A method and apparatus for forearm blood vessel image segmentation based on HAMU-Net deep learning network

Abstract

This invention provides a method and apparatus for forearm blood vessel image segmentation based on the HAMU-Net deep learning network, belonging to the field of medical image processing technology. The method includes: acquiring forearm blood vessel images using a near-infrared acquisition device to obtain an original image; preprocessing the original image based on a Gaussian filter and a preset background grayscale threshold to obtain a processed image; performing multi-stage image enhancement processing using a residual convolutional autoencoder based on the processed image using Reinforcement-CLAHE technology to obtain an enhanced image; constructing a HAMU-Net blood vessel segmentation model based on a Hybrid Attention module, a Triplet Attention module, a Spatial Attention module, and a U-Net network structure; and inputting the enhanced image into the HAMU-Net blood vessel segmentation model to obtain a segmented blood vessel image. This invention is an efficient and accurate near-infrared forearm blood vessel image segmentation method based on a hybrid attention module.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method and apparatus for segmenting forearm blood vessel images based on the HAMU-Net deep learning network. Background Technology

[0002] In recent years, near-infrared optical imaging technology has gradually become an important non-invasive imaging tool in the biomedical field, widely used for the visualization and monitoring of subcutaneous vascular structures. However, the visibility of subcutaneous vascular structures is often limited by factors such as absorption, scattering, and light attenuation by biological tissues, making the results unsuitable for direct practical application. Therefore, there is a strong need for an effective image enhancement and segmentation method to improve the visualization quality of subcutaneous veins, providing strong support for assisting in venous puncture procedures in clinical medicine.

[0003] Medical image segmentation is a core task in the field of medical image processing. Currently, commonly used medical image segmentation methods are mainly divided into traditional segmentation methods and deep learning segmentation methods based on convolutional neural networks. Traditional segmentation methods can be further categorized into threshold-based, edge-based, and region-based methods. Threshold-based methods, such as full-threshold segmentation and multi-threshold segmentation, segment images based on thresholds of pixel grayscale values. These methods are simple and computationally fast, suitable for processing images with high contrast, but sensitive to complex backgrounds and noise. Edge-based methods, such as edge detection and active contour models, aim to detect edges or contours in images, but perform poorly on images with complex textures and large lighting variations. Region-based methods, such as region growing and region segmentation, divide image pixels into several adjacent regions or segmentation units, resulting in continuous and smooth edges, but are sensitive to the selection of seed points. In recent years, the rise of deep learning technology has led to a revolutionary change in the field of medical image segmentation. Deep learning methods, especially convolutional neural networks and semantic segmentation networks, achieve a high degree of understanding of image semantic information by learning image features from large-scale data. These deep learning methods can use learned features for pixel-level classification and segmentation, thereby achieving more accurate and automated medical image segmentation. However, compared to traditional methods, deep learning often requires a large amount of manually labeled training data to train the model. Medical images are more difficult to obtain than other types of images, and the labeling process usually requires the labeler to have certain professional knowledge and consumes a lot of time.

[0004] U-Net, a variant of Fully Convolutional Networks (FCNs), is the most commonly used and simplest segmentation model. This model is efficient, easy to build, and supports training on small datasets with few samples. U-Net was originally designed to solve medical image segmentation problems, especially cell-level segmentation tasks. Each sub-module of U-Net is replaced with a residual module to design Residual U-Net (Res-Unet), which is then applied to retinal image segmentation. To enhance relevant features and suppress irrelevant features, Attention U-Net is proposed, which introduces an attention mechanism before concatenating the features at each resolution of the encoder with the corresponding features in the decoder. A Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) is proposed to readjust the encoder's output features. This method combines residual connections and recurrent convolutions to replace the original sub-modules in U-Net, maintaining network depth while mitigating the vanishing gradient effect. Finally, a fully automatic segmentation model based on spatial attention and residual recurrent convolution is proposed. The U-Net (AttR2U-Net) model is based on the classic U-Net and incorporates advanced mechanisms such as spatial attention, residual connections, recurrent convolution, and normalization to improve segmentation performance. The U-Net network based on residual multi-scale fusion and attention skipping layers (ResidualAttention U-Net, RAU-Net) adds an enhanced attention module to fuse multi-level features to solve the specular reflection problem in cataract surgery instrument segmentation.

