Pancreatic image segmentation method and system based on adaptive learning framework

Abstract

The invention discloses a pancreas image segmentation method and system based on an adaptive learning framework, and belongs to the technical field of medical image segmentation. Inputting the pancreas image into an adaptive learning framework for segmentation to obtain a segmentation result; the adaptive learning framework comprises a different kernel feature fusion module and an adaptive attention contrast learning module, a pancreas image to be segmented is divided into a plurality of input image blocks, the input image blocks are input to the adaptive attention contrast learning module, a plurality of feature maps are obtained through convolution operations of different kernel sizes, and the feature maps are input to the adaptive attention contrast learning module. And inputting the plurality of feature maps into different kernel feature fusion modules to perform inter-class aggregation between the same features and inter-class separation between different features, and then outputting a plurality of classification features through a feature comparison learning process to obtain a segmentation result. And feature information is obtained through convolution kernels of different sizes and collaborative double attention branches, so that learning of pancreas details and overall structure features is enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image segmentation, and in particular relates to a pancreatic image segmentation method and system based on a co-adaptive learning framework. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] Pancreatic cancer is a highly lethal malignancy with an extremely low five-year survival rate. Early diagnosis and timely intervention can significantly slow disease progression. Accurate segmentation of the pancreas from CT images is essential for clinicians to monitor pancreatic abnormalities, assist in prevention, early diagnosis, and surgical treatment of pancreatic cancer. Since manual annotation of pancreatic boundaries by radiologists is time-consuming and laborious and there are inter-observer differences, there is an urgent need to explore deep learning methods that can automatically identify and segment the pancreas.

[0004] As a landmark architecture in the field of medical image segmentation, U-Net has been widely used for automatic pancreas segmentation. However, due to the complex morphology of the pancreas, unclear boundaries with adjacent organs, and less than 0.5% of the total abdominal volume, the classic U-Net algorithm faces major challenges. Therefore, researchers are committed to enhancing the U-Net model and its variants to improve feature extraction capabilities. Multi-scale feature fusion can be used to effectively capture multi-scale features, while large kernel convolution provides a simpler structure while providing a larger receptive field and high-resolution features. However, direct multi-scale feature fusion may lead to information redundancy. The attention mechanism can not only alleviate the problem of redundant features, but also reduce the uncertainty of pixel classification. However, the existing attention mechanism does not fully utilize the similarity learning between features of different scales, which is crucial to enhancing the model's ability to extract similar features and suppress non-critical difference features. Summary of the invention

[0005] To overcome the above-mentioned deficiencies of the prior art, the present invention provides a pancreatic image segmentation method and system based on an adaptive learning framework, which acquires feature information through convolution kernels of different sizes and collaborative dual attention branches, thereby enhancing the learning of pancreatic details and overall structural features.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, the present invention provides a pancreatic image segmentation method based on a co-adaptive learning framework, comprising: Acquire a pancreas image to be segmented; The pancreatic image is input into a co-adaptive learning framework for segmentation to obtain a segmentation result; the co-adaptive learning framework includes a different kernel feature fusion module and an adaptive attention contrast learning module, the pancreatic image to be segmented is divided into a plurality of input image blocks, the plurality of input image blocks are input into the adaptive attention contrast learning module, a plurality of feature maps are obtained after convolution operations with different kernel sizes, the plurality of feature maps are input into a different kernel feature fusion module for inter-class aggregation between the same features and inter-class separation between different features, and then a plurality of classification features are output through a feature contrast learning process to obtain a segmentation result.

[0007] According to a further technical solution, the different kernel feature fusion module includes a parallel small kernel convolution module and a large kernel convolution module, a parallel global context perception branch and a detail sensitive feature enhancement branch.

[0008] A further technical solution is that after the global context-aware branch performs different convolution operations on the output features of the small-core convolution module and the output features of the large-core convolution module, complementary features are generated by element-by-element addition, and the complementary features are adaptively average pooled to generate multiple compressed features, and then a comprehensive weight map is generated by element-by-element addition.

[0009] A further technical solution adopts a recalibration mechanism to adjust the importance of each feature element in the comprehensive weight map to obtain globally context-aware perceptual features.

[0010] According to a further technical solution, the complementary features in the detail-sensitive feature enhancement branch are sequentially subjected to the operations of GELU nonlinear activation function, convolution, and Sigmoid activation function to obtain interactive features, and the interactive features are element-wise multiplied with the complementary features to obtain detail-sensitive features.

