Ultrasonic image segmentation method and system based on multi-level feature extraction

Abstract

The present invention discloses a method and system for ultrasonic image segmentation based on multi-level feature extraction. The present invention comprises using an ultrasonic image segmentation network model to perform ultrasonic image segmentation on an input ultrasonic image. The ultrasonic image segmentation network model comprises a dual-branch encoder, a multi-scale feature fusion module, a multi-scale feature attention module, and a decoder. The dual-branch encoder is used to extract small detail features, large-scale context features, and global features at different scales; the multi-scale feature fusion module fuses features and inputs them into the decoder; the multi-scale feature attention module enhances the features of each layer of the CNN encoder and inputs them into the decoder. After decoding by the decoder, a segmentation prediction map is obtained. The present invention aims to solve the problem that most existing ultrasonic image segmentation methods based on deep learning only focus on capturing local details and global correlations while ignoring large-scale context information, thereby improving the information capture efficiency of the input image and enhancing the segmentation performance of the network.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasound (US) image processing, and in particular to an ultrasound image segmentation method and system based on multi-level feature extraction. Background Art

[0002] Ultrasound imaging is a commonly used method for clinical diagnosis. It can observe the internal structure and location of organs and nerves, as well as their contractions and movements. Due to its real-time nature, ease of use, and harmlessness, ultrasound imaging is more widely accepted than other medical imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). However, ultrasound image analysis and lesion annotation rely heavily on physician experience, resulting in low efficiency. Ultrasound image segmentation plays a vital role in early diagnosis and subsequent treatment of diseases. It improves clinical diagnostic efficiency and reduces physician workload. Traditional segmentation methods such as region growing and graph cuts can achieve automatic segmentation of objects, but these methods suffer from suboptimal segmentation accuracy and limited generalization. In recent years, the development of deep learning has achieved significant success in the field of computer vision. Compared to traditional segmentation methods, deep learning-based image segmentation methods offer superior performance. Commonly used image segmentation methods are based on deep learning models, such as UNet, TransUNet, and ConvNext. While these methods offer significantly improved segmentation performance compared to traditional methods, most focus solely on extracting local and global features, neglecting the large-scale contextual features necessary for large objects. Furthermore, the use of large kernel convolutional networks results in excessively large model parameters, resulting in suboptimal training results. Image segmentation methods based on deep learning models continue to require continuous improvement and innovation to enhance network feature extraction capabilities and improve segmentation performance. Summary of the Invention

[0003] Technical problem to be solved by the present invention: In response to the above-mentioned problems in the prior art, a method and system for ultrasound image segmentation based on multi-level feature extraction are provided. The present invention aims to solve the problem that most existing ultrasound image segmentation methods based on deep learning only focus on capturing local details and global correlations while ignoring large-scale contextual information, thereby improving the information capture efficiency of the input image and enhancing the segmentation performance of the network.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for ultrasonic image segmentation based on multi-level feature extraction includes using an ultrasonic image segmentation network model to perform ultrasonic image segmentation on an input ultrasonic image to obtain a segmentation prediction map, wherein the ultrasonic image segmentation network model includes a dual-branch encoder, a multi-scale feature fusion module, a multi-scale feature attention module and a decoder, and the dual-branch encoder includes a CNN encoder and a Transformer encoder. The method for performing ultrasonic image segmentation on the input ultrasonic image to obtain a segmentation prediction map includes: extracting small detail features and large-scale context features of different scales from the input ultrasonic image through the CNN encoder and adaptively selecting and enhancing the small detail features and large-scale context features of different scales using adaptive weighted weights; extracting global features from the input ultrasonic image through the Transformer encoder; using the multi-scale feature attention module to enhance the shallow features output by each layer of the CNN encoder and outputting them to the decoder, and decoding the fused multi-level features and the shallow features enhanced by the multi-scale feature attention module layer by layer through the decoder to obtain the final segmentation prediction map.

[0006] Optionally, the CNN encoder includes N scale encoding branches, the encoding branch of the first scale includes a 1×1 convolution layer and a large and small kernel attention module, and the encoding branches of the remaining N-1 scales all include a downsampling module, a channel dimension splicing module and a large and small kernel attention module. The output of the 1×1 convolution layer in the encoding branch of the first scale serves as the input of the downsampling module in the encoding branch of the second scale, and the input of the remaining downsampling modules is the output of the channel dimension splicing module in the encoding branch of the previous scale. The output of all channel dimension splicing modules also serves as the input of the large and small kernel attention modules in the encoding branch of the same scale. The first N-1 The output of the large and small kernel attention modules in the encoding branch of a scale is used as one input of the channel dimension splicing module in the encoding branch of the next scale. The output of the large and small kernel attention modules in the encoding branch of the Nth scale is used as the input of the multi-scale feature fusion module CNN encoder end. The decoder includes N-1 levels of decoding modules connected in cascade. The multi-scale feature fusion module serves as the input of the first-level decoding module. The other input of each level of decoding module is the output of the large and small kernel attention modules in the corresponding scale encoding branch, and the output feature is extracted by the multi-scale attention module. The final segmentation prediction map is decoded and output by the N-1th level decoding module.

[0007] Optionally, the processing of the input feature map by the large and small core attention module includes: extracting large-scale context features from the input feature map of the large and small core attention module through a large-scale feature selection module, and extracting small detail features from the input feature map of the large and small core attention module through a small-core convolution module, and then adding the large-scale context features and the small detail features element by element; then passing the features obtained by element-by-element addition through a feature interaction module to obtain fully integrated small details and large-scale context output data, and the feature interaction module performs maximum pooling and average pooling on the features obtained by element-by-element addition and then splices them in the channel dimension, and then passes through a convolution layer with a convolution kernel size of 7×7 and a Sigmoid activation function, and then performs matrix multiplication operation on the features obtained by element-by-element addition, and finally adds the features obtained by the matrix multiplication operation to the small detail features element by element, and then passes through a 3×3 convolution layer, a 3×3 void convolution layer with a void rate d of 2, and a 1×1 convolution layer to obtain fully integrated small details and large-scale context output data.

