed pictures, and obtain . Use the four mixed images to train the network. The entire model framework consists of a teacher network and a student network are the parameters of the teacher network, are the parameters of the student network. Figure 3 and Figure 4 correspond to the pre - training and training stages of the entire model respectively.

[0025] The present invention was compared with various state-of-the-art semi-supervised segmentation methods on the PROMISE 12 and LA datasets, and semi-supervised experiments were conducted for different labeling ratios. Four evaluation metrics were selected, including region-sensitive metrics: Dice coefficient (%) and Jaccard similarity coefficient (%), and edge-sensitive metrics: 95% Hausdorff distance (95HD) and average surface distance (ASD). The present invention achieved the best performance on all four evaluation metrics, significantly outperforming other competitors. The Dice scores of the present invention on the PROMISE 12 dataset for two labeling ratios (20% and 40%) were increased by 3.70% and 4.08% respectively. The Dice scores on the LA dataset for two labeling ratios (10% and 20%) were increased by 1.04% and 0.57% respectively.

[0026] From a non-image-level perspective, the present invention increases the data diversity by merging labeled and unlabeled data in the form of patches, and encourages the model to better learn the shared semantics of the two types of data. A large number of experiments on the PROMISE 12 and LA datasets have demonstrated the effectiveness of this method. The excellent segmentation performance indicates more effective utilization of labeled data.

[0027] Corresponding to Figure 1 the method described above, an embodiment of the present invention also provides a semi-supervised medical image segmentation data augmentation system based on cropping and stitching, which specifically includes: a data acquisition module, a data preprocessing module, a data partitioning module, a model training module, and a segmentation result module; The data acquisition module is connected to the input end of the data preprocessing module, and is used to acquire medical image data, and perform horizontal cropping and vertical cropping on the medical image data to obtain a mixed picture dataset; The data preprocessing module is connected to the input end of the data partitioning module, and is used to preprocess the obtained mixed picture dataset; The data partitioning module is connected to the input end of the model training module, and is used to partition the preprocessed mixed picture dataset into a pre-training set, a training set, and a test set; The model training module is connected to the input end of the segmentation result module, and is used to input the pre-training set into the MeanTeacher model for pre-training to obtain a pre-trained Mean Teacher model, input the training set into the pre-trained Mean Teacher model, and obtain a trained Mean Teacher model after several iterations; The segmentation result module is connected to the output end of the model training module, and is used to input the test set into the trained MeanTeacher model to evaluate the medical image segmentation result.

[0028] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For related parts, reference can be made to the description in the method section.

[0029] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A semi-supervised medical image segmentation data augmentation method based on cropping and stitching, characterized in that, Including the following steps: S1. Obtain data: Obtain medical image data, and perform horizontal cropping and vertical cropping on the medical image data to obtain a mixed picture dataset; S2. Data preprocessing: Perform preprocessing on the obtained mixed picture dataset; S3. Data division: Divide the preprocessed mixed picture dataset into a pre-training set, a training set, and a test set; S4. Model training: Input the pre-training set into the Mean Teacher model for pre-training to obtain the pre-trained Mean Teacher model. Input the training set into the pre-trained Mean Teacher model, and after several iterations, obtain the trained Mean Teacher model; S5. Segmentation result: Input the test set into the trained Mean Teacher model to evaluate the medical image segmentation result.

2. A semi-supervised medical image segmentation data augmentation method based on cropping and stitching according to claim 1, characterized in that In S1, 50% of the obtained medical image data is labeled, and the medical image data is horizontally cropped and vertically cropped, and a mixed picture dataset is generated through component stitching. The mixed picture dataset includes mixed labeled picture data, mixed unlabeled picture data, mixed picture data, and mixed label picture data.

3. A semi-supervised medical image segmentation data augmentation method based on cropping and stitching according to claim 1 or 2, characterized in that The pre-training set in S3 is composed of mixed labeled picture data, and the training set and the test set are composed of mixed picture data.

4. A semi-supervised medical image segmentation data augmentation method based on cropping and stitching according to claim 1, characterized in that The Mean Teacher model in S4 consists of a teacher network and a student network . Among them, are the parameters of the teacher network, are the parameters of the student network.