[0011] A further technical solution is to design a contrast loss between the global context-aware branch and the detail-sensitive feature enhancement branch. The loss function is defined as follows:

[0012]

[0013]

[0014] in, Represents contrast loss The weight of represents the contrast loss, Expressed as the loss weight of each layer encoder, Indicates sensory loss, represents the Euler number, Indicated in The perceptual features in the layer encoder The value of a voxel, Indicated in The detail-sensitive features in the layer encoder The value of a voxel, Indicated in The perceptual features in the layer encoder belong to Channel No. The value of a voxel, Indicated in The detail-sensitive features in the layer encoder belong to Channel No. The value of the voxel.

[0015] In a further technical solution, the mutual adaptation learning framework uses a boundary sensitive region optimization strategy and defines a boundary sensitive region optimization loss function, which is expressed as:

[0016]

[0017] in, represents the weight coefficient, represents the Dice loss, represents the area loss, represents the number of model training iterations, Represents the total number of model training iterations.

[0018] In a second aspect, the present invention provides a pancreatic image segmentation system based on a co-adaptive learning framework, comprising: A data acquisition module is configured to: acquire a pancreas image to be segmented; The image segmentation module is configured as follows: the pancreatic image is input into a co-adaptive learning framework for segmentation to obtain a segmentation result; the co-adaptive learning framework includes a different kernel feature fusion module and an adaptive attention contrast learning module, the pancreatic image to be segmented is divided into a plurality of input image blocks, the plurality of input image blocks are input into the adaptive attention contrast learning module, a plurality of feature maps are obtained after convolution operations with different kernel sizes, the plurality of feature maps are input into a different kernel feature fusion module for inter-class aggregation between the same features and inter-class separation between different features, and then a plurality of classification features are output through a feature contrast learning process to obtain a segmentation result.

[0019] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the pancreatic image segmentation method based on the co-adaptive learning framework as described in the first aspect.

[0020] In a fourth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the pancreatic image segmentation method based on the co-adaptive learning framework as described in the first aspect are implemented.

[0021] One or more of the above technical solutions have the following beneficial effects: The present invention utilizes the collaborative dual attention branches in the different kernel feature fusion module (AFDK) to fuse global context information and local detail features at the same time, thereby enhancing the network's feature extraction capability. Through adaptive attention contrastive learning (AACL), the present invention can dynamically adjust the similarity between features extracted by different kernels and optimize feature representation. In addition, the present invention designs a boundary sensitive region optimization (BSRO) strategy, which improves segmentation accuracy by integrating boundary information and region optimization to balance the model's attention to the core region and detail boundaries.

[0022] The present invention proposes a different kernel feature fusion module (AFDK), which acquires feature information through convolution kernels of different sizes and collaborative dual attention branches, thereby enhancing the learning of pancreatic details and overall structural features.

[0023] The present invention develops an adaptive attention contrastive learning module (AACL), which promotes dynamic similarity learning between features extracted by different kernels and optimizes the representation of key features.

[0024] The present invention designs a boundary-sensitive region optimization strategy (BSRO) to balance the model's attention to the core area and detail boundaries during training, thereby improving the performance of pancreatic segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0026] Figure 1 It is an overall structural diagram of the mutual adaptive learning framework of an embodiment of the present invention; Figure 2 It is a global context-aware branch structure diagram in different core feature fusion modules in an embodiment of the present invention; Figure 3 is a diagram of detail-sensitive feature enhancement branch structures in different core feature fusion modules according to an embodiment of the present invention; Figure 4 is a structural diagram of an adaptive attention contrast learning module according to an embodiment of the present invention; Figure 5 is the real label in the pancreas segmentation label of the embodiment of the present invention; Figure 6 It is the border position information of the pancreas label corresponding to the BLI visualization in the embodiment of the present invention; Figure 7 These are three segmentation examples of pancreas segmentation on the NIH dataset according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0029] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0030] Embodiment 1 like Figure 1 As shown, this embodiment discloses a pancreatic image segmentation method based on a co-adaptive learning framework, the method comprising the following steps: S1: Acquire the pancreas image to be segmented; S2: Input the pancreatic image into a mutually adaptive learning framework (DKMAL) for segmentation to obtain a segmentation result; the mutually adaptive learning framework includes a different kernel feature fusion module and an adaptive attention contrast learning module, and divides the pancreatic image to be segmented into multiple input image blocks. The multiple input image blocks are input into the adaptive attention contrast learning module, and multiple feature maps are obtained after convolution operations with different kernel sizes. The multiple feature maps are input into the different kernel feature fusion module to perform inter-class aggregation between the same features and inter-class separation between different features, and then multiple classification features are output through a feature contrast learning process to obtain a segmentation result.