[0008] Optionally, the large-scale feature selection module extracts large-scale context features, including: extracting preliminary features from the input feature map of the large-scale feature selection module through a 1×1 convolution layer and a 3×3 convolution layer; respectively passing the preliminary features through large-core decomposition modules with different receptive fields to obtain two large-scale context features, adding the two large-scale context features element by element, and then performing feature adaptive enhancement through a feature adaptive enhancement module, wherein the feature adaptive enhancement module includes a global average pooling, two linear layers, and a Softmax function connected in sequence, wherein the two linear layers are used for dimensionality reduction and dimensionality increase, respectively; then, the two large-scale context features obtained by the large-core decomposition module are matrix-multiplied with the weights output by the Softmax function, and then added element by element to obtain the final large-scale context features; the small-core convolution module consists of a 1×1 convolution layer, a 5 × 5 convolutional layers are connected in series; the said passing the preliminary features through large kernel decomposition modules with different receptive fields respectively means that the preliminary features are respectively passed through the first large kernel decomposition module and the second large kernel decomposition module with different receptive fields for large-scale context feature extraction, and the first large kernel decomposition module and the second large kernel decomposition module each include a 5×5 depthwise separable convolution layer, a 5×5 void convolution module, a 7×7 void convolution module, a 3×3 void convolution module and a 1×1 convolution layer connected in sequence, wherein the number of groups of the void group convolution is 8, the void rate d of the 5×5 void convolution module, the 7×7 void convolution module and the 3×3 void convolution module in the first large kernel decomposition module is 1, and the void rate d of the 5×5 void convolution module, the 7×7 void convolution module and the 3×3 void convolution module in the second large kernel decomposition module is 3.

[0009] Optionally, the multi-scale feature fusion module fuses the multi-level features extracted by the CNN encoder and the Transformer encoder, including: splicing the multi-level features extracted by the CNN encoder and the Transformer encoder through the channel dimension, and then passing the channel dimension from 2048 dimensions to 1024-dimensional features through a 1×1 convolution layer, and then extracting features from the 1024-dimensional features through three branches of a 1×1 convolution layer, a 3×3 convolution layer, and a 3×3 void convolution layer with a void rate d of 2, and then splicing the features extracted by the three branches with the original 1024-dimensional features in the channel dimension, and then passing the channel dimension through a 1×1 convolution layer to reduce the channel dimension to 1024 dimensions to obtain the fused multi-level features.

[0010] Optionally, the multi-scale attention module extracts output features including: extracting features from the input features of the multi-scale attention module through a first branch consisting of a 1×1 convolution layer and a 3×3 convolution layer, and extracting features through a second branch consisting of a 1×1 convolution layer and a 3×3 dilated convolution layer with a dilation rate d of 2, concatenating the channel dimensions of the features extracted by the two branches, and then reducing the channel dimensions through a 1×1 convolution layer, and then performing maximum pooling and average pooling on the features after reducing the channel dimensions, and then concatenating the channel dimensions, and then performing matrix multiplication on the features after reducing the channel dimensions after dimensionality reduction by a linear layer, dimensionality increase by a linear layer, and activation by a Sigmoid activation function to obtain output features.

[0011] Optionally, the Transformer encoder is composed of a patch embedding layer, a linear layer and multiple transformer layers connected in sequence, wherein the image block is embedded as a convolution layer with a convolution kernel size of 16×16 and a stride of 16. The ultrasound image with a dimension of H×W×3 is passed through the image block embedding layer to obtain a feature with a dimension of H / 16×W / 16×D, where H and W are the height and width of the ultrasound image, and D is the output data dimension. The transformer layer is composed of an embedded image block layer (embedded Patches), two layers of normalization layers, a multi-head attention layer, a multi-layer perceptron layer and two residual connections. The input feature data passes through the embedded image block layer, the first layer of normalization layer, and the multi-head attention layer, and then is added element-by-element with the feature data output by the embedded image block layer through the first residual connection. The obtained feature data then passes through the second layer of normalization layer and the multi-layer perceptron layer, and then is added element-by-element with the first residual connection to obtain the feature data as the output data of the transformer layer; the decoding module consists of a transposed convolution with a convolution kernel size of 2×2 and two 3×3 convolution layers connected in sequence.

[0012] In addition, the present invention also provides an ultrasound image segmentation system based on multi-level feature extraction, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction.

[0013] In addition, the present invention also provides a computer-readable storage medium, which stores a computer program or instruction. The computer program or instruction is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction through a processor.

[0014] In addition, the present invention also provides a computer program product, including a computer program or instructions, which is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction through a processor.

[0015] Compared with the prior art, the present invention mainly has the following advantages:

[0016] 1. The multi-level feature extraction design of the ultrasound image segmentation network model of the present invention realizes the extraction of small detail features, large-scale context features to global features, improves the information capture efficiency of the input image, and can enhance the ultrasound image segmentation performance of the ultrasound image segmentation network model.

[0017] 2. The large and small kernel attention modules of this invention can capture features of objects of different sizes in the input image. Small kernel convolution layers are used to extract features such as edge details, while decomposed large kernel modules are used to more accurately capture large-scale contextual features of large-sized objects. The feature interaction module makes a significant contribution to improving the ability to fuse and express features of different scales within the network and preventing overfitting.

[0018] 3. The large-scale feature selection module of the present invention uses convolutional layers of multiple scales to extract features of different scales and adaptively selects and enhances important scales. This improves the network's feature extraction capabilities and the effective expression of features of different scales in the image, obtaining richer and more accurate feature data.

[0019] 4. The large kernel decomposition of the present invention can reduce the number of parameters of the ultrasound image segmentation network model and improve the trainability of the ultrasound image segmentation network model while fully extracting large-scale context features through an effective decomposition method. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the basic process of the method of the embodiment of the present invention.

[0021] Figure 2 Schematic diagram of the network structure of the ultrasound image segmentation network model in an embodiment of the present invention.

[0022] Figure 3 Schematic diagram of the network structure of the large and small core attention modules in an embodiment of the present invention.

[0023] Figure 4 Schematic diagram of the network structure of the large-scale feature selection module in an embodiment of the present invention.

