Semi-supervised medical image segmentation method based on contrast manifold regularization and related device

Abstract

The invention discloses a semi-supervised medical image segmentation method based on contrast manifold regularization and a related device. The method comprises the following steps: importing and preprocessing a medical image data set; initializing a teacher model and a student model, training the teacher model by using the annotation data, and generating a pseudo tag; calculating the similarity between the annotation data and the pseudo tag to obtain a manifold regularization item; constructing positive and negative sample pairs and calculating and comparing loss items; performing weighted summation to obtain a contrast manifold regularization item, and combining the contrast manifold regularization item with supervision loss to serve as a loss function of the student model; and iteratively training the student model to obtain a segmentation result. Compared with the prior art, by combining comparative learning and manifold regularization, the problem of data dependence in semi-supervised medical image segmentation is effectively relieved, the model generalization ability and segmentation precision are improved, and the method is particularly excellent in performance in a small target focus segmentation scene.

Description

Technical Field

[0001] The present disclosure relates to the technical field of medical image processing, and in particular, to a semi-supervised medical image segmentation method and related device based on contrast manifold regularization. Background Art

[0002] In the field of medical image analysis, deep learning technology has been widely applied to image segmentation tasks, providing important support for clinical diagnosis and treatment planning. However, existing technical solutions still face significant challenges in semi-supervised medical image segmentation scenarios.

[0003] Currently, medical image segmentation methods based on deep learning highly rely on large-scale labeled data. In actual clinical scenarios, accurate annotation of medical images requires radiologists to spend a large amount of time making layer-by-layer delineations, resulting in extremely high costs for obtaining labeled data. This data scarcity directly limits the performance of supervised learning frameworks, especially in segmentation tasks for rare cases and special diseases.

[0004] To alleviate the data dependence problem, semi-supervised learning methods have gradually become a research hotspot. Existing technologies usually adopt a technical route combining self-supervised learning and contrastive learning to improve the generalization ability of the model through feature learning of unlabeled data. However, such methods have inherent defects at the feature representation level: when the proportion of labeled data is lower than a specific threshold, the model performance will significantly decline, making it difficult to meet the requirements of clinical applications.

[0005] In terms of the feature constraint mechanism, traditional manifold regularization methods improve the model generalization by maintaining the intrinsic geometric structure of data, but their effects highly depend on the accuracy of the manifold assumption. Most existing technical solutions introduce manifold constraints as an independent module into the training process, failing to form effective cooperation with contrastive learning, resulting in inconsistent distributions in the feature space, specifically manifested as a relatively high dispersion of features of the same class samples and an overlapping phenomenon of features of different class samples.

[0006] For the small target lesion segmentation scenario, existing semi-supervised methods face dual challenges: on the one hand, limited labeled data is difficult to support the effective learning of small target features; on the other hand, traditional data augmentation strategies have limitations in applicability in the three-dimensional medical image scenario, resulting in generally low sensitivity for small target detection.

[0007] At the clinical deployment level, the real-time performance of existing semi-supervised segmentation frameworks is difficult to meet the rigid requirements such as surgical navigation. Most solutions take too long to run and are not specifically optimized for intraoperative scenarios, restricting the practical application value of the technology in clinical practice. Summary of the Invention

[0008] To solve the above technical problems, the present invention provides a semi-supervised medical image segmentation method and related device based on contrast manifold regularization.

[0009] A first aspect provides a semi-supervised medical image segmentation method based on contrast manifold regularization. The method includes: 1. Import and preprocess a medical image dataset, where the medical image dataset includes labeled data and unlabeled data; 2. Initialize a teacher model and a student model, train the teacher model using the labeled data to obtain teacher weights; screen the unlabeled data to obtain screened unlabeled data; based on the teacher weights, use the teacher model to generate pseudo-labels for the screened unlabeled data to obtain pseudo-labeled data; 3. Calculate the similarity between the labeled data and the pseudo-labeled data to obtain a reproducing kernel Hilbert space manifold regularization term; 4. Based on the labeled data and the pseudo-labeled data, perform random cropping to construct positive and negative sample pairs; where the positive sample pair is two different randomly cropped images of the same medical image; the negative sample pair is different medical images; use the positive sample pair and the negative sample pair to calculate a contrast loss function to obtain a contrast loss term; 5. Perform weighted summation of the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain a contrast manifold regularization term; 6. Obtain a supervised loss based on the labeled data, and then perform weighted summation of the contrast manifold regularization term and the supervised loss as the loss function of the student model; 7. Based on the labeled data and the pseudo-labeled data, use the loss function to iteratively train the student model to obtain student weights; based on the student weights, use the student model to perform an image segmentation task to obtain a segmentation result.

