Self-paced fuzzy clustering method and system for incomplete multiple views of medical image

Abstract

The invention discloses a self-paced fuzzy clustering method and system for incomplete multiple views of a medical image, and relates to the field of intelligent analysis of medical images. The problems that an existing multi-view clustering technology cannot effectively process missing part of view data, organization structure fuzziness is caused and a progressive dynamic sample selection mechanism is lacked are solved. The method comprises the steps of collecting multi-view medical image data, constructing a view missing mark matrix, and performing spatial registration and feature standardization processing; complementing missing view features based on a K neighbor neighborhood weighted reconstruction strategy of a feature space; standard multi-view fuzzy clustering is adopted, a combined fuzzy clustering objective function fusing view missing marks is constructed, an adaptive view weight distribution strategy and a self-paced learning strategy are adopted, and the confidence coefficient of a clustering model of fuzzy membership entropy is defined; and respectively updating the membership degree, the clustering center and the view weight of the clustering model by using an alternate optimization strategy, and completing self-step fuzzy clustering for the incomplete multiple views of the medical image.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical image analysis technology, specifically to a self-synchronizing fuzzy clustering method and system for incomplete multi-view medical images. Background Technology

[0002] Medical imaging is an important tool for clinical diagnosis and disease analysis. Clinically, multiple views (such as axial, coronal, and sagittal) of the same lesion or tissue structure are often acquired to achieve more accurate diagnosis and analysis. According to relevant research reports, in actual clinical data acquisition, due to the high radiation risk and cost of certain imaging methods, the scarcity of research subjects, patient privacy, and equipment limitations, some view image data are often missing. This situation is called the "incomplete multiview problem".

[0003] Existing multi-view image analysis methods typically assume data completeness. For example, Chinese patent application publication number CN110188825A discloses an image clustering method based on discrete multi-view clustering to cluster complete multi-view images. However, this patented method is difficult to apply to actual clinical data with high view missing rates. Most existing methods for processing missing view data use simple imputation (such as mean imputation), which easily accumulates errors, resulting in insufficient clustering stability and poor accuracy. In addition, tissue boundaries in medical images are often blurred and gradual, and traditional hard clustering algorithms (such as K-means and spectral clustering) cannot effectively capture this characteristic. Although fuzzy clustering methods can handle the problem of blurred boundaries, such as Chinese patent application publication number CN103353986A, which discloses a method for segmenting brain MR images using fuzzy clustering, it only applies to single-view data and cannot effectively utilize multi-view information.

[0004] In recent years, Self-Paced Learning (SPL) has been widely applied in image classification and object detection as a progressive learning method from easy to difficult. For example, Chinese patent application publication number CN112598060A discloses a multi-view clustering method based on self-paced learning. This method constructs the graph structure of each view through adaptive graph learning and uses the SPL weight matrix to measure the complexity of the samples, transitioning from fewer "simple" samples to more "complex" samples to improve robustness to noise and outliers. However, this method has the following shortcomings: First, it assumes that the multi-view data is complete, making it difficult to handle image data with high missing rates in clinical settings; second, the clustering process is based on hard segmentation and does not utilize fuzzy membership output, failing to characterize the fuzzy and gradual boundary changes in medical images for clinical data; third, it selects samples only through node-level complexity measurement, lacking a comprehensive evaluation of view-level feature reconstruction errors and clustering confidence. Currently, there is no effective method for applying SPL to the field of incomplete multi-view clustering of medical images. Therefore, how to simultaneously address the issues of missing views and blurred tissue boundaries in medical imaging data, and improve the accuracy and stability of cluster analysis, is a pressing technical challenge in the field. Summary of the Invention