[0024] Figure 5 Schematic diagrams of the network structures of the first and second large core decomposition modules in an embodiment of the present invention, wherein (a) is a schematic diagram of the network structure of the first large core decomposition module, and (b) is a schematic diagram of the network structure of the second large core decomposition module.

[0025] Figure 6 Schematic diagram of the network structure of the multi-scale feature fusion module in an embodiment of the present invention.

[0026] Figure 7 Schematic diagram of the network structure of the multi-scale attention module in an embodiment of the present invention.

[0027] Figure 8 Schematic diagram of the network structure of the transformer layer in the transformer encoder in an embodiment of the present invention.

[0028] Figure 9 The first part of the segmentation results of BUSI and BUSC ultrasound images using different segmentation methods in the embodiment of the present invention, where (a-1) to (e-1) are the label images of the first sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method; (a-2) to (e-2) are the label images of the second sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method; (a-3) to (e-3) are the label images of the third sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method. tion-UNet method; (a-4) to (e-4) are the label images of the 4th sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method; (a-5) to (e-5) are the label images of the 5th sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method; (a-6) to (e-6) are the label images of the 6th sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method.

[0029] Figure 10 The second part of the segmentation results of BUSI and BUSC ultrasound images using different segmentation methods in the embodiment of the present invention, where (a-1) to (e-1) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method and MDA-Net method for the first sample respectively; (a-2) to (e-2) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method and MDA-Net method for the second sample respectively; (a-3) to (e-3) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method and MDA-Net method for the third sample respectively. The segmentation prediction maps obtained by the SKUNet method and the MDA-Net method; (a-4) to (e-4) are the segmentation prediction maps obtained by the SKUNet method, TransUNet method, CMU-Net method, CMUNext method and MDA-Net method for the 4th sample respectively; (a-5) to (e-5) are the segmentation prediction maps obtained by the SKUNet method, TransUNet method, CMU-Net method, CMUNext method and MDA-Net method for the 5th sample respectively; (a-6) to (e-6) are the segmentation prediction maps obtained by the SKUNet method, TransUNet method, CMU-Net method, CMUNext method and MDA-Net method respectively.

[0030] Figure 11 These are the first part segmentation results of UNS ultrasound images using different segmentation methods in the embodiments of the present invention, where (a-1) to (e-1) are the label images of the first sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method, respectively; (a-2) to (e-2) are the label images of the second sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method, respectively; (a-3) to (e-3) are the label images of the third sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method, respectively; (a-4) to (e-4) are the label images of the fourth sample and the segmentation prediction maps obtained using the method of this embodiment, the UNet method, the UNet++ method, and the Attention-UNet method, respectively.

[0031] Figure 12These are the second part segmentation results of the UNS neural ultrasound image using different segmentation methods in the embodiments of the present invention, where (a-1) to (e-1) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method, and MDA-Net method for the first sample, respectively; (a-2) to (e-2) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method, and MDA-Net method for the second sample, respectively; (a-3) to (e-3) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method, and MDA-Net method for the third sample, respectively; (a-4) to (e-4) are the segmentation prediction maps obtained by using the SKUNet method, TransUNet method, CMU-Net method, CMUNext method, and MDA-Net method for the fourth sample, respectively. DETAILED DESCRIPTION

[0032] The present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0033] like Figure 1 As shown, the ultrasonic image segmentation method based on multi-level feature extraction in this embodiment includes using an ultrasonic image segmentation network model to perform ultrasonic image segmentation on the input ultrasonic image to obtain a segmentation prediction map, the ultrasonic image segmentation network model includes a dual-branch encoder, a multi-scale feature fusion module, a multi-scale feature attention module and a decoder, the dual-branch encoder includes a CNN encoder and a Transformer encoder, and the use of the ultrasonic image segmentation network to perform ultrasonic image segmentation on the input ultrasonic image to obtain a segmentation prediction map includes: extracting small detail features and large-scale context features of different scales from the input ultrasonic image through the CNN encoder and adaptively selecting and enhancing the small detail features and large-scale context features of different scales using adaptive weighted weights; extracting global features from the input ultrasonic image through the Transformer encoder; fusing the multi-level features extracted by the CNN encoder and the Transformer encoder through the multi-scale feature fusion module; using the multi-scale feature attention module to enhance the shallow features output by each layer of the CNN encoder and output them to the decoder, and decoding the fused multi-level features and the shallow features enhanced by the multi-scale feature attention module layer by layer through the decoder to obtain the final segmentation prediction map.

[0034] like Figure 2As shown in Figure 1, the CNN encoder includes N scale encoding branches. The encoding branch of the first scale includes a 1×1 convolution layer and a large and small kernel attention module. The encoding branches of the remaining N-1 scales all include a downsampling module, a channel dimension splicing module and a large and small kernel attention module. The output of the 1×1 convolution layer in the encoding branch of the first scale serves as the input of the downsampling module in the encoding branch of the second scale. The input of the remaining downsampling modules is the output of the channel dimension splicing module in the encoding branch of the previous scale. The output of all channel dimension splicing modules also serves as the input of the large and small kernel attention modules in the encoding branch of the same scale. The output of the large and small kernel attention modules in the encoding branch of the Nth scale is used as one input to the channel dimension splicing module in the encoding branch of the next scale. The output of the large and small kernel attention modules in the encoding branch of the Nth scale is used as the input to the multi-scale feature fusion module CNN encoder. The decoder includes N-1 levels of decoding modules connected in cascade. The multi-scale feature fusion module is used as the input to the first level decoding module. The other input of each level of decoding module is the output of the large and small kernel attention modules in the encoding branch of the corresponding scale. The output features are extracted by the multi-scale attention module, and the final segmentation prediction map is decoded and output by the N-1th level decoding module. Figure 2 In this embodiment, N=5. Figure 2 Among the five scale coding branches in the middle, from top to bottom are the coding branch of the first scale, the coding branch of the second scale, the coding branch of the third scale, the coding branch of the fourth scale, and the coding branch of the fifth scale. Figure 2 In the four-level decoding module on the right side, from bottom to top, there are the first-level decoding module, the second-level decoding module, the third-level decoding module, and the fourth-level decoding module. The method of this embodiment uses the CNN encoder to extract small detail features and large-scale context features and adaptively selects and enhances features of different scales. The encoding branch constitutes the multi-scale input of the CNN encoder-side features, and the Transformer encoder is used to extract global features; the multi-scale feature fusion module is used to perform information interaction between the multi-level features extracted by the dual encoder; the multi-scale feature attention module embedded in the jump connection is used to enhance the important features of the shallow layer output of each layer of the CNN encoder to promote the subsequent deep and shallow feature fusion; the decoder decodes the fused multi-level features layer by layer to output the final segmentation prediction map; the small details and large-scale context features extracted by the large and small kernel attention modules of the CNN encoder, and the global features extracted by the Transformer encoder are subjected to information interaction by the multi-scale feature fusion module, and the output features are passed through the decoder composed of convolutional layers. The deep features of the decoder are combined with the enhanced shallow features of the CNN encoder, and the final segmentation prediction map is obtained through layer-by-layer decoding.

