Cerebral stroke focus segmentation method based on adaptive feature enhancement

By introducing an iterative joint attention module and a cross-level feature enhancement aggregation module in stroke lesions segmentation, adaptively extracting and fusion of feature information, the problems of insufficient sample size and loss of feature information are solved, and more accurate and clear lesion segmentation is achieved Effect.

CN119068191AActive Publication Date: 2024-12-03HANGZHOU DIANZI UNIV

Patent Information

Application Number
CN202411113964.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-12-03
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

The prior art has problems such as insufficient sample size, loss of feature information and poor segmentation in image segmentation of stroke lesion areas. Especially when dealing with small target areas, it is difficult to obtain multi-scale feature information, resulting in poor segmentation effect.

Method used

It is designed to design an iterative joint attention module and a cross-level feature enhancement aggregation module. By iterating attention residual blocks and information entropy metrics, feature information is adaptively extracted and fused, the semantic and spatial information of deep features is enhanced, and the absence of detailed information is reduced. And decoded to the image space through the decoder to obtain the segmentation result.

Benefits of technology

It improves the accuracy and real-time nature of stroke lesions, can better segment lesion areas of different sizes, significantly improving the clarity and accuracy of segmentation results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068191A_ABST
    Figure CN119068191A_ABST
Patent Text Reader

Abstract

The invention discloses a cerebral apoplexy focus segmentation method based on adaptive feature enhancement, and the method comprises the steps: firstly carrying out the data preprocessing of MRI magnetic resonance imaging data: carrying out the zooming of the data to a uniform spatial resolution through a linear interpolation method, and enabling the data to serve as an image to be segmented; secondly, inputting the to-be-segmented image into an encoder for multi-level feature extraction, wherein the encoder comprises a primary feature extraction module, three encoding modules and two cross-level feature enhanced clustering modules; and finally, an output feature map of an encoding module in the encoder and output feature maps of the two cross-level feature enhanced clustering modules are input into a decoder, a segmentation result map is output, and the decoder is formed by cascading three decoding modules and a segmentation head. The method is suitable for segmentation of lesion areas of different sizes, the feature channel which is more beneficial to the final segmentation result is selected, and the segmentation result is more accurate.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of stroke lesion segmentation based on deep learning, and in particular to a stroke lesion segmentation method based on adaptive feature enhancement. Background Art

[0002] Stroke is a common clinical disease with extremely serious consequences, characterized by high disability and mortality rates. Usually, clinical diagnosis of stroke requires doctors to observe the patient's brain by CT (Computerized Tomography) or MRI (Magnetic Resonance Imaging). A set of brain MRIs often has many slices, and it is necessary to rely on the experience of professional doctors to determine whether there are corresponding lesions. In addition, medical images will generate some noise during imaging, resulting in mottled, granular, textured or snowflake-like appearances. In response to the above problems, the use of automatic image segmentation technology can help improve the diagnostic efficiency and accuracy of doctors.

[0003] In recent years, with the development of deep learning technology, architectures such as U-Net (U-shaped Network), DeepLab3+ (encoder-decoder network based on dilated separable convolution semantic image segmentation), and CLCI-Net (Cross-level fusion and Context Inference Networks, an image segmentation network with cross-level fusion and context inference) have shown significant effects in improving the speed and accuracy of clinical medical image processing. However, in the image segmentation task of stroke lesion area, there are still some shortcomings and limitations, which bring some problems and challenges. First, the lesions caused by stroke usually show a relatively limited distribution in the data set. Although a set of brain MRI slices covers a large number of images, most of them are MRI images of healthy brains, resulting in a relatively small number of samples with lesion areas. Secondly, the pathological images of stroke patients show diversity and heterogeneity, that is, the size of the lesion area varies greatly. For small target areas, due to multiple downsampling, the existing models cannot obtain multi-scale feature information well, resulting in a large loss of feature information, and then cannot accurately segment small target areas, resulting in many traditional image segmentation methods performing poorly in distinguishing lesion areas from normal tissues. Summary of the invention