[0005] The U-Net series of models has achieved significant advantages in the field of medical image segmentation, but current research mostly focuses on segmenting human organs or retinal vessels, with almost no results applied to near-infrared forearm vessel images. Furthermore, the performance of existing models is highly dependent on the size of the dataset and the proportion of effective regions within the image. Their segmentation results are less than ideal for forearm vessel images, which have a relatively small number of images, complex contours, but whose distribution is less dense than that of retinal vessels.

[0006] In the existing technology, there is a lack of an efficient and accurate near-infrared forearm blood vessel image segmentation method based on a hybrid attention module. Summary of the Invention

[0007] To address the technical problems of existing technologies, such as the high dependence of model performance on dataset size and the proportion of effective regions within images, and the unsatisfactory segmentation results for forearm blood vessel images with a small number of images, complex contours, and sparse distribution, this invention provides a forearm blood vessel image segmentation method and apparatus based on the HAMU-Net deep learning network. The technical solution is as follows:

[0008] On the one hand, a forearm blood vessel image segmentation method based on the HAMU-Net deep learning network is provided. This method is implemented by a forearm blood vessel image segmentation device and includes:

[0009] Forearm vascular images are acquired using a near-infrared acquisition device to obtain raw images;

[0010] Based on a Gaussian filter, the original image is preprocessed according to a preset background grayscale threshold to obtain a processed image;

[0011] Based on the Reinforcement-CLAHE technique, a multi-stage image enhancement process is performed using a residual convolutional autoencoder on the processed image to obtain an enhanced image.

[0012] The HAMU-Net vessel segmentation model is constructed based on the Hybrid Attention module, Triplet Attention module, Spatial attention module, and U-Net network structure.

[0013] The enhanced image is input into the HAMU-Net blood vessel segmentation model to obtain a blood vessel segmentation image.

[0014] On the other hand, a forearm blood vessel image segmentation device based on the HAMU-Net deep learning network is provided. This device is applied to the forearm blood vessel image segmentation method based on the HAMU-Net deep learning network. The device includes:

[0015] The image acquisition module is used to acquire images of forearm blood vessels using a near-infrared acquisition device to obtain raw images.

[0016] The image preprocessing module is used to preprocess the original image based on a Gaussian filter and according to a preset background grayscale threshold to obtain a processed image.

[0017] The image enhancement module is used to perform multi-stage image enhancement processing based on the Reinforcement-CLAHE technology, using a residual convolutional autoencoder to obtain an enhanced image;

[0018] The model building module is used to build the HAMU-Net blood vessel segmentation model based on the Hybrid Attention module, Triplet Attention module, Spatial attention module, and U-Net network structure.

[0019] The image segmentation module is used to input the enhanced image into the HAMU-Net blood vessel segmentation model to obtain a blood vessel segmentation image.

[0020] On the other hand, a forearm blood vessel image segmentation device is provided, the forearm blood vessel image segmentation device comprising: a processor; a memory storing computer-readable instructions, wherein when the computer-readable instructions are executed by the processor, any one of the forearm blood vessel image segmentation methods based on the HAMU-Net deep learning network described above is implemented.

[0021] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described forearm blood vessel image segmentation methods based on the HAMU-Net deep learning network.

[0022] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0023] This invention proposes a forearm vessel image segmentation method based on the HAMU-Net deep learning network. Through multiple effective segmentation network architectures and modules, including an encoder-decoder structure, a U-Net backbone network, HybridAttention, Triplet Attention, and Spatial Attention, as well as a Deeplabv3 segmentation head and an FCN auxiliary segmentation head, the HAMU-Net vessel segmentation network model possesses significant advantages such as multi-level feature extraction, multi-scale receptive fields, and high segmentation accuracy, thus improving the model's robustness and generalization ability in forearm vessel image segmentation tasks. This invention is an efficient and accurate near-infrared forearm vessel image segmentation method based on hybrid attention modules. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1This is a flowchart of a forearm blood vessel image segmentation method based on the HAMU-Net deep learning network provided by an embodiment of the present invention;