5. A semi-supervised medical image segmentation data augmentation method based on cropping and stitching according to claim 1, characterized in that In S4, in the pre-training stage, the data in the pre-training set is used, and the mixed label picture data is generated using the corresponding real mixed labeled picture data for supervision; in the training stage, the mixed picture data in the training set is used, and the mixed label picture data is used for supervision; the pre-training stage and the training stage are calculated in the following manner: Among them, is the training set data, is the mixed image after horizontal cropping sent to the teacher network for training, tb are top and bottom, t is the teacher network teacher, and are the first and second mixed images sent to the student network for training, is the mixed image after vertical cropping sent to the teacher network for training, lr are left and right, is the labeled image, is the unlabeled image, is element-wise multiplication, and are four zero-tensor masks, which are square, and the values of the four half-regions of the top, bottom, left, and right are set to 1 respectively; The following formula is used to calculate the generation of mixed label picture data using the corresponding real mixed labeled picture data for supervision in the pre-training stage: The following formula is used to calculate the use of mixed label picture data for supervision in the training stage: Among them, is the pseudo-label generated by the pre-trained teacher network, is and the corresponding true label and the mixed label after horizontal cropping and mixing, is and the corresponding true label and the mixed label after vertical cropping and mixing, is and the corresponding true label and the pseudo-label the mixed label after horizontal cropping and mixing, is and the corresponding pseudo-label and the true label the mixed label after horizontal cropping and mixing, is the corresponding true label, is the corresponding true label, is the corresponding pseudo-label, is the corresponding true label, is the corresponding pseudo-label, is the corresponding true label.

6. A semi-supervised medical image segmentation data augmentation method based on cropping and stitching according to claim 1, characterized in that In S4, in each iteration, first use the stochastic gradient descent method to optimize the student network parameters, and use the exponential moving average model of the student network parameters to update the teacher network parameters. In the pre-training stage of the teacher network, the loss function is defined as follows: The prediction result of the teacher network is calculated by the following formula: The total loss function in the pre-training stage is defined as: The total loss function in the training stage is defined as: Among them, is a linear combination of the loss and the cross-entropy loss, and is the prediction result of the teacher network, is the mixed image after horizontal cropping fed into the teacher network for training is the prediction result obtained through model prediction and the mixed label is the value of the loss function obtained through calculation, is the mixed image after vertical cropping fed into the teacher network for training is the prediction result obtained through model prediction and the mixed label is the value of the loss function obtained through calculation, and is and corresponding true labels and the mixed label after horizontal cropping and mixing, as well as and corresponding true labels and the mixed label after vertical cropping and mixing, is the value of the loss function obtained by calculating the prediction result of the mixed image and the corresponding mixed label, is the value of the loss function of the mixed image , is the value of the loss function of the mixed image , is the value of the loss function of the mixed image .

7. A semi-supervised medical image segmentation data augmentation system based on cropping and stitching, characterized in that, Applying a semi-supervised medical image segmentation data augmentation method based on cropping and splicing according to any one of claims 1-6, comprising: an acquisition data module, a data preprocessing module, a data partitioning module, a model training module, and a segmentation result module; The acquisition data module, connected to the input end of the data preprocessing module, is used to acquire medical image data, and perform horizontal cropping and vertical cropping on the medical image data to obtain a mixed picture data set; The data preprocessing module, connected to the input end of the data partitioning module, is used to preprocess the obtained mixed picture data set; The data partitioning module, connected to the input end of the model training module, is used to partition the preprocessed mixed picture data set into a pre-training set, a training set, and a test set; The model training module, connected to the input end of the segmentation result module, is used to input the pre-training set into the Mean Teacher model for pre-training to obtain a pre-trained Mean Teacher model, input the training set into the pre-trained Mean Teacher model, and obtain a trained Mean Teacher model after several iterations; The segmentation result module, connected to the output end of the model training module, is used to input the test set into the trained Mean Teacher model to evaluate the medical image segmentation result.

Images