[0004] In view of the above technical problems, the present invention proposes a stroke lesion segmentation method based on adaptive feature enhancement. The method extracts multi-level features through an encoder, designs an iterative joint attention module and a cross-level feature enhancement aggregation module for adaptive and efficient fusion of features; the extracted features are further decoded into the image space by a decoder to obtain the segmentation result of the lesion. The present invention includes two newly designed modules: an iterative attention residual block and a cross-level feature enhancement aggregation module. For the iterative attention residual block, the module is the basic component unit of the encoder. Each iterative attention residual block includes 3 convolution residual blocks, 1 iterative joint attention module and 1 ReLu activation layer. Among them, the convolution residual block can effectively extract multi-level features in the image, and the iterative joint attention module performs adaptive fusion of features by iteratively combining global and local attention. Since a deeper encoder will reduce the resolution, resulting in the problem of loss of detail information, the iterative attention residual block uses the attention mechanism as a residual connection. Under the influence of the attention mechanism, more rich detail information is fused into deep features, and more spatial information is integrated into deep features based on semantic information, thereby reducing the loss of detail information.

[0005] For the cross-level feature enhancement aggregation module, this module uses information entropy to measure the amount of information of each channel feature, screens the shallow features of the jump connection, reduces the weight of a large amount of background information in the shallow features, reduces the impact of this information on the segmentation of the target area, constructs enhanced features, and solves the problem of low efficiency of feature reuse and fusion.

[0006] The present invention combines iterative joint attention and information entropy to make the deep features contain rich abstract semantic information while also containing more detailed spatial feature information. It uses information entropy to measure the strength of information in the features, so that the features can be efficiently fused, thereby adaptively extracting important features in the image, and effectively solving the problems existing in the stroke image segmentation task.

[0007] The specific steps of the method of the present invention include:

[0008] S1 performs data preprocessing on the MRI magnetic resonance imaging data: scaling it to a uniform spatial resolution through linear interpolation as the image to be segmented.

[0009] S2 inputs the image to be segmented into the encoder for multi-level feature extraction. The encoder includes a primary feature extraction module, three encoding modules and two cross-level feature enhancement and aggregation modules CFA. The specific steps include:

[0010] S21 inputs the image to be segmented into a primary feature extraction module for feature extraction. The primary feature extraction module is composed of a convolutional layer and a maximum pooling layer cascaded together to extract the primary feature F0 of the image.

[0011] S22 inputs the feature F0 extracted in step S21, denoted as X, to the first encoding module. Each encoding module is composed of a cascade of several groups of iterative attention residual blocks, each of which is composed of a cascade of three convolutional residual blocks, an iterative joint attention module and a ReLu activation layer. For the i-th encoding module, its input feature is F i-1 , whose output feature is F i .

[0012] The specific steps of the encoding module include:

[0013] S221 inputs the input feature X into the first convolution residual block, and adds the residuals of the features calculated by the convolution layer, batch normalization layer, and ReLu activation layer to the input features to obtain the output feature of the first convolution residual block

[0014] S222 Input to the second convolution residual block to get the output features Will Input to the third convolution residual block to get the output features Among them, the calculation operations of the second and third convolution residual blocks are the same as the calculation step S221 of the first convolution residual block.

[0015] S223 combines the input features of this group of iterative attention residual blocks, i.e., the input features X of step S22, with the output features of the third group of convolutional residual blocks It is also denoted as Y and input into the iterative joint attention module to obtain the output feature Z. The module includes the following specific steps:

[0016] S2231 adds the input features X and Y by broadcasting to obtain feature x, and inputs feature x into two branches of the first global-local joint attention module: the global attention branch and the local attention branch.

[0017] S2232 For the global attention branch, the input feature x first passes through the global average pooling layer f AP (), then enter the first point-by-point convolutional layer f PW1 (), and then through the batch normalization layer f BN () and ReLu activation layer f ReLu () and then enter the second point-by-point convolution layer f PW2 (), and finally through the batch normalization layer f BN () and Sigmoid activation layer f Sigmoid () Calculate the global attention A global .

[0018] S2233 For the local channel attention branch bag, the input feature x first enters the first point-by-point convolution layer f PW1 (), and then through the batch normalization layer f BN () and ReLu activation layer f ReLu () After calculation, enter the second point-by-point convolution layer f PW2 (), and finally through the batch normalization layer f BN () and Sigmoid activation layer f Sigmoid () Calculate the local attention A local .

[0019] S2234 will global attention A global and local attention A local Multiply the input features X and Y element by element respectively, and add the two product results to get the feature x′.