[0005] This invention aims to address the problems of existing multi-view clustering techniques, such as their inability to effectively handle missing view data, ambiguous organizational structures, and the lack of a progressive dynamic sample selection mechanism. To solve these technical problems, this invention achieves the following technical solution: Option 1: This invention proposes a self-synchronizing fuzzy clustering method for incomplete multi-view medical images, the method comprising the following steps: Step 1: Collect multi-view medical image data, construct a view missing marker matrix based on the collected multi-view medical image data, and perform spatial registration and feature standardization processing; Step 2: Based on the K-nearest neighbor weighted reconstruction strategy in the feature space, the missing view features in Step 1 are completed to accurately estimate the feature vector of the missing view. Step 3: Based on the features completed in Step 2, standard multi-view fuzzy clustering is used to quickly lock the coarse clustering structure of the data, and the membership update, cluster center update and view weight update of the clustering model are output as the initial state of the self-step learning stage. Step 4: Construct a joint fuzzy clustering objective function for missing labels in the fused view, adopt an adaptive view weight allocation strategy, and introduce a local neighborhood structure to preserve the view's weights. Figure 1 To the point of being compatible; Step 5: Based on the self-stepping learning strategy in Step 4, define the confidence of the clustering model based on fuzzy membership entropy, and optimize the clustering model. Step 6: Use an alternating optimization strategy to update the membership degree, cluster center, and view weight of the clustering model in Step 4 until convergence, thus completing the self-synchronizing fuzzy clustering for incomplete multi-view medical images.

[0006] Furthermore, a preferred embodiment is provided, wherein the method for spatial registration and feature standardization of the acquired multi-view medical image data in step 1 is as follows: The image size was standardized to 256×256 pixels using a linear interpolation algorithm. The image grayscale values ​​are min-max normalized to the interval [0, 1]. Use the SimpleITK toolkit for rigid spatial registration to ensure spatial consistency across different views; Construct a view missing marker matrix, the matrix is This is used to indicate whether the data in each view is complete. .

[0007] Furthermore, a preferred implementation is provided, wherein step 2 uses a K-nearest neighbor weighted reconstruction strategy based on the feature space. The method for completing the missing view features in step 1 is as follows: Step 2.1: For each sample i, calculate the average vector of its observed view features, denoted as . , ,in This represents the set of views in which sample i exists. Step 2.2: Determine the weight vector using local linear regression. , , in Its closed-form solution is: ; Step 2.3: Perform feature completion on patient i with missing view v to obtain the completed vector. , .

[0008] Furthermore, a preferred embodiment is provided, wherein the method for calculating the membership degree of the output clustering model in step 3 is as follows: Membership update formula: ; in, ; Cluster center update formula: ; View weight update formula: ; When the number of iterations reaches the preset maximum or the update magnitude is extremely small, stop the iteration and assign the three values ​​obtained in the last round to the initial values. , , .

[0009] Furthermore, a preferred embodiment is provided, wherein the method for constructing the joint fuzzy clustering objective function for missing markers in the fused view in step 4 is as follows: Step 4.1: Define the main objective function for multi-view fuzzy clustering. : , in The filtering mask will be provided in step 5; fuzzy factor Take 2 experience points; Step 4.2: Calculate the local neighborhood preservation regularization term. : , in , The value is 0.1, determined by cross-validation.

[0010] Step 4.3, Calculate the view Figure 1 Consistency constraints : The value is 0.1, determined by cross-validation.

[0011] Step 4.4, Overall Objective Function: .

[0012] Furthermore, a preferred implementation is provided, wherein in step 5, the confidence level of the clustering model based on fuzzy membership entropy is defined through a self-stepping learning strategy, and the method for optimizing the clustering model is as follows: Step 5.1: For the current t-th iteration, calculate the membership degree of each sample i to cluster c:

[0013] in, , Step 5.2: Define the sample clustering confidence score: ; Step 5.3, Sample Screening Set the confidence threshold in each iteration With tolerance for missing views : Initial threshold =0.9, missing view tolerance number T_0=0; In each iteration, Decrease by 0.05, Increase by 1; Samples that meet the criteria will participate in cluster optimization, while samples that do not meet the criteria will be temporarily deferred. Sample selection markers: ; when When sample i is selected to participate in training in this iteration; when At that time, sample i was not included in the sample set for training in this iteration.