[0031] In this embodiment, a new end-to-end network, namely a diversified kernel-based adaptive learning framework (DKMAL), is provided, which includes three core components: a different kernel feature fusion module (AFDK), an adaptive attention contrastive learning module (AACL), and a boundary sensitive region optimization strategy (BSRO) for joint segmentation of the pancreas in abdominal CT images.

[0032] In this embodiment, if Figure 1As shown in the figure, the different kernel feature fusion module AFDK includes parallel small kernel convolution modules and large kernel convolution modules, parallel global context perception branches and detail sensitive feature enhancement branches. The small kernel convolution module is a convolution of size 3×3×3 (Conv2), and the large kernel convolution module is a convolution of size 7×7×7 (Conv1).

[0033] The AFDK module emphasizes the importance of feature extraction through convolution kernels of different sizes and feature fusion through collaborative dual attention branches, considering the interaction between global context information and local details in the segmentation task. The AFDK module uses two convolution kernels of different sizes to make full use of feature representations at multiple scales to improve model performance. Specifically, the convolution operation of size (7×7×7) is used to capture broader context information, which helps to learn the pancreatic structure. The convolution operation of size (3×3×3) is used to capture more detailed and localized features. Subsequently, the above features are fed into the collaborative dual attention branches, namely the global context-aware branch (GCAB) and the detail-sensitive feature enhancement branch (DFEB), for further information extraction.

[0034] like Figure 2 As shown in the figure, the global context-aware branch (GCAB) uses global average pooling operations to obtain feature maps in three different spatial dimensions (height (H), width (W) and depth (D)), and then performs weighted fusion to achieve global context awareness of the input image block. Specifically, the pancreatic image to be segmented is input to the small kernel convolution module and the large kernel convolution module for feature extraction, and the feature map is obtained from the standard small kernel convolution (small kernel convolution module S). , obtain the feature map from the large kernel convolution (large kernel convolution module L) , ( represents the number of channels), for global features conduct The convolution operation on local features conduct After different convolution operations, complementary features are generated by element-by-element addition. .

[0035] To evaluate the complementary characteristics The importance weights of different spatial positions in the present invention are Adaptive average pooling is performed to generate three compressed feature maps of different spatial dimensions (H, W, and D). Figure 2 As shown, after these features are broadcasted, a comprehensive weight map is generated by adding them element by element. The specific operations are as follows: (1) in, represents the element-by-element addition operation, express The weighted sum after global average pooling in the H, W, and D dimensions is the same as Stay consistent. Indicates the height position of the voxel, Indicates the width position of the voxel, Indicates the depth position of the voxel.

[0036] To model the comprehensive weight graph The present invention adopts a recalibration mechanism to adjust the importance of each feature element to obtain the perceptual features of global context awareness. The recalibration mechanism is the comprehensive weight map Pass by , GELU activation function, And Sigmoid activation function output features, the output features and feature map Perform element-by-element multiplication to obtain the features .

[0037] The process can be formalized as: (2) in, Represents an element-by-element multiplication operation; Corresponding to (1x1x1) convolution, it is used to keep the size and number of channels of the input features unchanged; It is a local feature; Corresponding to the 3D convolution operation that doubles the number of feature channels, the number of channels is halved while keeping the input feature size unchanged; The corresponding 3D convolution operation halved the number of feature channels, while keeping the input feature size unchanged and halving the number of channels; and Represent the GELU activation function and the Sigmoid activation function respectively.

[0038] like Figure 3 As shown in the figure, the detail-sensitive feature enhancement branch DFEB, as a supplement to GCAB, uses the GELU nonlinear activation function to enhance the interaction of local details in the input image and improve the network's ability to capture small structures, edges and other subtle but important details.

[0039] In the detail-sensitive feature enhancement branch, feature maps are , feature map Perform the same operation as in GCAB to obtain complementary features . Complementary features After the GELU nonlinear activation function, , and the Sigmoid activation function outputs the corresponding features, and the output features are complementary features Perform element-by-element multiplication to obtain detail-sensitive features .