[0035] The large and small kernel attention module uses the small kernel convolution module and the large-scale feature selection module to extract small detail features and large-scale context features, and adaptively selects and enhances large-scale context features of appropriate scales, and performs information interaction between features to obtain local small detail features and large-scale context features after feature fusion. The large and small kernel attention module includes the small kernel convolution module, the large-scale feature selection module and the feature interaction module, wherein the small kernel convolution module and the large-scale feature selection module are dual-branch parallel structures, and the features of the two branches are added and fused, and then the feature interaction module is used to realize information interaction between features of different scales. Figure 3 As shown, the processing of the input feature map by the large and small core attention module includes: extracting large-scale context features from the input feature map of the large and small core attention module through a large-scale feature selection module, and extracting small detail features from the input feature map of the large and small core attention module through a small-core convolution module, and then adding the large-scale context features and the small detail features element by element; then the features obtained by element-by-element addition are passed through the feature interaction module to obtain fully integrated small details and large-scale context output data, and the feature interaction module performs maximum pooling and average pooling on the features obtained by element-by-element addition and then splices them in the channel dimension, and then passes through a convolution layer with a convolution kernel size of 7×7 and a Sigmoid activation function, and then performs matrix multiplication operation with the features obtained by element-by-element addition, and finally adds the features obtained by the matrix multiplication operation to the small detail features element by element, and then passes through a 3×3 convolution layer, a 3×3 void convolution layer with a void rate d of 2, and a 1×1 convolution layer to obtain fully integrated small details and large-scale context output data. The feature interaction module consists of parallel max pooling and average pooling layers. The data from the two branches are concatenated in the channel dimension before being reduced using a 7×7 convolutional layer, converting the data channel dimension from two dimensions to one. The output data is activated using a sigmoid function to obtain corresponding spatial weights, which are then multiplied by the input data. The output data is then added to the output of the small-kernel convolution module and further fused through a 3×3 convolutional layer, a 3×3 dilated group convolutional layer with a dilation ratio of 2, and a 1×1 convolutional layer, resulting in fully integrated output data of small details and large-scale context. The large and small kernel attention modules achieve broader and deeper capture of features of objects at different scales in the input image. Small-kernel convolution layers are used to extract features such as edge details. For large-scale objects, a decomposed large kernel module is used to more accurately capture large-scale contextual features. The feature interaction module contributes significantly to improving the ability to fuse features of different scales within the network and preventing overfitting.

[0036] like Figure 4As shown, the large-scale feature selection module extracts large-scale context features, including: extracting preliminary features from the input feature map of the large-scale feature selection module through a 1×1 convolution layer and a 3×3 convolution layer; passing the preliminary features through large-core decomposition modules with different receptive fields to obtain two large-scale context features, adding the two large-scale context features obtained by the large-core decomposition module element-by-element and then performing feature adaptive enhancement through a feature adaptive enhancement module, wherein the feature adaptive enhancement module includes a global average pooling, two linear layers and a Softmax function connected in sequence, wherein the two linear layers are used for dimensionality reduction and dimensionality increase, respectively; then the two large-scale context features obtained by the large-core decomposition are matrix-multiplied with the weights output by the Softmax function, and then added element-by-element to obtain the final large-scale context feature. The large-scale feature selection module is implemented by two large-core decomposition modules of different scales and a feature adaptive enhancement module; the large-core decomposition module consists of depthwise separable convolution, hole group convolution and 1×1 convolution, with two scales and overall receptive fields of 17 and 41 respectively; the input data passes through a 1×1 convolution layer and a 3×3 convolution layer and is input to the two large-core decomposition modules respectively, and the obtained features are then adaptively enhanced to the appropriate scale by the feature adaptive enhancement module to obtain the enhanced large-scale context output data; the feature adaptive enhancement module consists of a global average pooling layer and two linear layers, which are used for downscaling respectively. The network consists of a convolutional layer with two dimensions, N×H×W×C, and a softmax function. The input data matrix has dimensions N×H×W×C. After a global average pooling layer, the output matrix has dimensions N×1×1×C. Two linear layers sequentially convert the channel dimension of the data from 2 / C to C. An activation function then outputs two channel weights α and β. These weights are multiplied by the two large kernel decomposition modules, and the two data are summed to obtain the output data extracted by the large-scale feature selection module. N represents the batch size of the input images, H and W represent the height and width of the input images, and the matrix dimensions are both N×1×1×C. The batch size is the number of images used in a training run. For example, a batch size of 8 indicates that the network trains on 8 ultrasound images at a time. This refers to the batch size of the input images. Furthermore, the size of the original ultrasound images and the labeled images are the same. The large-scale feature selection module uses convolutional layers of multiple scales to extract features at different scales and adaptively enhances the most important scales. This improves the network's feature extraction capabilities and the effective representation of features at different scales in the image, resulting in richer and more accurate feature data.

[0037] like Figure 3 As shown in the figure, the small-kernel convolution module consists of a 1×1 convolution layer and a 5×5 convolution layer in series; the input data of the small-kernel convolution module is processed by a 1×1 convolution layer and a 5×5 convolution layer in sequence, and then added to the input data of the large-scale feature selection module to obtain output data of different scale features.