[0014] Furthermore, a preferred embodiment is provided, wherein the method for updating the membership degree, cluster center, and view weight of the clustering model in step 4 using an alternating optimization strategy in step 6 is as follows: Membership update formula: ; Cluster center update formula: ; View weight update formula: ; The above updates are performed alternately, and the sample set for self-learning is gradually expanded.

[0015] Option 2: A self-synchronizing fuzzy clustering system for incomplete multi-view medical images, the system comprising: The data standardization module is used to collect multi-view medical image data, construct a view missing marker matrix based on the collected multi-view medical image data, and perform spatial registration and feature standardization processing. The feature completion module is used to complete the missing view features in the data standardization module based on the K-nearest neighbor weighted reconstruction strategy in the feature space, so as to accurately estimate the feature vector of the missing view. The multi-view fuzzy clustering module is used to apply standard multi-view fuzzy clustering to the features completed by the feature completion module, quickly lock the coarse clustering structure of the data, and output the membership update, cluster center update and view weight update of the clustering model as the initial state of the self-step learning stage. The fuzzy clustering objective function construction module is used to construct a joint fuzzy clustering objective function for fused views with missing labels. It adopts an adaptive view weight allocation strategy and introduces local neighborhood structure to preserve the view's weights. Figure 1 To the point of being compatible; The clustering model optimization module is used to define the confidence of the clustering model based on fuzzy membership entropy and optimize the clustering model by using a self-stepping learning strategy based on the fuzzy clustering objective function construction module. The self-stepping fuzzy clustering module is used to update the membership degree, cluster center, and view weight of the clustering model in the fuzzy clustering objective function construction module using an alternating optimization strategy until convergence, thus completing the self-stepping fuzzy clustering for incomplete multi-view medical images.

[0016] Option 3: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in Option 1.

[0017] Option 4: A computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the method described in Option 1.

[0018] The advantages of this invention are: The self-synchronized fuzzy clustering method and system for incomplete multi-view medical images described in this invention adopts a cross-view feature completion strategy based on a combination of neighborhood weighting and local linear regression. This strategy can more accurately reconstruct missing view features in environments with high missing rates, reduce completion errors, and thus enhance the stability of the clustering process. This invention incorporates view weight adaptive optimization, local neighborhood structure preservation, and cross-view optimization. Figure 1 Integrating consistency constraints into the fuzzy clustering objective function is more helpful in capturing the fuzzy tissue structure features in medical images than traditional hard clustering or single-view fuzzy clustering methods. This invention combines a confidence measure based on membership entropy with the SPL sample selection mechanism, which can gradually introduce samples in an order of "from easy to difficult". Theoretically, this helps to avoid optimization oscillations caused by noise or missing data, thereby improving the robustness of the overall clustering. The method described in this invention has a clear structure, well-defined parameters, is easy to implement, and is applicable to various important medical image analysis scenarios such as brain, lung, and liver, demonstrating significant clinical applicability and promotional value.

[0019] In summary, the method proposed in this invention effectively solves the problems of missing views, blurred tissue structures, and lack of robust optimization mechanisms in medical image analysis, improves the accuracy and robustness of medical image clustering analysis, and has important theoretical and clinical application value.

[0020] This invention is also applicable to fields such as medical image processing, pattern recognition, and the intersection of artificial intelligence. Attached Figure Description

[0021] Figure 1 This is a flowchart of the self-synchronizing fuzzy clustering method for incomplete multiple views of medical images as described in Implementation Method 1. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0023] Implementation Method 1: This implementation method proposes a self-synchronizing fuzzy clustering method for incomplete multi-view medical images. The method includes the following steps: Step 1: Collect multi-view medical image data, construct a view missing marker matrix based on the collected multi-view medical image data, and perform spatial registration and feature standardization processing; Step 2: Based on the K-nearest neighbor weighted reconstruction strategy in the feature space, the missing view features in Step 1 are completed to accurately estimate the feature vector of the missing view. Step 3: Based on the features completed in Step 2, standard multi-view fuzzy clustering is used to quickly lock the coarse clustering structure of the data, and the membership update, cluster center update and view weight update of the clustering model are output as the initial state for the subsequent SPL stage. Step 4: Construct a joint fuzzy clustering objective function for missing labels in the fused view, adopt an adaptive view weight allocation strategy, and introduce a local neighborhood structure to preserve the view's weights. Figure 1 To the point of being compatible; Step 5: Based on the self-stepping learning strategy in Step 4, define the confidence of the clustering model based on fuzzy membership entropy, and optimize the clustering model. Step 6: Use an alternating optimization strategy to update the membership degree, cluster center, and view weight of the clustering model in Step 4 until convergence, thus completing the self-synchronizing fuzzy clustering for incomplete multi-view medical images.