[0010] Combined with any embodiment of the present application, the screening of the unlabeled data to obtain screened unlabeled data includes: Screening based on data quality, where the data quality includes image clarity and contrast; screening based on the similarity between the unlabeled data and the labeled data, and the similarity is measured by calculating the feature distance between the unlabeled data and the labeled data.

[0011] Combined with any embodiment of the present application, the calculation of the similarity between the labeled data and the pseudo-labeled data to obtain a reproducing kernel Hilbert space manifold regularization term includes: Select a kernel function; calculate a similarity matrix, where the similarity matrix includes similarity values between the labeled data and the pseudo-labeled data; construct the reproducing kernel Hilbert space manifold regularization term based on the similarity matrix.

[0012] Combined with any embodiment of the present application, performing random cropping based on the labeled data and the pseudo-labeled data to construct positive and negative sample pairs includes: Set the cropping size, which is determined according to the resolution of the medical image and the size of the lesion; Randomly select the cropping positions, which are independently selected on the labeled data and the pseudo-labeled data; Construct positive sample pairs, where the positive sample pairs are two different randomly cropped images of the same medical image; construct negative sample pairs, where the negative sample pairs are different medical images.

[0013] Combined with any embodiment of the present application, weighted summation of the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain the contrast manifold regularization term includes: The contrast loss term is used to measure the distance difference between the positive sample pairs and the negative sample pairs in the feature space, and the formula is: , where, and are the feature representations of the positive sample pairs, is the feature representation of the negative sample pairs, is the similarity function, is the temperature parameter; Perform weighted summation of the manifold regularization term and the contrast loss term to obtain the contrast manifold regularization term, and the formula is: , where, is the reproducing kernel Hilbert space manifold regularization term, is the contrast loss term, is the weight of the contrast loss term;

[0014] Combined with any embodiment of the present application, obtaining the supervised loss based on the labeled data, and then weighted summing the contrast manifold regularization term and the supervised loss as the loss function of the student model includes: The supervised loss function is obtained by the student model predicting the labeled data, and is used to measure the difference between the prediction result of the student model on the labeled data and the true label; Perform weighted summation of the supervised loss term and the contrast manifold regularization term to obtain the loss function of the student model, and the formula is: , where, is the supervised loss function, is the weight of the contrast manifold regularization term; The loss function of the student model is used to guide the training process of the student model.

[0015] Combined with any implementation manner of the present application, based on the labeled data and the pseudo-label data, using the loss function to iteratively train the student model to obtain student weights, including: Set the iteration end condition and initialize the training parameters; Use the loss function to calculate the loss values of the student model on the labeled data and the pseudo-label data; Update the parameters of the student model according to the loss values; repeat the iteration process until the iteration end condition is satisfied.

[0016] In a second aspect, a semi-supervised medical image segmentation device based on contrast manifold regularization is provided, including: An input unit for importing and preprocessing a medical image data set, where the medical image data set includes labeled data and unlabeled data; A contrast learning unit, based on the labeled data and the pseudo-label data, performs random cropping to construct positive and negative sample pairs; wherein, the positive sample pair is two different randomly cropped images of the same medical image; the negative sample pair is different medical images; use the positive sample pair and the negative sample pair to calculate a contrast loss function to obtain a contrast loss term; A manifold regularization unit for calculating the similarity between the labeled data and the pseudo-label data to obtain a reproducing kernel Hilbert space manifold regularization term; A model training unit for initializing a teacher model and a student model, using the labeled data to train the teacher model to obtain teacher weights; screening the unlabeled data to obtain screened unlabeled data; based on the teacher weights, using the teacher model to generate pseudo-labels for the screened unlabeled data to obtain pseudo-label data; is also used to perform weighted summation of the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain a contrast manifold regularization term; obtain a supervised loss based on the labeled data, and then perform weighted summation of the contrast manifold regularization term and the supervised loss as the loss function of the student model; based on the labeled data and the pseudo-label data, use the loss function to iteratively train the student model to obtain student weights; An output unit for performing an image segmentation task using the student model based on the student weights to obtain a segmentation result.