[0040] DFEB enhances the model’s focus on local features, ensuring that the network does not ignore subtle local differences while grasping the global context. The operation of DFEB can be expressed as: (3) in, For detailed features, For complementary features, and Respectively represent the GELU activation function and the Sigmoid activation function, Corresponding to (1x1x1) convolution, it is used to keep the size and number of channels of the input features unchanged.

[0041] In this embodiment, if Figure 4 As shown in the figure, the traditional contrastive learning method obtains universal feature representation by constructing positive and negative sample pairs. This method requires clear definition of positive and negative samples to make the learning objectives clearer. However, in practical applications, defining and constructing appropriate positive and negative sample pairs usually faces certain challenges. To this end, the present invention proposes a novel adaptive attention contrastive learning module (AACL module), which does not require explicit construction of positive and negative sample pairs and supports end-to-end training to optimize the pancreas segmentation task.

[0042] The AACL module introduces a new paradigm using two input image patches. and As the initial input of contrastive learning. Through convolutions with different kernel sizes and a simple feature contrastive learning process, the model generates four different feature maps. Specifically, the two input image blocks and After convolution processing with size 3 (small kernel convolution module S), we get and , two input image patches and After convolution with size 7 (large kernel convolution module L), we get and ; and , and After achieving intra-class aggregation between features of the same image block and inter-class separation between different image blocks on the two attention branches of GCAB and DFEB, comparative learning is performed to obtain four different feature maps.

[0043] These feature maps achieve intra-class aggregation between features of the same image block and inter-class separation between different image blocks on the two attention branches, the global context-aware branch (GCAB) and the detail-sensitive feature enhancement branch (DFEB). Specifically, in order to enhance the contrast and complementarity of the four features generated by the two attention branches GCAB and DFEB, the present invention designs a contrast loss between GCAB and DFEB, focusing on capturing the similarity of key features extracted by different convolution kernels throughout the learning process. The AACL loss function is defined as follows: (4) (5) (6) in, represents the loss function of the AACL method, represents the contrast loss, represents the number of layers of the network encoder, represents the loss weight of each layer encoder, Indicates the sequence number of the voxel point, represents the Euler number, and Respectively represent the context-aware features and detail-sensitive features mentioned above, so Indicated in The context-aware features in the layer encoder The value of a voxel, Indicated in The detail-sensitive features in the layer encoder The value of the voxel; Indicated in The context-aware features in the layer encoder belong to Channel No. The value of a voxel, Indicated in The detail-sensitive features in the layer encoder belong to Channel No. The value of the voxel; Represents the total number of elements, Indicates the total number of channels, represents the weight of contrast loss, Indicates that it is used to enhance and Sensory loss of the ability to distinguish between.

[0044] In order to better adapt to the needs of the model at different training stages, the present invention proposes an adaptive contrast learning strategy to further learn more accurate feature information. Specifically, the adaptive contrast learning strategy means that when the count threshold is not improved When the preset value is reached, Added to the training phase. Weight The smoothing reward factor and parameters (like Figure 4 The system can adjust itself through adaptive strategies (as shown in the flowchart in ). Calculated by the Exponential Moving Average (EMA) method: (7) (8) (9) Represents the final calculated reward factor. is the number of training iterations; is an integer indicating how many times the network iterates before performing a validation; is the smoothing coefficient in the EMA method; Indicated in The total network loss at iteration and The exponential moving average of . It is a combination of Dice loss (focusing on global structural consistency) and cross entropy loss (sensitive to the classification of each pixel), as shown in the following formula (13).

[0045] (10) in, is the total number of model training iterations.

[0046] like Figure 4 As shown in the flow chart, Remains at zero until reaches the preset value, then Participate in the optimization process of DKMAL. and is a uniformly distributed random number), randomly generate a The values ​​in the range are added to In. When and or hour, Added to .

[0047] In this embodiment, the present invention designs a boundary sensitive region optimization (BSRO strategy), which integrates the pancreatic boundary information to guide the network to pay more attention to balancing the attention of the pancreatic core area and the detail boundary. Considering the small volume of the pancreas in the three-dimensional abdominal CT image, the BSRO strategy aims to solve the inherent class imbalance problem in pancreatic segmentation, thereby improving the model's boundary sensitivity and segmentation accuracy for small targets. The present invention defines the BSRO loss function (Formula 11), this function not only focuses on the consistency of the global structure, but also has sensitivity to pixel-level classification.