[0038] In this embodiment, decomposing the preliminary large-scale context features through large kernels with different receptive fields respectively means decomposing the preliminary large-scale context features through a first large kernel decomposition module (LKDM1) and a second large kernel decomposition module (LKDM2) with different receptive fields respectively. Figure 5 As shown in the figure, both the first and second large kernel decomposition modules consist of a sequentially connected 5×5 depthwise separable convolutional layer, a 5×5 dilated convolutional module, a 7×7 dilated convolutional module, a 3×3 dilated convolutional module, and a 1×1 convolutional layer. The number of dilated group convolutions is 8. The dilation ratio d of the 5×5 dilated convolutional module, the 7×7 dilated convolutional module, and the 3×3 dilated convolutional module in the first large kernel decomposition module is 1, while the dilation ratio d of the 5×5 dilated convolutional module, the 7×7 dilated convolutional module, and the 3×3 dilated convolutional module in the second large kernel decomposition module is 3. Spatial feature data is extracted from the input data through depthwise separable convolution. Large-scale contextual features are further extracted through three dilated group convolutions with different kernel sizes. These features are then fed into a 1×1 convolutional layer for channel-wise feature interaction, resulting in the large-scale contextual feature data output by the large kernel decomposition module. The first large kernel decomposition module and the second large kernel decomposition module are different from conventional standard large kernel convolution. The appropriate and reasonable decomposition method enables the module to fully express the features of various levels of the image, greatly improving the network's feature extraction and generalization capabilities.

[0039] like Figure 6 As shown in the figure, the multi-scale feature fusion module fuses the multi-level features extracted by the CNN encoder and the Transformer encoder, including: splicing the multi-level features extracted by the CNN encoder and the Transformer encoder through the channel dimension, and then passing the channel dimension from 2048 dimensions to 1024-dimensional features through a 1×1 convolution layer. The 1024-dimensional features are then extracted through three branches: a 1×1 convolution layer, a 3×3 convolution layer, and a 3×3 dilated convolution layer with a dilation rate d of 2. The features extracted by the three branches are then spliced ​​with the original 1024-dimensional features in the channel dimension and then passed through a 1×1 convolution layer to obtain the fused multi-level features. In this embodiment, the multi-scale feature fusion module splices the data output by the dual-branch encoder in the channel dimension, uses a 1×1 convolution layer to convert the channel dimension from 2048 dimensions to 1024 dimensions, and then inputs them into a 1×1 convolution layer, a 3×3 convolution layer, a 3×3 dilated convolution layer with a dilation rate of 2, and a direct connection operation respectively. After splicing the data output by each convolution layer in the channel dimension, a 1×1 convolution layer is used to reduce the channel dimension to obtain the output data of the multi-scale feature fusion module.

[0040] like Figure 7As shown in the figure, the multi-scale attention module extracts output features including: extracting features from the input features of the multi-scale attention module through the first branch composed of a 1×1 convolution layer and a 3×3 convolution layer, and extracting features through the second branch composed of a 1×1 convolution layer and a 3×3 dilated convolution layer with a hole rate d of 2, then concatenating the channel dimensions of the features extracted by the two branches and reducing the channel dimensions through a 1×1 convolution layer, then performing maximum pooling and average pooling on the features after reducing the channel dimensions and then concatenating the features in the channel dimensions, and then performing matrix multiplication on the features after reducing the channel dimensions after dimensionality reduction and dimensionality increase by a linear layer and activation by a Sigmoid activation function to obtain the output features. In this embodiment, the multi-scale attention module inputs the output data from each layer of the CNN encoder into the module. First, the data is input into a branch consisting of a 1×1 convolution layer, a 3×3 convolution layer, a 1×1 convolution layer, and a 3×3 dilated convolution layer with a dilated rate of 2. The output features are concatenated in the channel dimension and then a 1×1 convolution layer is used to reduce the channel dimension. The data is then input into a channel attention layer to enhance the channel dimension of the feature data to obtain the output data of the multi-scale attention module.

[0041] like Figure 2 As shown in the figure, the Transformer encoder consists of a patch embedding layer, a linear layer, and n transformer layers connected in sequence. The patch embedding layer is a convolutional layer with a kernel size of 16×16 and a stride of 16. The ultrasound image with a dimension of H×W×3 is passed through the patch embedding layer to obtain features with a dimension of H / 16×W / 16×D, where H and W are the height and width of the ultrasound image, and D is the output data dimension (the default is 768). Figure 2 , where transformer layer (1) to transformer layer (n) are n transformer layers. In this embodiment, n=12, that is, the transformer encoder used in the present invention includes 12 transformer layers.

[0042] like Figure 8 As shown in the figure, the transformer layer consists of an embedded image block layer, two normalization layers, a multi-head attention layer, a multi-layer perceptron layer and two residual connections. The input feature data passes through the embedded image block layer, the first normalization layer, and the multi-head attention layer, and is then added element-by-element to the feature data output by the embedded image block layer through the first residual connection. The obtained feature data then passes through the second normalization layer and the multi-layer perceptron layer, and the feature data obtained by adding the first residual connection element-by-element is added as the output data of the transformer layer.

[0043] In this embodiment, the decoding module consists of a series of 2×2 transposed convolutions and two 3×3 convolutional layers. The decoder uses the decoding module to decode the extracted features to accurately locate the target. The decoder uses four identical decoding modules to restore the image size and output the final segmentation prediction map of the target area.

[0044] In this embodiment, the parameters of the convolutional layers in the large and small kernel attention modules, the Transformer encoder, the multi-scale feature fusion module, the multi-scale attention module, and the decoder network are determined by optimizing the loss function when training with a training set, so that the loss function is minimized; wherein, the training set includes ultrasound images and corresponding label images; the loss function refers to the error between the segmentation prediction map obtained by using the ultrasound image segmentation network model and the corresponding label image. As an optional implementation, the loss function used in this embodiment when training the ultrasound image segmentation network model is The function expression is:

[0045] ,

[0046] ,

[0047] ,

[0048] In the above formula, is the binary cross entropy loss, is the dice loss, is the sample size, is the label of the i-th pixel value, is the pixel value of the i-th pixel value, is the label graph, is the final segmentation prediction map.