[0020] S2235 inputs the output feature x′ of step S2234 into the second global-local joint attention module, and repeats step S2232 to obtain the global attention A global′ , repeat step S2233 to get local attention A local′ , and perform weighted addition on the features X and Y input in step S223 respectively to obtain the output Z of the second global-local joint attention module.

[0021] S224 performs a ReLu activation operation on the feature Z to obtain the output features of this group of iterative attention residual blocks.

[0022] S225 repeats steps S221 to S224 several times, the number of repetitions is the number of groups of iterative attention residual blocks in the encoding module, and the output features of the last group of iterative attention residual blocks are the output features of the encoding module. For the i-th encoding module, the input features of its first group of iterative attention residual blocks are F i-1 , the input features of other groups of iterative attention residual blocks are the output features of the previous group of iterative attention residual blocks.

[0023] S23 inputs the output feature F1 of the first encoding module into the second encoding module, and calculates the output feature F2 of the second encoding module; inputs the output feature map F2 of the second encoding module into the third encoding module, and calculates the output feature map F3 of the third encoding module. The operation of the second and third encoding modules is the same as that of the first encoding module, that is, step S22.

[0024] S24 inputs the input feature F0 (shallow feature) and output feature F1 (deep feature) of the first encoding module into the first cross-level feature enhancement aggregation module (CFA) to obtain the enhanced aggregation feature The specific steps of this module are:

[0025] S241 calculates the channel information entropy E of shallow features i .

[0026] S242 For shallow features, each channel feature f i and channel information entropy E i Multiply them together to get the information entropy weighted features.

[0027] S243 adds the information entropy weighted features to the deep features to obtain enhanced aggregated features.

[0028] S25 inputs the output features F2 (shallow features) of the second encoding module and F3 (deep features) of the third encoding module into the second cross-level feature enhancement aggregation module to obtain the enhanced aggregation features. The calculation steps of the second cross-level feature enhancement aggregation module are the same as the calculation steps of the first cross-level feature enhancement aggregation module, namely steps S241 to S243.

[0029] S3 extracts the features f3, and Input to the decoder and output the segmentation result map. The decoder consists of three decoding modules and a segmentation head in cascade, where each decoding module is composed of a deconvolution layer, a normalization layer, and a Relu activation function layer in cascade.

[0030] Beneficial effects of the present invention: In order to solve the problems of poor real-time performance and insufficient feature extraction of the stroke lesion segmentation model, the present invention proposes a stroke lesion segmentation method based on adaptive feature enhancement. First, the iterative attention residual block in the encoder is used to extract features of the input stroke image, so that the deep features contain rich abstract semantic information and also contain rich spatial detail feature information. After feature extraction, a cross-level feature enhancement aggregation module is used for efficient adaptive feature fusion. The module measures the amount of information in the feature by calculating the channel information entropy, and assigns higher weights to channels with rich information and lower weights to channels with less information. The present invention has the following benefits: 1. An iterative joint attention module is designed. The iterative joint attention module can achieve channel attention at multiple scales by adjusting the size of the spatial pool, so that it is suitable for the segmentation of lesion areas of different sizes; 2. A cross-level feature enhancement aggregation module is designed. The features are weighted by information entropy to fuse features, thereby selecting feature channels that are more conducive to the final segmentation result, making the segmentation result more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Architecture diagram of the stroke lesion segmentation method based on adaptive feature enhancement;

[0032] Figure 2 Iterative joint attention module structure diagram;

[0033] Figure 3 Cross-level feature reinforcement aggregation and module structure diagram;

[0034] Figure 4 The effect diagram of the technology of the present invention on the segmentation of stroke lesions. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0036] As a preferred implementation form of the present invention, a stroke lesion segmentation method based on adaptive feature enhancement is provided, and its architecture is as follows: Figure 1 As shown, the following steps are included:

[0037] S1 scales the MRI magnetic resonance imaging data to a uniform 224×224 spatial resolution through linear interpolation as the image to be segmented.

[0038] S2 inputs the image to be segmented into the encoder for multi-level feature extraction. The encoder includes a primary feature extraction module, three encoding modules, and two cross-level feature enhancement aggregation modules (CFA). The specific steps include:

[0039] S21 inputs the image to be segmented into the encoder for multi-level feature extraction to extract the primary feature F0 of the image. The primary feature extraction module is composed of a convolutional layer and a maximum pooling layer cascaded together, where the convolution kernel size of the convolution layer is 7×7, the number of convolution kernels is 64, the step size is 2, and the padding is 3; the convolution kernel size of the maximum pooling layer is 3×3, the number of convolution kernels is 64, the step size is 2, and the padding is 1.