[0017] In a third aspect, an electronic device is provided, comprising: a processor, a communication module, a sensor, a user interface, and a storage unit, wherein the storage unit is used to store computer program code, wherein the program code includes computer instructions. When the processor executes these instructions, the electronic device will execute the method described in the second aspect and any embodiment thereof.

[0018] In a fourth aspect, another electronic device is provided, comprising: a processor, a wireless communication module, a touch screen, a speaker, and a storage unit, wherein the storage unit is used to store computer program code, wherein the program code includes computer instructions. When the processor executes these instructions, the electronic device will execute the method described in the second aspect and any embodiment thereof.

[0019] In a fifth aspect, a computer-readable storage medium is provided, wherein a computer program is stored, wherein the program includes program instructions. When these instructions are executed by a processor, the processor will execute the method described in the second aspect and any embodiment thereof.

[0020] In a sixth aspect, a computer program product is provided, the computer program product comprising a computer program or instructions. When the computer program or instructions are run on a computer, the computer will execute according to the method described in the second aspect and any of its embodiments.

[0021] It should be understood that the above general description and the following detailed description are only used as examples and explanations and do not impose any limitations on the present application.

[0022] In this application, compared with the prior art, the present invention first obtains a small amount of labeled medical image data and a large amount of unlabeled data in a ratio of 1:N (N≥1); then constructs a semi-supervised learning framework combining contrastive learning and manifold regularization, defines positive and negative sample pairs through a contrastive learning module and trains the model using a contrastive loss function, and at the same time uses a manifold regularization module to impose regularization constraints on the prediction results in a reproducing kernel Hilbert space (RKHS); finally, the labeled and unlabeled data are input into the framework to complete the model training.

[0023] The present invention integrates contrastive learning to mine the intrinsic semantic features of data and manifold regularization to preserve the manifold structure information of the data, breaking through the limitations of traditional supervised learning that only relies on labeled data and the low accuracy of unsupervised learning, and solving the problems of insufficient accuracy and poor noise resistance of existing semi-supervised medical image segmentation methods in complex images. It significantly improves the segmentation accuracy and robustness, and is particularly suitable for medical image scenarios where labeled data are scarce. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] To more clearly illustrate the technical solutions in the embodiments of the present application or the background art, the following will describe the drawings required for use in the embodiments of the present application or the background art.

[0025] The accompanying drawings herein are incorporated into the specification and form a part of this specification. These drawings illustrate embodiments consistent with the present application and, together with the specification, are used to explain the technical solutions of the present application.

[0026] FIG. 1 is a schematic flow chart of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application; FIG. 2 is a schematic detailed flowchart of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application; FIG. 3 is a schematic structural diagram of a semi-supervised medical image segmentation device based on contrast manifold regularization provided by an embodiment of the present application; FIG. 4 is a schematic hardware architecture diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0027] To enable professionals in the technical field to more comprehensively understand the technical solutions of the present application, the following will explain the technical solutions of the present application in detail and clearly through the accompanying drawings. It should be particularly noted that the described embodiments are only partial examples of the present application and do not represent all of them. Based on these embodiments, those skilled in the art can directly derive all other possible implementation schemes without creative thinking, and these are also included within the protection scope of the present application.

[0028] In the specification, claims, and related drawings of the present application, the terms "first", "second", etc. are only used to distinguish different elements and do not imply any specific order. At the same time, the use of "including" and "having" and their variants means non-exclusive inclusion. This means that if a process, method, system, product, or device includes a series of steps or components, it indicates that the process, method, system, product, or device is not limited to the listed steps or components, and may also include other steps or components not listed, or other steps or units inherent to it.