[0048] (11) (12) in, express The weight coefficient is used to balance the and This makes The individual components of the model can be adjusted at a more fine-grained level, allowing the model to better capture each feature during training based on individual data; represents the Dice loss, represents the area loss, Represents the total number of model training iterations.

[0049] (13) in, Represents pixels Predicted to belong to the category The probability of Represents the label value (0 or 1), referring to the pixel Belongs to category ,and represents the total number of categories, Represents the total number of pixels in the image.

[0050] (14) (15) in, Indicates the distance in pixels between labels The nearest boundary pixel, Represents the value of the label map generated according to the label of the network input image, Indicates the category of the label value.

[0051] like Figure 6 As shown, Used to encapsulate boundary location information (BLI), which can enhance the model's ability to learn boundary details under the guidance of geometric prior knowledge.

[0052] Figure 5 and Figure 6 Visualization examples of labels and their corresponding BLIs are shown. Figure 5 Only black and white areas with clear regions are shown, and the two colors represent areas inside and outside the category; Figure 6 Shows a gradient color change with sharp boundaries, and elements outside the category area are set to 1 (black).

[0053] Data set introduction and experimental parameter settings: The method of the present invention was evaluated on the NIH pancreas dataset (NIHPancreas), which contains 82 non-contrast enhanced abdominal CT scans and their corresponding pancreas labels manually annotated by NIH radiologists using ITK-SNAP software. The size of each image is 512×512×[94-198], the slice thickness is usually between 1 mm and 2.5 mm, and the image provides high-resolution details. The present invention uses a five-fold cross-validation method to conduct experiments, dividing the dataset into five subsets, each containing 17, 17, 16, 16, and 16 samples, respectively. In each round of cross-validation, the present invention selects four subsets for model training and uses the remaining subset for model testing.

[0054] The DKMAL framework of the present invention is implemented based on the PyTorch framework and runs on a Linux system equipped with an NVIDIA GeForce RTX 3090 GPU with 24GB memory. The maximum number of training iterations is set to 20,000, and each iteration processes a batch of data. The present invention uses the Adam optimizer with an initial learning rate of 0.0001 to optimize the network parameters. In terms of data preprocessing, the present invention normalizes the image intensity values ​​from -100 to 240 to a range of 0.0 to 1.0, and the values ​​outside this range are clipped. The data set is balanced by random clipping to ensure that each clip contains a positive sample (foreground) and a negative sample (background), and the clipping volume is fixed to 96×96×96. In order to enhance the diversity of the data, the present invention introduces a random intensity change of up to 10% with a probability of 50%, and applies random affine transformations such as rotation and scaling. The present invention uses four evaluation indicators to evaluate the segmentation performance: Dice similarity coefficient (DSC), average symmetric surface distance (ASSD), positive predictive value (PPV) and sensitivity (SEN).

[0055] Test results: The DKMAL framework was experimented on the NIH pancreas dataset. DKMAL performed well in segmenting the pancreas with an average DSC of 88.03±2.60% and a minimum DSC of 74.91%, indicating that DKMAL is robust in handling the most challenging segmentation cases. In addition, the standard deviation of the average DSC of DKMAL across different pancreas cases is 2.60%, further demonstrating the stability of the method on the NIH dataset.

[0056] Figure 7 The following are three segmentation examples (Example 1, Example 2, and Example 3) of the pancreas region in the NIHPancreas dataset. The pancreatic region is marked with a solid arrow, and the segmented region is marked with a dotted arrow.

[0057] Embodiment 2 This embodiment discloses a pancreatic image segmentation system based on a co-adaptive learning framework, including: A data acquisition module is configured to: acquire a pancreas image to be segmented; The image segmentation module is configured as follows: the pancreatic image is input into a co-adaptive learning framework for segmentation to obtain a segmentation result; the co-adaptive learning framework includes a different kernel feature fusion module and an adaptive attention contrast learning module, the pancreatic image to be segmented is divided into a plurality of input image blocks, the plurality of input image blocks are input into the adaptive attention contrast learning module, a plurality of feature maps are obtained after convolution operations with different kernel sizes, the plurality of feature maps are input into a different kernel feature fusion module for inter-class aggregation between the same features and inter-class separation between different features, and then a plurality of classification features are output through a feature contrast learning process to obtain a segmentation result.