[0049] To verify the effectiveness of the method in this embodiment, the BUSI and BUSC breast ultrasound image datasets, as well as the UNS neural ultrasound image dataset, were used for testing. The experiment used ultrasound image datasets with a one-to-one correspondence between original images and labels. The method in this embodiment was compared with four classic segmentation methods: UNet, UNet++, Attention-UNet, and SKUNet; two segmentation methods combining CNN and Transformer, TransUNet and CMUNext; and two large-kernel convolutional segmentation methods, CMU-Net and MDA-Net. Objective evaluation metrics used included intersection over union (IOU), Dice coefficient, Hausdorff distance, accuracy, and precision. The results are shown in Tables 1, 2, and 3.

[0050] Table 1: Objective evaluation indicators of ultrasound image segmentation results of the BUSI dataset

[0051]

[0052] Table 2: Objective evaluation indicators of ultrasound image segmentation results of BUSC dataset

[0053]

[0054] Table 3: Objective evaluation indicators of ultrasound image segmentation results of UNS dataset

[0055]

[0056] As shown in Tables 1, 2, and 3, compared with the comparison method, the method of this embodiment achieves relatively excellent performance in terms of five indicators: intersection over union (IOU), Dice coefficient, Hausdorff distance, accuracy, and precision on the BUSI, BUSC, and UNS ultrasound image datasets.

[0057] In addition, the comparison results of some samples between the method of this embodiment and the existing method are as follows: Figure 9 、 Figure 10 、 Figure 11 and Figure 12 As shown in Figure 9 and Figure 10In the figure, the ultrasound images in the first row (a-1) to (e-1) through the fourth row (a-4) to (e-4) are from the BUSI dataset, with the categories benign, benign, malignant, and normal, respectively. In the first row (a-1) to (e-1), the labeled image (a-1) has two lesions. Some networks (such as TransUNet and CMUNext) can only predict one of the lesions, while most other networks (such as SKUNet and CMU-Net) have low lesion prediction accuracy. In the second row (a-2) to (e-2), the target region in the labeled image is solid. More than half of the segmentation predictions for the comparison methods are hollow. In the third row (a-3) to (e-3), the segmentation predictions for half of the segmentation methods (i.e., UNet, UNet++, Attention-UNet, TransUNet, and MDA-Net) contain redundant regions. The remaining methods predict the lesion boundaries more coarsely. For the normal images in the fourth row (a-4) to (e-4), there are no lesions in the label image. However, all the comparison methods predicted the lesion area. The prediction errors of half of the segmentation methods (i.e., Attention-UNet method, SKUNet method, TransUNet method, CMU-Net method, and CMUNext method) are too large. It is worth emphasizing that the method of this embodiment is completely correct in predicting normal images in the BUSI dataset. Compared with other methods, the method of this embodiment has obvious advantages. The input images of the fifth row (a-5) to (e-5) and the sixth row (a-6) to (e-6) are from the BUSC dataset. The input images of the fifth row (a-5) to (e-5) contain benign breast areas, and the inputs of the sixth row (a-6) to (e-6) are malignant breast images. It can be seen that the predicted masks of many comparison methods are relatively rough for the edge details of the target area. In Figure 11 and Figure 12In the first rows (a-1) to (e-1), some comparison methods (i.e., Attention-UNet, SKUNet, TransUNet, CMUNext, and MDA-Net) predict an extra brachial plexus region in the upper right corner. In the second rows (a-2) to (e-2), segmentation predictions from many comparison methods (i.e., UNet, UNet++, SKUNet, and CMU-Net) show incomplete brachial plexus regions, resulting in the loss of many edge details within the target. In the last two rows, where no brachial plexus regions are present, most segmentation methods exhibit incorrect predictions. Compared to other segmentation methods, the method in this embodiment achieves superior segmentation performance in the brachial plexus region. It effectively captures image edge details, suppresses interference from irrelevant tissue surrounding the target, and globally distinguishes regions similar to the target. This method enhances the robustness of the target segmentation network in complex backgrounds.

[0058] In order to verify the contribution of each module in the ultrasound image segmentation network model of the method of this embodiment, this embodiment makes adjustments on the baseline model respectively. By comparing the four improved parts, namely the large and small kernel attention module (LSKA), the multi-scale attention module (MSA), the Transformer module and the multi-scale feature fusion module (MSF2), ablation experiments are conducted. The target segmentation accuracy of the BUSI dataset is experimentally tested to evaluate the effectiveness of each part. The experimental results are shown in Table 4.

[0059] Table 4: Objective evaluation indicators of the ablation experiment segmentation results of the method of the present invention on BUSI ultrasound images

[0060]

[0061] Table 4 clearly shows that the segmentation network proposed in this embodiment significantly improves performance compared to the baseline network (Unet). Specifically, the intersection over union (IOU) metric improved by 7.80%, the Dice coefficient improved by 6.80%, the Hausdorff distance improved by 5.26, the accuracy improved by 0.36%, and the precision improved by 1.97%. In the second row of Table 3, the proposed LSKA module achieves a 4.86% improvement in IOU. This demonstrates that the combination of the large and small kernel feature extraction modules enables the segmentation algorithm to learn more reliable breast ultrasound image details and large-scale contextual features.

[0062] In addition, in order to verify the contribution of different levels of feature extraction modules in the method of this embodiment to the overall performance, this embodiment conducts ablation experiments on the ultrasound image segmentation network model of the method of this embodiment by removing the small kernel module (SKM), the multi-scale feature selection module (LSFS), and the Transformer module. The target segmentation accuracy of the BUSI dataset is experimentally tested to evaluate the effectiveness of each part. The experimental results are shown in Table 5.