[0029] As used herein, "embodiment" refers to any instance that combines specific features, structures, or characteristics, and these instances may belong to at least one embodiment of the present application. The "embodiments" mentioned in the text do not necessarily refer to the same specific case, nor do they indicate that they are mutually independent or exclusive alternative solutions. Those skilled in the art should understand that the embodiments described in the text can be used together with other embodiments. It should be clear that in the present application, "at least one" includes one or more instances, "a plurality" means two or more instances, and "at least two" means two or more instances.

[0030] It should be understood that the method embodiments of the present application can also be implemented by a processor executing computer program code. The embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application.

[0031] Please refer to Figure 1 , FIG. 1 is a schematic flowchart of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application. The specific steps are as follows:

[0032] Step 101: Data acquisition and preprocessing: Import a medical image dataset, which includes labeled data and unlabeled data, and perform preprocessing operations on this data so that the subsequent model can better perform learning and segmentation tasks.

[0033] In this embodiment, the modalities of the dataset include but are not limited to CT, enhanced CT, ultrasound, MRI, etc., but it is best to avoid inputting data of mixed modalities.

[0034] In this embodiment, the preprocessing operations include but are not limited to normalization, data augmentation, data cleaning, and data anonymization, etc.

[0035] In another possible implementation method, in some cases, data augmentation techniques can be used to amplify the labeled medical image data. For example, operations including but not limited to rotation, scaling, and cropping are performed on the labeled images.

[0036] Step 102: Model initialization and pseudo-label generation: Initialize the teacher model and the student model, train the teacher model using the labeled data to obtain the teacher weights. Then screen the unlabeled data to obtain the screened unlabeled data. Based on the teacher weights, use the teacher model to generate pseudo-labels for the screened unlabeled data, thereby obtaining pseudo-labeled data. For detailed content, please refer to Figure 2 , Figure 2 , which is a schematic diagram of the process details of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application.

[0037] In another possible implementation manner, the teacher model can be initialized with a pre-trained model to improve the initial performance and training efficiency of the model.

[0038] In yet another possible implementation manner, the process of screening the unlabeled data can be based on the characteristics or distribution characteristics of the data to ensure that the screened unlabeled data is representative.

[0039] Step 103: Obtaining the popularity regularization term: Calculate the similarity between the labeled data and the pseudo-labeled data, and obtain the reproducing kernel Hilbert space manifold regularization term through a specific calculation method. This regularization term helps the model learn the intrinsic structure and distribution characteristics of the data. For detailed content, please refer to Figure 2 。 Figure 2 This is a schematic diagram of the process details of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application.

[0040] In another possible implementation, different kernel functions can be used to calculate the reproducing kernel Hilbert space manifold regularization term to adapt to different types of data distributions.

[0041] In yet another possible implementation, the manifold regularization term can be optimized through an optimization algorithm to improve the performance and generalization ability of the model.

[0042] Step 104: Obtaining the contrast loss term: Based on the labeled data and the pseudo-labeled data, perform random cropping operations to construct positive and negative sample pairs. Among them, the positive sample pair is two different randomly cropped images of the same medical image, and the negative sample pair is different medical images. Use these positive and negative sample pairs to calculate the contrast loss function to obtain the contrast loss term, which can prompt the model to learn to distinguish the differences between different samples. For detailed content, please refer to Figure 2 , Figure 2 This is a schematic diagram of the process details of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application.

[0043] In another possible implementation, different cropping strategies can be adopted to construct positive and negative sample pairs to increase the diversity of the sample pairs.

[0044] In yet another possible implementation, the contrast loss function can adopt different loss function forms, including but not limited to InfoNCE loss or triplet loss, etc., to adapt to different task requirements.

[0045] Step 105: Constructing the loss function: Weightedly sum the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain the contrast manifold regularization term. Then obtain the supervised loss based on the labeled data, and weightedly sum the contrast manifold regularization term and the supervised loss. Finally, construct the loss function of the student model, which combines the advantages of supervised learning and contrast learning. For detailed content, please refer to Figure 2 , Figure 2 This is a schematic diagram of the process details of a semi-supervised medical image segmentation method based on contrast manifold regularization provided by an embodiment of the present application.

[0046] In another possible implementation, different weighting strategies can be adopted to balance the weights of the manifold regularization term and the contrast loss term, so as to optimize the performance of the model.