[0024] Implementation Method Two: This implementation method further defines the self-synchronizing fuzzy clustering method for incomplete multi-view medical images described in Implementation Method One. The method for spatial registration and feature standardization of the acquired multi-view medical image data in step 1 is as follows: The image size was standardized to 256×256 pixels using a linear interpolation algorithm. The image grayscale values ​​are min-max normalized to the interval [0, 1]. Use the SimpleITK toolkit for rigid spatial registration to ensure spatial consistency across different views; Construct a view missing marker matrix, the matrix is This is used to indicate whether the data in each view is complete.

[0025] This marker is used to accurately guide subsequent feature completion.

[0026] Implementation Method 3: This implementation method further defines the self-synchronizing fuzzy clustering method for incomplete multi-view medical images described in Implementation Method 1. The method for completing the missing view features in Step 1 using the K-nearest neighbor weighted reconstruction strategy based on the feature space in Step 2 is as follows: Step 2.1: For each sample i, calculate the average vector of its observed view features, denoted as . : ,in This represents the set of views in which sample i exists. Step 2.2: Determine the weight vector using local linear regression. :

[0027] in Its closed-form solution is: ; Step 2.3: Perform feature completion on patient i with missing view v to obtain the completed vector. : .

[0028] Implementation Method Four: This implementation method further defines the self-synchronizing fuzzy clustering method for incomplete multi-view medical images described in Implementation Method One. The method for calculating the membership degree of the output clustering model in step 3 is as follows: Membership update formula: ; in, ; Cluster center update formula: ; When the number of iterations reaches the preset maximum or the update magnitude is extremely small, stop the iteration and assign the three values ​​obtained in the last round to the initial values. , , .

[0029] Implementation Method 5: This implementation method further defines the self-synchronizing fuzzy clustering method for incomplete multi-view medical images described in Implementation Method 1. The method for constructing the joint fuzzy clustering objective function for missing labels in fused views in step 4 is as follows: Step 4.1: Define the main objective function for multi-view fuzzy clustering. : , in The filtering mask will be provided in step 5; fuzzy factor Take 2 experience points; Step 4.2: Calculate the local neighborhood preservation regularization term. , in , The recommended value is 0.1, determined through cross-validation. Step 4.3, Calculate the view Figure 1 Consistency constraints : The value is 0.1; Step 4.4, Overall Objective Function: .

[0030] Implementation Method Six: This implementation method provides a preferred embodiment. In step 5, the confidence level of the clustering model based on fuzzy membership entropy is defined through a self-stepping learning strategy, and the method for optimizing the clustering model is as follows: Step 5.1: For the current t-th iteration, calculate the membership degree of each sample i to cluster c:

[0031] in,

[0032] Step 5.2: Define the sample clustering confidence score: ; Step 5.3, Sample Screening Set the confidence threshold in each iteration With tolerance for missing views : Initial threshold =0.9, missing view tolerance number T_0=0; In each iteration, Decrease by 0.05, Increase by 1; Samples that meet the criteria will participate in cluster optimization, while samples that do not meet the criteria will be temporarily deferred. Sample selection markers: ; when When sample i is selected to participate in training in this iteration; when At that time, sample i was not included in the sample set for training in this iteration.

[0033] Implementation Method Seven: This implementation method further defines the preferred implementation method described in Implementation Method Six. The method for updating the membership degree, cluster center, and view weight of the clustering model in Step 4 using an alternating optimization strategy in step 6 is as follows: Membership update formula: ; Cluster center update formula: ; View weight update formula: ; The above updates are performed alternately, and the sample set for self-learning is gradually expanded.