[0063] Table 5: Objective evaluation indicators of the ablation experiment segmentation results of different levels of feature extraction modules on BUSI ultrasound images

[0064]

[0065] The first row of Table 5 shows the baseline network without the Small Kernel Module (SKM), Multi-Scale Feature Selection Module (LSFS), and Transformer module. The three levels of feature extraction modules can extract local small details, large-scale context, and global features. The second to fourth rows of Table 5 show the ablation experiments of the baseline combined with the three separate feature extraction modules. Figure 5 As can be observed, each independent module achieves better segmentation results than the baseline module. Rows 5 to 7 show the network results for the baseline combined with the two modules. The results show that the network combining the two modules outperforms each individual module. The last row shows the proposed method combining the baseline with three levels of feature extraction modules. The segmentation performance of this embodiment method outperforms other networks. Ablation experiment results confirm that the multi-level feature extraction network has an impact on improving ultrasound image segmentation accuracy.

[0066] In summary, the ultrasound image segmentation network model of the ultrasound image segmentation method based on multi-level feature extraction in this embodiment is a dual-encoder single-decoder structure. The input image is subjected to the large and small kernel attention modules of the CNN encoder to extract small detail features and large-scale local context features. The small detail features and large-scale context features of different scales are adaptively selected and enhanced by adaptive weighting. The global features of the image are extracted based on the Transformer encoder. The multi-level features extracted by the two encoders are then subjected to the multi-scale feature fusion module for information interaction between different-level features. In addition, this embodiment designs a multi-scale attention module on the jump connection to retain important shallow features to facilitate the subsequent fusion of deep and shallow features. The fused multi-level features are decoded layer by layer by the decoder to obtain the final segmentation prediction map. The present invention adopts the large and small kernel attention modules to adaptively select and enhance features of different scales. The large kernel decomposition method reduces the number of model parameters while achieving a large receptive field, and also improves the network's feature extraction ability and model generalization ability. The multi-level feature extraction method designed by this method makes the obtained features more sufficient, which is more conducive to improving image segmentation performance. The method of this embodiment fully extracts small detail features, large-scale context features, and global features from the image through a parallel dual encoder structure. It uses a small convolution kernel module and a large kernel decomposition module of different scales to extract small detail features and large-scale features, as well as adaptively select and enhance large-scale features. It also uses convolution and attention layers to interact information between features. It uses two large kernel decomposition modules with different receptive fields to extract large-scale context features. After adding the output data, it obtains two different attention weights through a channel weight generation layer. Each weight is multiplied by the corresponding large kernel decomposition module result to obtain adaptively selected and enhanced large-scale context features. The large kernel is decomposed into a combination of convolutions of different types and scales. While achieving large-scale feature extraction, it can also significantly reduce the number of model parameters. Compared with traditional decomposition forms, the design of this unit is more efficient and lightweight for feature extraction in images. Compared with the existing technology, the method of this embodiment has the following significant advantages: 1. The multi-level feature extraction design of the ultrasound image segmentation network model of this embodiment realizes the extraction of small detail features, large-scale context features, and global features, improving the information capture efficiency of the input image and further improving the segmentation performance of the network. 2. The large and small kernel attention modules in this embodiment can capture features of objects of different sizes in the input image. Small kernel convolution layers are used to extract features such as edge details, while decomposed large kernel modules are used to more accurately capture large-scale contextual features of large-scale objects. The feature interaction module makes a significant contribution to improving the ability to fuse and express features of different scales within the network and preventing overfitting.3. The large-scale feature selection module of this embodiment utilizes convolutional layers of multiple scales to extract features at different scales and adaptively selects and enhances important scales. This improves the network's feature extraction capabilities and the effective representation of features at different scales in the image, resulting in richer and more accurate feature data. 4. The large-core decomposition module of this embodiment uses an effective decomposition method to reduce the number of model parameters while fully extracting large-scale contextual features, thereby improving the model's trainability.

[0067] In addition, this embodiment also provides an ultrasound image segmentation system based on multi-level feature extraction, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction.

[0068] In addition, this embodiment also provides a computer-readable storage medium, which stores a computer program or instruction. The computer program or instruction is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction through a processor.

[0069] In addition, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction through a processor.

[0070] Those skilled in the art should understand that the technical solutions provided by the embodiments of the present application may be in the form of methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0071] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An ultrasound image segmentation method based on multi-level feature extraction, characterized in that: The method comprises performing ultrasonic image segmentation on an input ultrasonic image using an ultrasonic image segmentation network model to obtain a segmentation prediction map, wherein the ultrasonic image segmentation network model comprises a dual-branch encoder, a multi-scale feature fusion module, a multi-scale feature attention module and a decoder, wherein the dual-branch encoder comprises a CNN encoder and a Transformer encoder, and performing ultrasonic image segmentation on the input ultrasonic image using the ultrasonic image segmentation network to obtain a segmentation prediction map comprises: extracting small detail features and large-scale context features of different scales from the input ultrasonic image through the CNN encoder and adaptively selecting and enhancing the small detail features and large-scale context features of different scales using adaptive weighted weights; extracting global features from the input ultrasonic image through the Transformer encoder; fusing the multi-level features extracted by the CNN encoder and the Transformer encoder through the multi-scale feature fusion module; enhancing the shallow features output by each layer of the CNN encoder using the multi-scale feature attention module and outputting the enhanced shallow features to the decoder, and decoding the fused multi-level features and the shallow features enhanced by the multi-scale feature attention module layer by layer through the decoder to obtain a final segmentation prediction map; The CNN encoder includes N scale encoding branches, the first scale encoding branch includes a 1×1 convolution layer and a large and small kernel attention module, and the remaining N-1 scale encoding branches include a downsampling module, a channel dimension splicing module and a large and small kernel attention module. The output of the 1×1 convolution layer in the first scale encoding branch serves as the input of the downsampling module in the second scale encoding branch, and the input of the remaining downsampling modules is the output of the channel dimension splicing module in the previous scale encoding branch. The output of all channel dimension splicing modules also serves as the input of the large and small kernel attention modules in the encoding branch of the same scale. The output of the large and small kernel attention modules in the encoding branch of the next scale is used as one input of the channel dimension splicing module in the encoding branch of the next scale, and the output of the large and small kernel attention modules in the encoding branch of the Nth scale is used as the input of the multi-scale feature fusion module CNN encoder end. The decoder includes N-1 levels of decoding modules connected in cascade, and the multi-scale feature fusion module is used as the input of the first level decoding module. The other input of each level of decoding module is the output of the large and small kernel attention modules in the encoding branch of the corresponding scale, and the output feature is extracted by the multi-scale attention module. The final segmentation prediction map is decoded and output by the N-1th level decoding module; The processing of the input feature map by the large and small core attention module includes: extracting large-scale context features from the input feature map of the large and small core attention module through a large-scale feature selection module, and extracting small detail features from the input feature map of the large and small core attention module through a small-core convolution module, and then adding the large-scale context features and the small detail features element by element; then the features obtained by element-by-element addition are passed through a feature interaction module to obtain fully integrated small detail and large-scale context output data, and the feature interaction module performs maximum pooling and average pooling on the features obtained by element-by-element addition and then splices them in the channel dimension, and then passes through a convolution layer with a convolution kernel size of 7×7 and a Sigmoid activation function, and then performs matrix multiplication operation with the features obtained by element-by-element addition, and finally adds the features obtained by the matrix multiplication operation to the small detail features element by element, and then passes through a 3×3 convolution layer, a 3×3 void convolution layer with a void rate d of 2, and a 1×1 convolution layer to obtain fully integrated small detail and large-scale context output data.