[0047] In yet another possible implementation, the construction of the loss function can incorporate other regularization terms or constraints to further improve the generalization ability and stability of the model.

[0048] Step 106: Model training and segmentation task execution: Based on the labeled data and pseudo-labeled data, use the constructed loss function above to iteratively train the student model to obtain the student weights. Finally, based on the student weights, use the student model to perform the segmentation task on the medical image to obtain the final segmentation result.

[0049] In yet another possible implementation, the execution of the segmentation task can incorporate post-processing operations, such as morphological operations or filtering, etc., to further optimize the quality and accuracy of the segmentation result.

[0050] In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the methods described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be elaborated here.

[0051] The above has elaborated in detail the methods of the embodiments of the present application. Below, the devices of the embodiments of the present application are provided.

[0052] Please refer to Figure 3 , FIG. 3 is a structural schematic diagram of a semi-supervised medical image segmentation device based on contrast manifold regularization provided by an embodiment of the present application. The segmentation device 1 includes: an input unit 11, a contrast learning unit 12, a manifold regularization unit 13, a model training unit 14, and an output unit 15. Specifically: The input unit 11 is used to import and preprocess the medical image dataset, and the medical image dataset includes labeled data and unlabeled data; The contrast learning unit 12 randomly crops based on the labeled data and the pseudo-labeled data to construct positive and negative sample pairs; wherein, the positive sample pair is two different randomly cropped images of the same medical image; the negative sample pair is different medical images; use the positive sample pair and the negative sample pair to calculate the contrast loss function to obtain the contrast loss term; The manifold regularization unit 13 is used to calculate the similarity between the labeled data and the pseudo-labeled data to obtain the reproducing kernel Hilbert space manifold regularization term; A model training unit 14, configured to initialize a teacher model and a student model, train the teacher model by using the labeled data to obtain teacher weights; screen the unlabeled data to obtain screened unlabeled data; generate pseudo-labels for the screened unlabeled data by using the teacher model based on the teacher weights to obtain pseudo-labeled data; is further configured to perform weighted summation on the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain a contrast manifold regularization term; obtain a supervised loss based on the labeled data, and then perform weighted summation on the contrast manifold regularization term and the supervised loss as a loss function of the student model; iteratively train the student model by using the loss function based on the labeled data and the pseudo-labeled data to obtain student weights; An output unit 15, configured to perform an image segmentation task by using the student model based on the student weights to obtain a segmentation result.

[0053] Please refer to Figure 4 , FIG. 4 shows a schematic hardware architecture diagram of an electronic device described in an embodiment of the present application. The electronic device 2 mainly includes a processor 21 and a memory 22. In addition, the device may further include an input device 23 and an output device 24. The processor 21, the memory 22, the input device 23, and the output device 24 are interconnected through connection components, and these connection components may be various interfaces, data lines, or communication buses, etc., and the embodiments of the present application do not make specific regulations thereon.

[0054] The processor 21 may be one or more graphics processing units (GPUs). If the processor 21 is a GPU, the GPU may be single-core or multi-core. As an option, the processor 21 may also be a processor group composed of multiple GPUs, and these processors are interconnected through one or more buses. In addition, the processor may also be other types of processors, and the embodiments of the present application do not make specific limitations thereon.

[0055] The memory 22 is designed to store instructions of a computer program and various program codes required to execute the solution of the present application. As an option, the memory may include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or a compact disc read-only memory (CD-ROM), and these memories are used to store relevant instructions and data.

[0056] The input device 23 is used to input data and / or signals, and the output device 24 is used to output data and / or signals. The input device 23 and the output device 24 may be independent devices or an integrated device.

[0057] It should be recognized that in the embodiments of the present application, the memory 22 can not only store relevant instructions but also store relevant data. The specific data content stored in the memory is not specifically defined in the embodiments of the present application.

[0058] It should be understood that FIG. 3 only shows a simplified design of an electronic device. In actual use, the electronic device may also include other necessary components, such as different numbers of input / output devices, processors, memories, etc. All electronic devices capable of implementing the embodiments of the present application fall within the protection scope of the present application.