[0034] Implementation Method 8: This implementation method proposes a self-synchronizing fuzzy clustering system for incomplete multi-view medical images. The system includes: The data standardization module is used to collect multi-view medical image data, construct a view missing marker matrix based on the collected multi-view medical image data, and perform spatial registration and feature standardization processing. The feature completion module is used to complete the missing view features in step 1 based on the K-nearest neighbor weighted reconstruction strategy of the feature space, so as to accurately estimate the feature vector of the missing view. The multi-view fuzzy clustering module is used to apply standard multi-view fuzzy clustering to the features completed by the feature completion module, quickly lock the coarse clustering structure of the data, and output the membership update, cluster center update and view weight update of the clustering model as the initial state for the subsequent SPL stage. The fuzzy clustering objective function construction module is used to construct a joint fuzzy clustering objective function for fused views with missing labels. It adopts an adaptive view weight allocation strategy and introduces local neighborhood structure to preserve the view's weights. Figure 1 To the point of being compatible; The clustering model optimization module is used to define the confidence of the clustering model based on fuzzy membership entropy and optimize the clustering model by using a self-stepping learning strategy based on the fuzzy clustering objective function construction module. The self-stepping fuzzy clustering module is used to update the membership degree, cluster center, and view weight of the clustering model in the fuzzy clustering objective function construction module using an alternating optimization strategy until convergence, thus completing the self-stepping fuzzy clustering for incomplete multi-view medical images.

[0035] Implementation Method Nine: This implementation method provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in any one of Implementation Methods One to Seven. Implementation Method 10: This implementation method provides a computer device, including a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the method described in any one of Implementation Methods 1 to 7.

[0036] Implementation Method Eleven: This implementation method provides an example, which is used to explain the above-described implementation methods one through ten. The specific example is as follows: See Figure 1 This implementation method is described with reference to Figure 1As shown, this embodiment proposes a self-paced fuzzy clustering method for incomplete multi-view medical images. The core technology of this invention lies in the innovative combination of a self-paced learning (SPL) strategy with a fuzzy clustering model optimized for missing multi-view data, so as to achieve more efficient and robust automatic clustering analysis of medical image data.

[0037] The specific technical solution of this embodiment includes the following key steps: Step 1: Collect multi-view medical image data, construct a view missing marker matrix, clearly identify the missing status of each sample view, and perform spatial registration and feature standardization preprocessing; Step 2: Using a K-nearest neighbor weighted reconstruction strategy based on feature space, missing view features are filled in to accurately estimate the feature vector of the missing view. Step 3: Run standard multi-view fuzzy clustering based on all completed features. This stage assumes all views exist and does not perform sample filtering, aiming to quickly determine the coarse clustering structure of the data. The membership distribution, cluster centers, and view weights output from this stage are used as the initial state for subsequent SPL stages. Step 4: Construct a joint fuzzy clustering objective function for missing labels in the fused view, adopt an adaptive view weight allocation strategy, and introduce local neighborhood structure preservation and view... Figure 1 Consistency regularization terms are used to improve the accuracy and spatial consistency of cluster analysis. Step 5: Innovatively introduce a self-stepping learning strategy, define clustering confidence based on fuzzy membership entropy, design a dynamic sample introduction mechanism that gradually increases sample difficulty and tolerance for missing views from easy to difficult, and robustly optimize the clustering model; Step 6: Use an alternating optimization strategy to update the membership degree, cluster center, and view weight of the clustering model until convergence, and obtain the accurate clustering analysis results of medical images.