2. The ultrasonic image segmentation method based on multi-level feature extraction according to claim 1, characterized in that: The large-scale feature selection module extracts large-scale context features, including: extracting preliminary features from the input feature map of the large-scale feature selection module through a 1×1 convolution layer and a 3×3 convolution layer; respectively passing the preliminary features through large-core decomposition modules with different receptive fields to obtain two large-scale context features, adding the two large-scale context features element by element, and then performing feature adaptive enhancement through a feature adaptive enhancement module, wherein the feature adaptive enhancement module includes a global average pooling, two linear layers, and a Softmax function connected in sequence, wherein the two linear layers are used for dimensionality reduction and dimensionality increase, respectively; then, the two large-scale context features obtained by the large-core decomposition are matrix-multiplied with the weights output by the Softmax function, and then added element by element to obtain the final large-scale context features; the small-core convolution module is composed of a 1×1 convolution layer and a 5×5 convolution layer in series. The method comprises the following steps: passing the preliminary large-scale context features through the large-core decomposition modules with different receptive fields respectively, extracting the preliminary features through the first large-core decomposition module and the second large-core decomposition module with different receptive fields respectively, and the first large-core decomposition module and the second large-core decomposition module each include a 5×5 depthwise separable convolution layer, a 5×5 dilated convolution module, a 7×7 dilated convolution module, a 3×3 dilated convolution module and a 1×1 convolution layer connected in sequence, wherein the number of groups of the dilated group convolution is 8, the dilated rate d of the 5×5 dilated convolution module, the 7×7 dilated convolution module and the 3×3 dilated convolution module in the first large-core decomposition module is 1, and the dilated rate d of the 5×5 dilated convolution module, the 7×7 dilated convolution module and the 3×3 dilated convolution module in the second large-core decomposition module is 3.

3. The ultrasonic image segmentation method based on multi-level feature extraction according to claim 1, characterized in that: The multi-scale feature fusion module fuses the multi-level features extracted by the CNN encoder and the Transformer encoder, including: splicing the multi-level features extracted by the CNN encoder and the Transformer encoder through the channel dimension, and then passing the channel dimension from 2048 dimensions to 1024-dimensional features through a 1×1 convolution layer, and then extracting features from the 1024-dimensional features through three branches of a 1×1 convolution layer, a 3×3 convolution layer, and a 3×3 hole convolution layer with a hole rate d of 2, and then splicing the features extracted by the three branches with the original 1024-dimensional features in the channel dimension, and then passing the 1×1 convolution layer to reduce the channel dimension to 1024 dimensions to obtain the fused multi-level features.

4. The ultrasonic image segmentation method based on multi-level feature extraction according to claim 2, characterized in that: The multi-scale attention module extracts output features including: extracting features from the input features of the multi-scale attention module through a first branch consisting of a 1×1 convolution layer and a 3×3 convolution layer, and extracting features through a second branch consisting of a 1×1 convolution layer and a 3×3 dilated convolution layer with a dilation rate d of 2, concatenating the channel dimensions of the features extracted by the two branches, and then reducing the channel dimensions through a 1×1 convolution layer, performing maximum pooling and average pooling on the features after reducing the channel dimensions, and then concatenating the channel dimensions, and then performing matrix multiplication on the features after reducing the channel dimensions after dimensionality reduction by a linear layer, dimensionality increase by a linear layer, and activation by a Sigmoid activation function to obtain output features.

5. The ultrasonic image segmentation method based on multi-level feature extraction according to claim 1, characterized in that: The Transformer encoder consists of an image block embedding layer, a linear layer and multiple transformer layers connected in sequence, wherein the image block is embedded as a convolution layer with a convolution kernel size of 16×16 and a stride of 16. The ultrasound image with a dimension of H×W×3 is passed through the image block embedding layer to obtain features with a dimension of H / 16×W / 16×D, where H and W are the height and width of the ultrasound image, and D is the output data dimension. The transformer layer consists of an embedded image block layer, two normalization layers, a multi-head attention layer, The invention consists of a multi-layer perceptron layer and two residual connections. The input feature data passes through the embedded image block layer, the first normalization layer, and the multi-head attention layer, and is then added element-by-element to the feature data output by the embedded image block layer through the first residual connection. The obtained feature data passes through the second normalization layer and the multi-layer perceptron layer, and is then added element-by-element to the first residual connection through the second residual connection. The obtained feature data is added as the output data of the transformer layer; the decoding module consists of a transposed convolution with a convolution kernel size of 2×2 and two 3×3 convolution layers connected in sequence.

6. An ultrasound image segmentation system based on multi-level feature extraction, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction according to any one of claims 1 to 5 through a processor.

8. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the ultrasound image segmentation method based on multi-level feature extraction according to any one of claims 1 to 5 through a processor.

Images