[0040] S22 For the primary feature F0, denoted as X, three encoding modules are used to further extract multi-level features. The first encoding module is composed of three groups of attention residual blocks in cascade, and the second and third encoding modules are composed of four and six groups of attention residual blocks in cascade, respectively. Each group of iterative attention residual blocks (steps S221 to S224) consists of three convolution residual blocks, one iterative joint attention module (whose structure is as follows Figure 2 As shown in Figure 2, the input feature of the i-th encoding module is F i-1 , whose output feature is F i .

[0041] The specific steps of the encoding module include:

[0042] S221 inputs feature X into the first convolution residual block, and adds the residual of the feature calculated by the 1×1 convolution layer, batch normalization layer, and ReLu activation layer to the input feature to obtain the output feature of the first convolution residual block

[0043] S222 Input to the second convolution residual block to get the output features Will Input to the third convolution residual block to get the output features Among them, the calculation steps of the second and third convolution residual blocks (where the convolution kernel sizes of the convolution layer are 3×3 and 1×1 respectively) are the same as the calculation step S221 of the first convolution residual block.

[0044] S223 combines the input features of this group of iterative attention residual blocks, i.e., the input features X of step S22, with the output features of the third group of convolutional residual blocks It is also denoted as Y and input into the iterative joint attention module, which includes the following specific steps:

[0045] S2231 adds the input features X and Y by broadcasting to obtain feature x, and inputs feature x into two branches of the first global-local joint attention module: the global attention branch and the local attention branch.

[0046] The S2232 global attention branch contains a global average pooling layer, two point-wise convolutional layers, and two activation function layers. First, the input feature x passes through the global average pooling layer f AP (), then enter the first point-by-point convolutional layer f PW1 (), and then through the batch normalization layer f BN () and ReLu activation layer f ReLu () and then enter the second point-by-point convolution layer f PW2 (), and finally through the batch normalization layer f BN () and Sigmoid activation layer f Sigmoid () Calculate the global attention A global , the specific calculation formula is:

[0047] A global =f Sigmoid (f BN (f PW2 (f ReLu (f BN (f PW1 (f AP (x)))))))

[0048] The global average pooling layer transforms feature x into a feature of size C×1×1, where C is the number of channels and f PW1() and f PW2 The convolution kernel sizes of () are (C / r)×C×1×1 and C×(C / r)×1×1 respectively. global The dimension is C×1×1.

[0049] The S2233 local channel attention branch contains two point-by-point convolutional layers and two activation function operations. First, input feature x into the first point-by-point convolutional layer f PW1 (), and then through the batch normalization layer f BN () and ReLu activation layer f ReLu () After calculation, enter the second point-by-point convolution layer f PW2 (), and finally through the batch normalization layer f BN () and Sigmoid activation layer f Sigmoid () Calculate the local attention A local , the specific calculation formula is:

[0050] A local =f Sigmoid (f BN (f PW2 (f ReLu (f BN (f PW1 (x))))))

[0051] f PW1 () and f PW2 The convolution kernel sizes of () are (C / r)×C×1×1 and C×(C / r)×1×1 respectively. local The dimensions are C×H×W.

[0052] S2234 will global attention A global and local attention A local Multiply the input features X and Y element by element, add the two product results, and get the feature x'. The calculation formula is:

[0053]

[0054] The above X and Y are obtained in step S223, where Represents element-wise multiplication.

[0055] S2235 inputs the output feature x′ of step S2234 into the second global-local joint attention module, and repeats step S2232 to obtain the global attention A global′ , repeat step S2233 to get local attention A local′ , and perform weighted addition on the feature maps X and Y input in step S223 to obtain the output Z of the second global-local joint attention module. The specific formula is:

[0056]

[0057] S224 performs a ReLu activation operation on the feature Z to obtain the output features of this group of iterative attention residual blocks.

[0058] S225 repeats steps S221 to S224 several times, the number of repetitions is the number of groups of iterative attention residual blocks in the encoding module (the first, second, and third encoding modules contain three, four, and six groups of iterative attention residual blocks, respectively), and the output features of the last group of iterative attention residual blocks are the output features of the encoding module. For the i-th encoding module, the input features of its first group of iterative attention residual blocks are F i-1 , the input features of other groups of iterative attention residual blocks are the output features of the previous group of iterative attention residual blocks.