[0038] This invention uses multi-view brain MRI medical images as an example to implement a specific self-stepping fuzzy clustering method for cases with incomplete multi-view images (i.e., missing views). In this embodiment, the following notation is used: the number of iterations is set as t, and the initial value is 0. It is assumed that there are a total of... Samples were collected from 10 patients, and samples were collected from each patient. A medical image view; for each image, a dimension of [dimension not specified] is extracted using a convolutional neural network. The feature vector. Furthermore, the preset number of clusters in cluster analysis is [number missing]. When performing cross-view feature completion, the nearest neighbor number K is used. The membership matrix is ​​as follows: , Let be the membership degree of the sample to each cluster center, and The cluster center matrix is The view weight vector is ,and These symbols will remain consistent throughout all subsequent formulas and algorithm descriptions. The specific process and formula derivation are as follows: Step 1: Standardization and Missing Data of Medical Imaging Multiview First, the following standardized procedures should be performed on clinically acquired brain MRI images (such as axial, coronal, and sagittal views): (1) Use a linear interpolation algorithm to unify the image size to 256×256 pixels; (2) Normalize the image grayscale values ​​to the interval [0, 1] using the minimum-maximum method; (3) Use the SimpleITK toolkit to perform rigid spatial registration to ensure spatial consistency of different views.

[0039] For each patient sample i, construct a view missing marker matrix, the matrix is ​​as follows: This is used to indicate whether the data in each view is complete.

[0040] This marker is used to accurately guide subsequent feature completion.

[0041] Step 2: Multi-view feature extraction and cross-view feature completion Each view uses a convolutional neural network (CNN) to extract features individually. The network structure is as follows: Convolutional layer 1: Convolutional kernel (3×3, 32 channels) + BatchNorm + ReLU; Convolutional layer 2: Convolutional kernel (3×3, 64 channels) + BatchNorm + ReLU; Maximum pooling layer: 2×2; Convolutional layer 3: Convolutional kernel (3×3, 128 channels) + BatchNorm + ReLU; Global average pooling outputs a 128-dimensional feature vector. .

[0042] In this scheme, to improve the accuracy and stability of missing view feature completion, we introduce a Local Linear Regression (LLR) mechanism on top of the classic KNN weighted reconstruction. Specifically, for each sample i, we first select the K nearest neighbors around its average features, and then automatically learn a set of optimal linear coefficients through least squares regression to minimize the reconstruction error within the neighborhood. The regression coefficients are dynamically adjusted according to the local sample distribution, which can more accurately characterize the neighborhood structure and improve the stability of completion.

[0043] First, for each sample i, calculate the average vector of its observed view features, denoted as . : ,in The set of views representing the existence of sample i.

[0044] For patient i with a missing view v, the average feature vector of existing views is based on samples in the feature space. Calculate the Euclidean distance to features of other samples, and take the K nearest neighbors, i.e. (Empirical choice: K=10) Feature completion is performed.

[0045] set up Constructing a matrix .

[0046] Then, local linear regression is used to determine the weight vector. : , in Its closed-form solution is: To prevent overfitting of sample features within a local neighborhood, a regularization parameter is introduced. Controlling the norm of the linear regression coefficients Smaller values ​​result in a more accurate and complete feature vector. However, this may lead to overfitting and unstable completion; larger values ​​result in smoother and more stable feature vector completion, but the completion error may be larger. Generally... =0.05, which can be further fine-tuned on the validation set.

[0047] Finally, feature completion is performed on patient i with missing view v to obtain the completed vector. ; Step 3: Initialize Global Fuzzy Clustering (FCM) Before proceeding to the SPL iteration, based on the view completed in step two, run a standard multi-view FCM once on all samples to obtain high-quality initial memberships. Cluster center With view weight The update formula for the t-th iteration is as follows: Membership update formula: ; in, .

[0048] Cluster center update formula: ; View weight update formula: ; When the number of iterations reaches the preset maximum or the update magnitude is extremely small, stop the iteration and assign the three values ​​obtained in the last round to the initial values. , , .

[0049] Step 4: Construction of Self-Propelled Multi-View Fuzzy Clustering Model Constructing missing markers in the fused view, preserving local neighborhood structure, and view Figure 1 Objective function for multi-view fuzzy clustering with consistency constraints: (1) Define the main objective function for multi-view fuzzy clustering : ; in The filtering mask will be provided in step five; fuzzy factor Take 2 experience points.