[0026] Figure 2 This is a schematic diagram of a target image enhancement process provided in an embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of the HAMU-Net blood vessel segmentation network model structure with the introduction of a hybrid attention module provided in an embodiment of the present invention;

[0028] Figure 4 This is a schematic diagram of the Hybrid Attention module, Triplet Attention module, and Spatial attention module provided in the embodiments of the present invention;

[0029] Figure 5 This is a schematic diagram of the Deeplabv3 segmentation head and the FCN auxiliary segmentation head provided in an embodiment of the present invention;

[0030] Figure 6 This is a block diagram of a forearm blood vessel image segmentation device based on the HAMU-Net deep learning network provided in an embodiment of the present invention;

[0031] Figure 7 This is a schematic diagram of the structure of a forearm blood vessel image segmentation device provided in an embodiment of the present invention. Detailed Implementation

[0032] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0033] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0034] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0035] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0036] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0037] This invention provides a forearm blood vessel image segmentation method based on the HAMU-Net deep learning network. This method can be implemented using a forearm blood vessel image segmentation device, which can be a terminal or a server. Figure 1 The flowchart shown is for a forearm blood vessel image segmentation method based on the HAMU-Net deep learning network. The processing flow of this method can include the following steps:

[0038] S1. Acquire raw images of forearm blood vessels using a near-infrared acquisition device.

[0039] In one feasible implementation, the present invention uses a near-infrared acquisition device to acquire initial raw images and sets the image size in batches to 960×720 pixels in order to perform subsequent preprocessing calculations.

[0040] S2. Based on a Gaussian filter, the original image is preprocessed according to a preset background grayscale threshold to obtain the processed image.

[0041] In one feasible implementation, since the grayscale distribution of the foreground and background differs greatly in near-infrared forearm blood vessel images, threshold segmentation can be used to remove interference caused by impure background colors, thereby improving the adaptability and generalization ability of the algorithm.

[0042] Optionally, based on a Gaussian filter and according to a preset background grayscale threshold, the original image is preprocessed to obtain a processed image, including:

[0043] Based on a preset background grayscale threshold, background noise is removed from the original image to obtain the first image;

[0044] The first image is input into a Gaussian filter to remove noise, thus obtaining the second image;

[0045] The second image is contrast-stretched to obtain the processed image.

[0046] In one feasible implementation, since the grayscale intensity of the background portion of the original image is significantly higher than that of the foreground portion, background noise can be removed by calculating a segmentation threshold. The degree of separation between the foreground and background can be controlled by adjusting the value of the threshold T. The threshold-based background removal method is as follows (1):

[0047] (1)

[0048] Where O(x, y) represents the pixel value of the output image O at coordinates (x, y); I(x, y) represents the pixel value of the input image I at coordinates (x, y); T represents a user-defined threshold value; for each pixel (x, y) of the input image I, if its grayscale value is greater than the threshold T, the corresponding position in the output image O is set to 255, representing the foreground; otherwise, it is set to 0, representing the background. Otherwise represents the case where the pixel value of the input image I at coordinates (x, y) is less than the threshold T.

[0049] To further suppress noise in the original image and improve image quality, this invention employs a Gaussian filter for noise removal. The Gaussian filter is a linear filter that effectively suppresses salt-and-pepper noise and Gaussian noise, smoothing the image.

[0050] The Gaussian function is used as the filter kernel to perform a convolution operation on the first image. The Gaussian function is a bell-shaped curve with a peak value, which represents the weight of a pixel's distance from the center pixel. The farther a pixel is from the center pixel, the lower its weight. Therefore, when smoothing an image, the Gaussian filter emphasizes the center pixel more, resulting in less blurring compared to the mean filter. When the mean μ = 0, the two-dimensional form of the Gaussian function is as follows (2):

[0051] (2)

[0052] Here, σ represents the standard deviation of the normal distribution. The value of the standard deviation determines the rate of decay of the function, controls the width of the Gaussian function, and affects the smoothness of the filter. A smaller standard deviation produces a narrower curve, preserving more details, while a larger standard deviation produces a wider curve, resulting in a stronger smoothing effect.

[0053] Near-infrared forearm blood vessel images after background removal and Gaussian filtering can show the basic blood vessel outline, but the contrast between the blood vessel area and other parts of the arm is poor. The contrast stretching algorithm can initially improve this problem.