[0059] S23 inputs the output feature F1 of the first encoding module into the second encoding module, and calculates the output feature F2 of the second encoding module; inputs the output feature map F2 of the second encoding module into the third encoding module, and calculates the output feature map F3 of the third encoding module. The steps of the second and third encoding modules are the same as those of the first encoding module, i.e., step S22.

[0060] S24 inputs the input feature F0 (shallow feature) and output feature F1 (deep feature) of the first encoding module into the first cross-level feature reinforcement aggregation module (CFA, whose structure is as follows: Figure 3 As shown), the characteristics after enhanced polymerization are obtained The specific steps of this module are:

[0061] S241 calculates the channel information entropy of the shallow features, and the specific steps include:

[0062] S2411 For each channel feature f i , extracting bigrams (C n ,A n ), where C n represents the eigenvalue of the center point of the nth 3×3 sliding window, A n Represents the average of the remaining eigenvalues ​​in this window.

[0063] S2412 Statistics of bigrams (C n ,A n ) appears N(C n ,A n ), calculate the pair (C n ,A n ) n , and its calculation formula is:

[0064] Pn =N(C n ,A n ) / (H×W)

[0065] Where H is the height of the feature and W is the width of the feature.

[0066] S2413 calculates channel feature f i Information entropy E i , and its calculation formula is:

[0067]

[0068] S242 For shallow features, each channel feature f i and channel information entropy E i By multiplying them together, we get the information entropy-weighted features, thereby focusing on channels with larger information entropy and obtaining richer information representation.

[0069] S243 adds the information entropy weighted features to the deep features to obtain enhanced aggregated features. The features obtained in this way have spatial information of shallow features and contain semantic information of deep features.

[0070] S25 inputs the output features F2 (shallow features) of the second encoding module and F3 (deep features) of the third encoding module into the second cross-level feature enhancement aggregation module to obtain the enhanced aggregation features. The calculation steps of the second cross-level feature enhancement aggregation module are the same as the calculation steps of the first cross-level feature enhancement aggregation module, namely steps S241 to S243.

[0071] S3 decodes the features extracted by the encoder into the image space to obtain the lesion segmentation result. Specifically, the features F3, and The decoder is composed of three decoding modules and a segmentation head in cascade, where each decoding module is composed of a deconvolution layer, a normalization layer, and a Relu activation function layer in cascade. The specific steps of the decoder include:

[0072] S31 inputs the feature F3 into the decoding module 3, and calculates the decoded feature f3 through the deconvolution layer, the normalization layer and the Relu activation function layer.

[0073] S32 will feature After being cascaded with feature f3 in the channel dimension, it is input into the decoding module 2, and the decoded feature f2 is calculated through the deconvolution layer, normalization layer and Relu activation function layer.

[0074] S33 will feature After being cascaded with feature f2 in the channel dimension, it is input into the decoding module 1, and the decoded feature f1 is calculated through the deconvolution layer, normalization layer and Relu activation function layer.

[0075] S34 inputs the feature f1 into the segmentation head, and obtains the segmented image through the convolution layer and the Sigmoid activation layer. The effect is as follows Figure 4 shown.

[0076] Example:

[0077] The steps of this embodiment are the same as those of the specific implementation method, and will not be described in detail here. The following is a demonstration of some implementation processes and implementation results.

[0078] The technology of the present invention is implemented in the dataset ISLES2022, which contains 250 case images, including a large number of images of non-lesion areas. In order to reduce the impact of too many negative samples on model training, the slices in the middle part were selected when splitting the image, and finally 15684 pairs of images were obtained. Among them, 12547 images were used as training sets and 3137 images were used as test sets. The above image data were cropped to a uniform size of 224×224. This model uses a batch size of 8 and a weight decay of 10 -8 The Adam (Adaptive Moment Estimation) optimizer is used for training, and the learning rate is set to 0.0001. In order to verify the effectiveness of the technology of the present invention, Figure 4 The segmentation effect diagram of the technology of the present invention after training on the data set ISLES2022 is displayed, and compared with the following methods respectively: U-Net (U-shaped Network): an image segmentation network based on convolutional neural network, CA-Net (Comprehensive Attention Network): an image segmentation network based on comprehensive attention mechanism, DeepLab3+ (DeepLab Version 3Plus): an encoder-decoder network based on void separable convolution semantic image segmentation, TransUNet (Transformers+U-Net): a medical image segmentation network based on Transformer and convolutional neural network, CLCI-Net (Cross-level fusion andContext Inference Networks): an image segmentation network with cross-level fusion and contextual reasoning, VA-TransUNet (Visual Attention with Transformers+U-Net): an image segmentation network using multi-scale visual attention. Figure 4As shown in the figure, compared with U-Net, CLCI-Net and other methods, the technology of the present invention can accurately segment the lesions, especially the lesion areas with smaller areas, which can also be segmented by the technology of the present invention. Compared with other methods, the segmented lesion areas have clearer contours and are more similar in shape to the labels.