[0059] Those skilled in the art should recognize that according to the components and algorithm steps described in the various examples in the embodiments disclosed herein, these functions can be implemented by electronic hardware or in a manner combining computer software and electronic hardware. Specifically, whether to execute these functions by hardware or software will be determined based on the specific application requirements and design limitations of the technical solution. Those skilled in the art can adopt different implementation methods according to the requirements of each specific application, but such implementation methods should not be regarded as exceeding the protection scope of the present application.

[0060] Professionals should understand that for the convenience of description and simplification of the description, the specific operation procedures of the above systems, devices, and components can refer to the corresponding steps in the foregoing method embodiments and will not be repeated here. At the same time, professionals should also understand that each embodiment in the present application has its own emphasis. For the convenience of description and simplification, the same or similar content may not be repeatedly described in different embodiments. Therefore, for parts not mentioned or not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0061] In several embodiments provided by the present application, it should be recognized that the disclosed systems, devices, and methods can also be implemented through other means. For example, the described device embodiments are only exemplary, and the division of the units therein is only a logical functional division. In actual implementation, there may be different division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or some steps may not be executed. In addition, the connections shown or discussed with each other, whether direct or indirect, whether coupled or communicatively connected, may be implemented in an electrical, mechanical, or other form through interfaces, devices, or units.

[0062] The units described as independent components may or may not actually be physically separated; the parts presented as units may or may not be physical entities, that is, they can be concentrated in one location or dispersed on multiple network nodes. According to actual needs, some or all of these units can be selected to achieve the objectives of this embodiment.

[0063] In addition, in various embodiments of the present application, each functional unit can either be integrated into a processing unit, exist independently physically, or combine two or more units into one unit. In the foregoing embodiments, the relevant functions can be fully or partially implemented through software, hardware, firmware, or any combination thereof. If software implementation is selected, it can be in the form of a computer program product in whole or in part. This computer program product contains one or more computer instructions. When these instructions are loaded and executed on a computer, they will generate all or part of the processes or functions that conform to the description of the embodiments of the present application. The aforementioned computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. These computer instructions can be stored in a computer-readable storage medium or transmitted through these media. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, DSL) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any computer-accessible available medium, or a data storage facility such as a server or data center that integrates one or more available media. These available media may include magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), semiconductor media (such as SSDs), etc. Those skilled in the art should understand that all or part of the processes for implementing the methods of the above embodiments can be completed through computer program instructions related to hardware, and these programs can be stored in a computer-readable storage medium. When these programs are executed, they will include the processes of the above method embodiments. The above storage media include, but are not limited to, various media that can store program codes such as read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

Claims

1. A semi-supervised medical image segmentation method based on contrast manifold regularization, characterized in that, The method includes: Importing and preprocessing a medical image dataset, where the medical image dataset contains labeled data and unlabeled data; Initializing a teacher model and a student model, training the teacher model using the labeled data to obtain teacher weights; screening the unlabeled data to obtain screened unlabeled data; based on the teacher weights, using the teacher model to generate pseudo-labels for the screened unlabeled data to obtain pseudo-labeled data; Calculating the similarity between the labeled data and the pseudo-labeled data to obtain a reproducing kernel Hilbert space manifold regularization term; Based on the labeled data and the pseudo-labeled data, performing random cropping to construct positive and negative sample pairs; where the positive sample pair is two different randomly cropped images of the same medical image; the negative sample pair is different medical images; using the positive sample pair and the negative sample pair to calculate a contrast loss function to obtain a contrast loss term; Performing weighted summation of the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain a contrast manifold regularization term; Obtaining a supervised loss based on the labeled data, and then performing weighted summation of the contrast manifold regularization term and the supervised loss as the loss function of the student model; Based on the labeled data and the pseudo-labeled data, using the loss function to iteratively train the student model to obtain student weights; based on the student weights, using the student model to perform an image segmentation task to obtain a segmentation result.

2. The method according to claim 1, wherein The screening of the unlabeled data to obtain screened unlabeled data includes: Screening based on data quality, where the data quality includes image clarity and contrast; Screening based on the similarity between the unlabeled data and the labeled data, and the similarity is measured by calculating the feature distance between the unlabeled data and the labeled data.