[0054] Contrast stretching can separate the gray levels of dark and bright areas in an image, making details more prominent and helping to visually identify objects or features in the image. Its principle is to remap the gray level range of the image, stretching the gray values ​​in the original image to a wider range, thereby increasing the difference between different gray levels and making the image look clearer and more suitable for subsequent analysis or visualization. The stretching process is represented by the following equation (3):

[0055] (3)

[0056] Among them, I out (x, y) are the pixel values ​​of the output image; I in (x, y) are the pixel values ​​of the input image; I min and I max These are the minimum and maximum pixel values ​​of the input image; O min and O max These are the minimum and maximum pixel values ​​required for the output image. The pixel values ​​of the input image are mapped to a new range in the output image, which is defined by O. min and O max Definition. Therefore, darker areas in the original image will be made brighter, and brighter areas will be made darker, expanding the contrast range between blood vessels and other tissues.

[0057] S3. Based on the Reinforcement-CLAHE technique, a multi-stage image enhancement process is performed using a residual convolutional autoencoder to obtain an enhanced image.

[0058] One feasible implementation method is, for example Figure 2 As shown, in the near-infrared image enhancement stage, this invention proposes a multi-stage image enhancement method to effectively enhance blood vessels and veins, preparing the model for automatic and accurate segmentation of blood vessels. The basic idea is to input the preprocessed image into a Residual Convolutional Autoencoder (RCAE) network and use the Reinforcement-CLAHE enhancement method to enhance the results. The enhanced image is obtained by proportionally combining the outputs of the RCAE and Reinforcement-CLAHE.

[0059] Optionally, based on the Reinforcement-CLAHE technique, a multi-stage image enhancement process is performed using a residual convolutional autoencoder on the processed image to obtain an enhanced image, including:

[0060] Based on the processed image, image enhancement processing is performed using a residual convolutional autoencoder to obtain the first enhanced image;

[0061] Based on the Reinforcement-CLAHE technique, the first enhanced image is subjected to image enhancement processing to obtain the second enhanced image;

[0062] The first enhanced image and the second enhanced image are weighted and fused to obtain the enhanced image.

[0063] In one feasible implementation, the Reinforcement-CLAHE of the present invention includes Contrast Limited Adaptive Histogram Equalization (CLAHE) and adaptive brightness contrast adjustment. By applying CLAHE to images at multiple scales, the large-scale structure and small-scale details of the image can be enhanced simultaneously, while suppressing excessive noise amplification. Using this method, the model can extract vascular information to the maximum extent under the objective constraints of small sample data, achieving pixel-level automatic segmentation. This method reasonably preserves the contour features of vascular images at each stage, and experiments have shown that the effective combination ratio achieves the purpose of significantly enhancing the vascular region and suppressing noise interference. The calculation process of the multi-stage image enhancement method is shown in the figure, and the linear weighted calculation process is shown in equation (4):

[0064] (4)

[0065] in, Represents the target output image, I a I represents an image enhanced only by the RCAE network. b This represents the image after RCAE and CLAHE enhancement, where η represents the weighting coefficient.

[0066] A schematic diagram of the RCAE network structure. The RCAE network consists of three blocks, each including a convolutional layer, an activation function layer, and a normalization layer. During RCAE network training, this paper considers the augmented image to be a linear combination of the actual input image and a binary mask with vein annotations. If x represents the input forearm blood vessel image and h is the manually annotated binary mask of the blood vessels, then the target image y is obtained as: y = αx + (1—α)h, where α is the weighting parameter.

[0067] Where x, h, and y must have the same dimension. The generation process of the target image is shown in equation (5):

[0068] (5)

[0069] Where λ is a fixed weight parameter, I ori Represents the input image, I mask A binary mask image representing vein annotations, O tar The target image generated by the model must have the same dimensions.

[0070] S4. Construct the HAMU-Net blood vessel segmentation model based on the Hybrid Attention module, Triplet Attention module, Spatial attention module, and U-Net network structure.