[0050] (2) Local neighborhood preservation regularization term : ; in , The recommended value is 0.1, determined by cross-validation.

[0051] (3) Vision Figure 1 Consistency constraints : ; The recommended value is 0.1, determined by cross-validation.

[0052] (4) Overall objective function: .

[0053] Step 5: Self-Progressive Learning (SPL) Dynamic Sample Selection Mechanism First, for the current t-th iteration, calculate the membership degree of each sample i to cluster c:

[0054] in,

[0055] Define the sample clustering confidence (using membership entropy): Set a confidence threshold for each iteration. With tolerance for missing views : (1) Initial threshold =0.9, missing view tolerance number T_0=0; (2) In each iteration, Decrease by 0.05, Increase by 1; (3) Samples that meet the conditions participate in cluster optimization, while samples that do not meet the conditions are temporarily suspended from being introduced.

[0056] Sample selection markers: ; when When sample i is selected to participate in training in this iteration; when At that time, sample i was not included in the sample set for training in this iteration.

[0057] Step Six: Alternate Optimization of Model Parameters (Iterative Update) Update sequentially using an alternating optimization strategy: Membership update formula: ; Cluster center update formula: ; View weight update formula: .

[0058] The above updates are performed alternately, and the sample set for self-learning is gradually expanded.

[0059] Step 7: Model Optimization Termination Conditions and Clustering Result Output The model stops iterating when the cluster center moves less than a set threshold or reaches a set iteration limit. The final model outputs the cluster classification and fuzzy membership degree for each sample under multi-view conditions.

[0060] In summary, the method proposed in this embodiment effectively solves the problems of missing views, blurred tissue structures, and lack of robust optimization mechanisms in medical image analysis, improves the accuracy and robustness of medical image clustering analysis, and has important theoretical and clinical application value.

[0061] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or technical solutions of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. This is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0062] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended technical solutions are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the present invention. Clearly, those skilled in the art can make various modifications and variations to the present invention without departing from its spirit and scope. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention also intends to include these modifications and variations.

Claims

1. A self-synchronizing fuzzy clustering method for incomplete multi-view medical images, characterized in that, The method includes the following steps: Step 1: Collect multi-view medical image data, construct a view missing marker matrix based on the collected multi-view medical image data, and perform spatial registration and feature standardization processing; Step 2: Based on the K-nearest neighbor weighted reconstruction strategy in the feature space, the missing view features in Step 1 are completed to accurately estimate the feature vector of the missing view. Step 3: Based on the features completed in Step 2, standard multi-view fuzzy clustering is used to quickly lock the coarse clustering structure of the data, and the membership update, cluster center update and view weight update of the clustering model are output as the initial state of the self-step learning stage. Step 4: Construct a joint fuzzy clustering objective function for missing labels in the fused view, adopt an adaptive view weight allocation strategy, and introduce a local neighborhood structure to maintain view consistency; Step 5: Based on the self-stepping learning strategy in Step 4, define the confidence of the clustering model based on fuzzy membership entropy, and optimize the clustering model. Step 6: Use an alternating optimization strategy to update the membership degree, cluster center, and view weight of the clustering model in Step 4 until convergence, thus completing the self-synchronizing fuzzy clustering for incomplete multi-view medical images.

2. The self-synchronizing fuzzy clustering method for incomplete multi-view medical images according to claim 1, characterized in that, The method for spatial registration and feature standardization of the acquired multi-view medical image data in step 1 is as follows: The image size was standardized to 256×256 pixels using a linear interpolation algorithm. The image grayscale values ​​are min-max normalized to the interval [0, 1]. Use the SimpleITK toolkit for rigid spatial registration to ensure spatial consistency across different views; Construct a view missing marker matrix, the matrix is This is used to indicate whether the data in each view is complete. 。 3. The self-synchronizing fuzzy clustering method for incomplete multi-view medical images according to claim 1, characterized in that, In step 2, the method for completing the missing view features from step 1 using the K-nearest neighbor weighted reconstruction strategy based on the feature space is as follows: Step 2.1: For each sample i, calculate the average vector of its observed view features, denoted as . , in, This represents the set of views in which sample i exists. Step 2.2: Determine the weight vector using local linear regression. : in Its closed-form solution is: ; Step 2.3: Perform feature completion on patient i with missing view v to obtain the completed vector. , 。 4. The self-synchronizing fuzzy clustering method for incomplete multi-view medical images according to claim 1, characterized in that, The method for calculating the membership degree of the output clustering model in step 3 is as follows: Membership update formula: ； in, ; Cluster center update formula: ； When the number of iterations reaches the preset maximum or the update magnitude is extremely small, stop the iteration and assign the three values ​​obtained in the last round to the initial values. , , .