3. The method according to claim 1, wherein The calculating the similarity between the labeled data and the pseudo-labeled data to obtain a reproducing kernel Hilbert space manifold regularization term includes: Selecting a kernel function; calculating a similarity matrix, where the similarity matrix contains similarity values between the labeled data and the pseudo-labeled data; constructing the reproducing kernel Hilbert space manifold regularization term based on the similarity matrix.

4. The method according to claim 1, wherein The performing random cropping based on the labeled data and the pseudo-labeled data to construct positive and negative sample pairs includes: Setting a cropping size, where the cropping size is determined according to the resolution of the medical image and the size of the lesion; Randomly selecting cropping positions, where the cropping positions are independently selected on the labeled data and the pseudo-labeled data; Constructing positive sample pairs, where the positive sample pair is two different randomly cropped images of the same medical image; constructing negative sample pairs, where the negative sample pair is different medical images.

5. The method according to claim 1, wherein The performing weighted summation of the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain a contrast manifold regularization term includes: The contrast loss term is used to measure the distance difference between positive and negative sample pairs in the feature space, and the formula is: ， Among them, and are the feature representations of the positive sample pairs, is the feature representation of the negative sample pairs, is the similarity function, is the temperature parameter; Performing weighted summation of the manifold regularization term and the contrast loss term to obtain the contrast manifold regularization term, and the formula is: ， Among them, is the reproducing kernel Hilbert space manifold regularization term, is the contrastive loss term, is the weight of the contrastive loss term.

6. The method according to claim 1, wherein Obtaining a supervised loss based on the labeled data, and then performing a weighted sum of the contrast manifold regularization term and the supervised loss as the loss function of the student model, includes: The supervised loss function is obtained by the student model predicting the labeled data, and is used to measure the difference between the prediction result of the student model on the labeled data and the true label; Performing a weighted sum of the supervised loss term and the contrast manifold regularization term to obtain the loss function of the student model, using the formula: ， Among them, is the supervision loss function, is the weight of the contrast manifold regularization term; The loss function of the student model is used to guide the training process of the student model.

7. The method according to claim 1, characterized in that, Based on the labeled data and the pseudo-labeled data, iteratively training the student model using the loss function to obtain student weights, includes: Setting an iteration end condition and initializing training parameters; Calculating the loss values of the student model on the labeled data and the pseudo-labeled data using the loss function, and updating the parameters of the student model according to the loss values; Repeating the iteration process until the iteration end condition is satisfied.

8. A semi-supervised medical image segmentation device based on contrast manifold regularization, characterized in that, Includes: An input unit for importing and preprocessing a medical image dataset, where the medical image dataset includes labeled data and unlabeled data; A contrast learning unit, based on the labeled data and the pseudo-labeled data, performing random cropping to construct positive and negative sample pairs; where the positive sample pair is two different randomly cropped images of the same medical image; the negative sample pair is different medical images; calculating a contrast loss function using the positive sample pair and the negative sample pair to obtain a contrast loss term; A manifold regularization unit for calculating the similarity between the labeled data and the pseudo-labeled data to obtain a reproducing kernel Hilbert space manifold regularization term; A model training unit for initializing a teacher model and a student model, training the teacher model using the labeled data to obtain teacher weights; screening the unlabeled data to obtain screened unlabeled data; generating pseudo-labels for the screened unlabeled data using the teacher model based on the teacher weights to obtain pseudo-labeled data; also for performing a weighted sum of the reproducing kernel Hilbert space manifold regularization term and the contrast loss term to obtain a contrast manifold regularization term; obtaining a supervised loss based on the labeled data, and then performing a weighted sum of the contrast manifold regularization term and the supervised loss as the loss function of the student model; based on the labeled data and the pseudo-labeled data, iteratively training the student model using the loss function to obtain student weights; An output unit for performing an image segmentation task using the student model based on the student weights to obtain a segmentation result.

9. An electronic device, characterized in that, Includes: A processor and a storage unit, where the storage unit is used to store computer program code, and the code includes computer instructions, and when the processor executes these instructions, the electronic device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, and the computer program includes program instructions, and when the program instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 7.

Images