[0071] In one feasible implementation, U-Net employs a unique U-shaped network structure, divided into downsampling (encoder) and upsampling (decoder) parts, to effectively segment images. Downsampling gradually reduces the spatial resolution of the image while capturing its high-level features, while upsampling expands the feature map to the resolution of the original input image through deconvolution operations. The encoder, the downsampling part of U-Net, consists of a series of convolutional layers, pooling layers, and activation functions. These layers extract feature representations of the image. Each pooling layer reduces the spatial size of the feature map while increasing the abstraction of the features. This allows the network to capture feature information at different scales, from lower-level texture to higher-level semantic information. The decoder gradually increases the spatial resolution of the feature map through deconvolution operations. Furthermore, U-Net employs a skip connection strategy, fusing features from the encoder with those from the decoder. This strategy helps preserve low-level feature information and improves the accuracy of the segmentation results.

[0072] The HAMU-Net vessel segmentation model includes the HAMU-Net backbone network, skip connection part, bottleneck layer and output layer.

[0073] One feasible implementation method is, for example Figure 3 As shown, the encoder is responsible for extracting feature information from the input image. It contains a series of convolutional and pooling layers that gradually reduce the spatial size of the feature map and increase the number of channels to capture different features in the image, such as blood vessel edges, texture, and shape. The encoder's output is a set of high-dimensional feature representations that contain abstract feature information from the input image.

[0074] The decoder restores the high-dimensional feature maps extracted by the encoder to the same size as the input image. It contains a series of transposed convolutional layers that progressively increase the spatial size of the feature maps while decreasing the number of channels. The decoder's task is to transform the high-dimensional feature maps into pixel-level outputs for vascular image enhancement. A residual connection is introduced between the encoder and decoder to generate a difference component, which helps the network learn the differences between the input data and the target.

[0075] The size of the convolution kernel is a crucial factor in determining the receptive field. Since the size and location of the effective region in a vascular image, as well as the width of the vascular contour, can vary randomly within a given range, the size of the convolution kernel cannot be estimated intuitively. The convolution kernels in the encoder section gradually decrease from 9×9 to 3×3, while the kernel size in the decoder section gradually increases.

[0076] A fully convolutional network (FCN) is added as an auxiliary segmentation head near the output layer of the decoder to increase the network's perceptual ability and improve segmentation performance. For example... Figure 4 As shown, the model contains detailed components including the Hybrid Attention module, Triplet Attention module, Spatial Attention module, Deeplabv3Head, and FCN Head.

[0077] The HAMU-Net backbone network includes an encoder and a decoder; the encoder includes nine 3×3 convolutional layers and four 2×2 downsampling layers; the decoder includes nine 3×3 convolutional layers, four 2×2 upsampling layers and one Spatialattention module.

[0078] The skip connection part includes three first skip connections and two second skip connections; the first skip connections introduce the Hybrid Attention module;

[0079] The bottleneck layer consists of one Triplet Attention module; the Triplet Attention module includes a first branch, a second branch, and a third branch;

[0080] The output layer includes a main segmentation head and an auxiliary segmentation head; the main segmentation head includes one 3×3 convolutional layer, one Deeplabv3 Head, and one 1×1 convolutional layer; the auxiliary segmentation head includes one 3×3 convolutional layer and one FCN Head.

[0081] In one feasible implementation, the encoder side includes nine 3×3 convolutional layers and four 2×2 downsampling layers; the decoder side includes nine 3×3 convolutional layers, four 2×2 upsampling layers, and one spatial attention module; there are four skip connections, the first three of which introduce a Hybrid Attention module; the bottleneck layer is a Triplet Attention module; the output layer includes two segmentation head branches, the main segmentation head branch includes one 3×3 convolutional layer, one Deeplabv3 Head, and one 1×1 convolutional layer, and the auxiliary segmentation head includes one 3×3 convolutional layer and one FCN Head.

[0082] A Triplet Attention module is introduced in the part connecting the deep encoder and the bottleneck layer to help the model capture global contextual information from three dimensions (height, width, and channels) of the feature map. This is very useful for complex scenarios in blood vessel segmentation tasks, because blood vessels usually extend throughout the entire image and require global information to understand their structure.