[0058] Embodiment 3 The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method of embodiment 1 when executing the program.

[0059] Embodiment 4 The purpose of this embodiment is to provide a computer-readable storage medium, a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the steps of the method of embodiment 1 are performed.

[0060] The steps involved in the apparatus of the above embodiments 3 and 4 correspond to the method embodiment 1, and the specific implementation method can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0061] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0062] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0063] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A pancreatic image segmentation method based on a co-adaptive learning framework, characterized in that: include: Acquire a pancreas image to be segmented; The pancreatic image is input into a co-adaptive learning framework for segmentation to obtain a segmentation result; the co-adaptive learning framework includes a different kernel feature fusion module and an adaptive attention contrast learning module, the pancreatic image to be segmented is divided into a plurality of input image blocks, the plurality of input image blocks are input into the adaptive attention contrast learning module, a plurality of feature maps are obtained after convolution operations with different kernel sizes, the plurality of feature maps are input into a different kernel feature fusion module for inter-class aggregation between the same features and inter-class separation between different features, and then a plurality of classification features are output through a feature contrast learning process to obtain a segmentation result.

2. The pancreatic image segmentation method based on the co-adaptive learning framework according to claim 1, characterized in that: The different kernel feature fusion module includes a parallel small kernel convolution module and a parallel large kernel convolution module, a parallel global context perception branch and a detail sensitive feature enhancement branch.

3. The pancreatic image segmentation method based on the co-adaptive learning framework as claimed in claim 2, characterized in that: The global context-aware branch performs different convolution operations on the output features of the small-core convolution module and the output features of the large-core convolution module, generates complementary features by element-by-element addition, performs adaptive average pooling on the complementary features to generate multiple compressed features, and then generates a comprehensive weight map by element-by-element addition.

4. The pancreatic image segmentation method based on the co-adaptive learning framework as claimed in claim 3, characterized in that: A recalibration mechanism is adopted to adjust the importance of each feature element in the comprehensive weight map to obtain global context-aware perceptual features.

5. The pancreatic image segmentation method based on the co-adaptive learning framework as claimed in claim 2, characterized in that: The complementary features in the detail-sensitive feature enhancement branch are sequentially subjected to the operations of the GELU nonlinear activation function, convolution, and Sigmoid activation function to obtain interactive features, and the interactive features are element-wise multiplied with the complementary features to obtain detail-sensitive features.

6. The pancreatic image segmentation method based on the co-adaptive learning framework according to claim 2, characterized in that: A contrast loss is designed between the global context-aware branch and the detail-sensitive feature enhancement branch, and the loss function is defined as follows: in, Represents contrast loss The weight of represents the contrast loss, Expressed as the loss weight of each layer encoder, Indicates sensory loss, represents the Euler number, Indicated in The perceptual features in the layer encoder The value of a voxel, Indicated in The first layer of detail-sensitive features in the encoder The value of a voxel, Indicated in The perceptual features in the layer encoder belong to Channel No. The value of a voxel, Indicated in The detail-sensitive features in the layer encoder belong to Channel No. The value of the voxel.

7. The pancreatic image segmentation method based on the co-adaptive learning framework according to claim 1, characterized in that: The co-adaptive learning framework uses a boundary-sensitive region optimization strategy and defines a boundary-sensitive region optimization loss function, which is expressed as: in, represents the weight coefficient, represents the Dice loss, represents the area loss, represents the number of model training iterations, Represents the total number of model training iterations.

8. A pancreatic image segmentation system based on a co-adaptive learning framework, characterized in that: include: A data acquisition module is configured to: acquire a pancreas image to be segmented; The image segmentation module is configured as follows: the pancreatic image is input into a co-adaptive learning framework for segmentation to obtain a segmentation result; the co-adaptive learning framework includes a different kernel feature fusion module and an adaptive attention contrast learning module, the pancreatic image to be segmented is divided into a plurality of input image blocks, the plurality of input image blocks are input into the adaptive attention contrast learning module, a plurality of feature maps are obtained after convolution operations with different kernel sizes, the plurality of feature maps are input into a different kernel feature fusion module for inter-class aggregation between the same features and inter-class separation between different features, and then a plurality of classification features are output through a feature contrast learning process to obtain a segmentation result.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the pancreatic image segmentation method based on the co-adaptive learning framework as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the pancreatic image segmentation method based on the co-adaptive learning framework as described in any one of claims 1 to 7 are implemented.

Images