[0079] In order to verify the effectiveness of the technology of the present invention, Table 1 uses the indicators Dice (measure the overlap between two sets), Recall (recall rate, measure the proportion of samples correctly predicted by the model as positive examples to all actual positive examples), VOE (Volume Overlap Error, measure the error of the volume overlap between the segmentation result and the true segmentation), RVD (Relative Volume Difference, measure the volume difference between the segmentation result and the true segmentation), and IoU (Intersection over Union, measure the overlap between the predicted result and the true annotation) to compare the technology of the present invention with the existing technology. As shown in Table 1, the technology of the present invention has achieved better results in many indicators, thus proving the effectiveness of the technology of the present invention.

[0080] Table 1 Quantitative comparison results between the invention technology and the prior art

[0081]

Claims

1. A stroke lesion segmentation method based on adaptive feature enhancement, characterized in that: The specific steps include: S1. Preprocess the MRI data: scale the data to a uniform spatial resolution using a linear interpolation method as the image to be segmented; S2. Input the image to be segmented into the encoder for multi-level feature extraction, the encoder includes a primary feature extraction module, three encoding modules and two cross-level feature enhancement and aggregation modules CFA; S3. Input the output feature map of the encoding module in the encoder and the output feature map of the two cross-level feature enhancement aggregation modules CFA into the decoder, and output the segmentation result map. The decoder is composed of three decoding modules and a segmentation head in cascade.

2. The method for stroke lesion segmentation based on adaptive feature enhancement according to claim 1, characterized in that: The specific implementation process of the encoder is as follows: S21. Input the image to be segmented into the primary feature extraction module for feature extraction. The primary feature extraction module is composed of a convolution layer and a maximum pooling layer cascaded to extract the primary feature F0 of the image; S22. Input feature F0, denoted as X, to the first encoding module; each encoding module is composed of a cascade of several groups of iterative attention residual blocks, each group of iterative attention residual blocks is composed of a cascade of three convolutional residual blocks, an iterative joint attention module and a ReLu activation layer; for the i-th encoding module, its input feature is F i-1 , whose output feature is F i ; S23. Input the output feature F1 of the first encoding module to the second encoding module, and calculate the output feature F2 of the second encoding module; input the output feature map F2 of the second encoding module to the third encoding module, and calculate the output feature map F3 of the third encoding module. The operations of the second and third encoding modules are the same as those of the first encoding module; S24. Input the input feature F0 and output feature F1 of the first encoding module to the first cross-level feature enhancement aggregation module CFA to obtain the enhanced aggregation feature S25. Input the output feature F2 of the second encoding module and the output feature F3 of the third encoding module into the second cross-level feature enhancement aggregation module to obtain the enhanced aggregation feature The calculation steps of the second cross-level feature enhancement aggregation module are the same as those of the first cross-level feature enhancement aggregation module.

3. The method for stroke lesion segmentation based on adaptive feature enhancement according to claim 2, characterized in that: The specific implementation process of the encoding module is as follows: S221. Input the input feature X to the first convolution residual block, and add the residuals of the features calculated by the convolution layer, batch normalization layer, and ReLu activation layer to the input features to obtain the output feature of the first convolution residual block S222. Input to the second convolution residual block to get the output features Will Input to the third convolution residual block to get the output features The computational operations of the second and third convolutional residual blocks are the same as the first convolutional residual block; S223. Combine the input features of this group of iterative attention residual blocks, i.e., feature X, with the output features of the third group of convolutional residual blocks Also denoted as Y, it is input into the iterative joint attention module to obtain the output feature Z; S224. Perform ReLu activation on feature Z to obtain the output features of this group of iterative attention residual blocks; S225. Repeat steps S221 to S224, the number of repetitions is the number of groups of iterative attention residual blocks in the encoding module, and the output features of the last group of iterative attention residual blocks are the output features of the encoding module; for the i-th encoding module, the input features of its first group of iterative attention residual blocks are F i-1 , the input features of other groups of iterative attention residual blocks are the output features of the previous group of iterative attention residual blocks.