[0083] Traditional Convolutional Block Attention Modules (CBAMs) and Squeeze-and-Excitation Networks (SENets) require a certain number of learnable parameters to establish inter-channel dependencies. The Triplet Attention mechanism used in this invention employs a novel approach, using an attention mechanism with almost no parameters to model channel attention and spatial attention.

[0084] The Triplet Attention module's architecture comprises three parallel branches. The first branch, similar to traditional CBAM, establishes channel attention; however, unlike CBAM, Triplet Attention requires fewer parameters to implement this branch. The second and third branches capture the cross-dimensional interactions between the channel dimension C and either the spatial dimension H or W. These branches are implemented with almost no parameters, making the model more lightweight. The Triplet Attention module simply averages and sums the outputs of these three branches to obtain comprehensive attention information across both the channel and spatial dimensions. This approach not only improves the model's computational efficiency but also makes it more suitable for resource-constrained environments while maintaining good performance.

[0085] Using the Spatial Attention module in the shallow contact output layer of the decoder allows the model to focus on key spatial locations in the feature map, especially the edges of blood vessels and the segmentation of small branches. The core idea of ​​Spatial Attention is to generate a weight map by analyzing information in the spatial dimensions (i.e., height and width) of the input feature map, which enhances the model's attention to important spatial locations.

[0086] Suppose the input feature map has dimensions (C × H × W), meaning height H, width W, and number of channels C. At this stage, the model has not yet determined the importance of specific spatial locations. Spatial Attention uses pooling operations to summarize the spatial information of the feature map, extracting globally representative information.

[0087] Global average pooling is performed on the feature map along the channel dimension, averaging the channel values ​​at each location. This yields the average feature response at each location, forming a spatially averaged feature map. Max pooling is then performed on the feature map along the channel dimension to extract the maximum activation value in the channel at each location, capturing the strongest response at each spatial location and forming a spatially maximum feature map.

[0088] After average pooling and max pooling, the two feature maps represent the global information at each spatial location. The results of max pooling and average pooling are then added element-wise to obtain a fused spatial feature map, which retains both the average response value and the maximum response value at each location.

[0089] The spatial feature map is further processed by convolution to generate weight values ​​for each spatial location. The convolution kernel sizes include 7×7 and 5×7, and multiple receptive fields capture spatial context information. The value of each location in the convolutional feature map represents the importance of that location. After batch normalization and the Sigmoid activation function, a spatial weight map is obtained. Finally, the generated spatial weight map is multiplied element-wise with the original input feature map, so that the model weights important spatial locations and weakens unimportant locations. The calculation process is as follows: (6) and (7):

[0090] (6)

[0091] (7)

[0092] Where X represents the input feature map, A represents the attention weight, Y represents the output feature map, σ is the Sigmoid activation function, ﹒ represents element-wise multiplication, Conv is the convolution operation, and Pool is the pooling operation.

[0093] In the skip connection part, the Hybrid Attention module is used, with Triplet Attention and Spatial Attention running sequentially. The model can capture global multidimensional information and refine local spatial features, enhancing these features at both the global and local levels, enabling the decoder to better utilize these detailed information. In the Hybrid Attention module, the calculation process of attention weights and feature maps is similar to that of the two attention methods described above, and will not be repeated here. The calculation process of attention weights and feature maps in the Hybrid Attention module is as follows: (8) and (9):

[0094] (8)

[0095] (9)

[0096] Where M is the input feature map, N is the feature map after Triplet Attention, and P is the output feature map after Spatial Attention.

[0097] like Figure 5 As shown, the output layer contains two branches: the Deeplabv3 Head and the FCN Head. The Deeplabv3 Head, as the main segmentation head of the model, receives features from the backbone network and is responsible for generating the final segmentation result. One of the core principles of the Deeplabv3 network is the use of dilated convolution, also known as attenuated convolution.

[0098] Traditional convolution operations operate on the input with a fixed kernel size, while dilated convolution introduces a dilation rate parameter, allowing the kernel to operate on the input with different strides. This enables the network to capture information from images at multiple receptive fields, better handling vascular structures of different sizes. The Atrous Spatial Pyramid Pooling (ASPP) module is a key component of the Deeplabv3 network. It includes multiple parallel dilated convolution branches, each with a different dilation rate. These branches extract features from the input image at different receptive fields and merge the results. This allows the network to simultaneously capture contextual information of vascular images at different scales, increasing the network's perceptual power and improving the model's resolution of vascular contour boundaries and details.