5. The self-synchronizing fuzzy clustering method for incomplete multi-view medical images according to claim 1, characterized in that, The method for constructing the joint fuzzy clustering objective function for missing markers in the fused view in step 4 is as follows: Step 4.1: Define the main objective function for multi-view fuzzy clustering. : in The filtering mask will be provided in step 5; fuzzy factor Take 2 experience points; Step 4.2: Calculate the local neighborhood preservation regularization term. : in , The value is 0.1, determined by cross-validation; Step 4.3: Calculate view consistency constraints. : The value is 0.1; Step 4.4, Overall Objective Function: .

6. The self-synchronizing fuzzy clustering method for incomplete multi-view medical images according to claim 1, characterized in that, In step 5, a self-stepping learning strategy is used to define the confidence level of the clustering model based on fuzzy membership entropy, and the method for optimizing the clustering model is as follows: Step 5.1: For the current t-th iteration, calculate the membership degree of each sample i to cluster c: in, ; Step 5.2: Define the sample clustering confidence score: ; Step 5.3, Sample Screening Set the confidence threshold in each iteration With tolerance for missing views ; Initial threshold =0.9, missing view tolerance number T_0=0; In each iteration, Decrease by 0.05, Increase by 1; Samples that meet the criteria will participate in cluster optimization, while samples that do not meet the criteria will be temporarily deferred. Sample selection markers: , when When sample i is selected to participate in training in this iteration; when At that time, sample i was not included in the sample set for training in this iteration.

7. The self-synchronizing fuzzy clustering method for incomplete multi-view medical images according to claim 1, characterized in that, The method for updating the membership degree, cluster center, and view weight of the clustering model in step 4 using the alternating optimization strategy in step 6 is as follows: Membership update formula: ; Cluster center update formula: ; View weight update formula: ; The above updates are performed alternately, and the sample set for self-learning is gradually expanded.

8. A self-synchronizing fuzzy clustering system for incomplete multi-view medical images, characterized in that, The system includes: The data standardization module is used to collect multi-view medical image data, construct a view missing marker matrix based on the collected multi-view medical image data, and perform spatial registration and feature standardization processing. The feature completion module is used to complete the missing view features in the data standardization module based on the K-nearest neighbor weighted reconstruction strategy in the feature space, so as to accurately estimate the feature vector of the missing view. The multi-view fuzzy clustering module is used to perform standard multi-view fuzzy clustering based on the features completed by the feature completion module. It is used to quickly lock the coarse clustering structure of the data and output the membership update, cluster center update and view weight update of the clustering model as the initial state for the subsequent SPL stage. The fuzzy clustering objective function construction module is used to construct a joint fuzzy clustering objective function that integrates missing labels in the view. It adopts an adaptive view weight allocation strategy and introduces a local neighborhood structure to maintain view consistency. The clustering model optimization module is used to define the confidence of the clustering model based on fuzzy membership entropy and optimize the clustering model by using a self-stepping learning strategy based on the fuzzy clustering objective function construction module. The self-stepping fuzzy clustering module is used to update the membership degree, cluster center, and view weight of the clustering model in the fuzzy clustering objective function construction module using an alternating optimization strategy until convergence, thus completing the self-stepping fuzzy clustering for incomplete multi-view medical images.

9. A computer storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-7.

10. A computer device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method of any one of claims 1-7.

Images