4. The method for stroke lesion segmentation based on adaptive feature enhancement according to claim 3, characterized in that: The implementation process of the iterative joint attention module is as follows: S2231. Add the input features X and Y by broadcasting to obtain feature x, and input feature x into two branches of the first global-local joint attention module: the global attention branch and the local attention branch; S2232. For the global attention branch, the input feature x first passes through the global average pooling layer f AP (), then enter the first point-by-point convolutional layer f PW1 (), and then through the batch normalization layer f BN () and ReLu activation layer f ReLu () and then enter the second point-by-point convolution layer f PW2 (), and finally through the batch normalization layer f BN () and Sigmoid activation layer f Sigmoid () Calculate the global attention A global ; S2233. For the local channel attention branch bag, the input feature x first enters the first point-by-point convolution layer f PW1 (), and then through the batch normalization layer f BN () and ReLu activation layer f ReLu () After calculation, enter the second point-by-point convolution layer f PW2 (), and finally through the batch normalization layer f BN () and Sigmoid activation layer f Sigmoid () Calculate the local attention A local ; S2234. Global attention A global and local attention A local Multiply the input features X and Y element by element, add the two product results, and get the feature x′; S2235 inputs feature x′ into the second global-local joint attention module and repeats step S2232 to obtain global attention A global′ , repeat step S2233 to get local attention A local′ , and perform weighted addition on features X and Y respectively to obtain the output feature Z of the second global-local joint attention module.

5. The method for stroke lesion segmentation based on adaptive feature enhancement according to claim 4, characterized in that: The specific implementation process of the cross-level feature enhancement aggregation module CFA is as follows: S241. Calculate the channel information entropy E of feature F0 i ; S242. For feature F0, each channel feature f i and channel information entropy E i Multiply them together to get the information entropy weighted features; S243. Add the information entropy weighted features to feature F1 to obtain enhanced aggregated features.

6. The method for stroke lesion segmentation based on adaptive feature enhancement according to claim 5, characterized in that: The specific calculation process of the channel information entropy of the feature is as follows: S2411. For each channel feature f i , extracting bigrams (C n ,A n ), C n represents the eigenvalue of the center point of the nth 3×3 sliding window, A n represents the average value of the remaining eigenvalues ​​in the window; S2412. Calculate the pair (C n ,A n ) n , and its calculation formula is: P n =N(C n ,A n ) / (H×W) Among them, N(C n ,A n ) is a binary group (C n ,A n ) appears, H is the height of the feature, and W is the width of the feature; S2413. Calculate channel feature f i Information entropy E i , and its calculation formula is:

7. The method for stroke lesion segmentation based on adaptive feature enhancement according to claim 6, characterized in that: The specific implementation process of the decoder is as follows: S31 inputs the feature F3 into the decoding module 3, and obtains the decoded feature f3 through the deconvolution layer, the normalization layer and the Relu activation function layer; S32.Characteristics After being cascaded with feature f3 in the channel dimension, it is input into decoding module 2, and the decoded feature f2 is calculated through the deconvolution layer, normalization layer and Relu activation function layer; S33.Characteristics After being cascaded with feature f2 in the channel dimension, it is input into the decoding module 1, and the decoded feature f1 is calculated through the deconvolution layer, normalization layer and Relu activation function layer; S34. Input feature f1 into the segmentation head, and obtain the segmented image through the convolution layer and the Sigmoid activation layer.

Citation Information

Patent Citations

  • Medical image segmentation method fusing multi-scale residual attention

    CN116563204A

  • Cerebral stroke image segmentation method based on multi-attention and dilated convolution

    CN117876689A

  • Medical image segmentation method based on multiple attention

    CN118071774A

  • KR20230166464A

Cited By

  • Tissue perception magnetic resonance image enhancement method, electronic equipment and storage medium

    CN122530010A