[0099] This section comprises multiple branches, three of which are structurally identical modules (designated as Module 3 in the diagram). These modules mainly consist of convolutional layers (the kernel size can be adjusted according to the dilation rate; in this invention, they are 3×3, 5×5, and 7×7), batch normalization layers, and linear rectified function layers. Another branch consists of an adaptive two-dimensional average pooling layer, Module 1, and a linear rectified function layer. The outputs from these branches are then processed through a stitching layer, Module 1, a random deactivation layer, Module 3, and a 1×1 convolutional layer to obtain the final image segmentation result.

[0100] The FCN Head, an auxiliary head, is another part of the decoder. In this model, it performs tasks in parallel with the main segmentation head but does not output results. It contains multiple 1×1 convolutional layers to adjust the number of channels to the desired number of classes, followed by batch normalization layers, rectified linear activation functions (CRMs), and random deactivation layers (one 1×1 convolutional layer, one normalization layer, and one CRM layer constitute a module, named Module 1 in the figure). During training, the difference between the auxiliary head's output and the true label is used to calculate the loss, which is then combined with the loss from the main segmentation head. This multi-segmentation head training strategy helps the network learn the semantic information of the image better. Simultaneously, it provides an auxiliary supervision signal, facilitating faster gradient propagation and mitigating the vanishing gradient problem. Furthermore, it can fully utilize the high-resolution features of intermediate layers to improve model performance, especially in terms of blood vessel edges and details.

[0101] S5. Input the enhanced image into the HAMU-Net blood vessel segmentation model to obtain a blood vessel segmentation image.

[0102] In one feasible implementation, this invention addresses the problems of low contrast between vascular lines and other tissues, and unclear vascular contours in near-infrared vascular images. By effectively combining the advantages of traditional image enhancement methods and deep learning image enhancement methods, the processing effect is more suitable for neural networks to learn vascular region features, significantly enhancing vascular structure. The proposed HAMU-Net vascular segmentation model can efficiently fuse global and local information and rationally utilize the correlation between pixels in the image. It solves the problem that the network cannot achieve accurate vascular image segmentation due to insufficient multi-scale feature learning under small sample conditions.

[0103] Optionally, the enhanced image is input into the HAMU-Net vessel segmentation model to obtain a vessel segmentation image, including:

[0104] The enhanced image is input to the encoder for feature extraction to obtain the feature map of the input image;

[0105] Based on the input image feature map, multi-dimensional feature capture is performed through the bottleneck layer to obtain the output feature map;

[0106] Based on the skip connection part, the output feature map is input into the decoder to generate an image and obtain an image of the forearm blood vessels;

[0107] Based on the forearm blood vessel image, image segmentation is performed through the output layer to obtain a blood vessel segmentation image.

[0108] In one feasible implementation, this invention introduces multiple attention modules as the feature extraction network for the blood vessel segmentation model based on the original U-Net, enabling the model to better focus on the overall distribution state and edge distribution information of blood vessels. The Deeplabv3 architecture containing the ASPP module is selected as the main segmentation head of the blood vessel segmentation model, allowing the network to better learn contextual information, effectively capture spatial and semantic relationships between features, and improve feature representation capabilities and information transmission effects.

[0109] Optionally, based on the input image feature map, multi-dimensional feature capture is performed through a bottleneck layer to obtain an output feature map, including:

[0110] The input image feature map is fed into the first branch for convolution to obtain the first attention feature map; based on the first attention feature map, the sigmoid function is used for data processing to obtain the first channel attention weights;

[0111] The input image feature map is fed into the second branch for convolution to obtain the second attention feature map; based on the second attention feature map, the sigmoid function is used for data processing to obtain the second channel attention weights;

[0112] The input image feature map is fed into the third branch for convolution to obtain the third attention feature map; based on the third attention feature map, the sigmoid function is used for data processing to obtain the third channel attention weights;

[0113] Based on the attention weights of the first, second, and third channels, the first, second, and third attention feature maps are weighted and fused to obtain the output feature map.

[0114] In one feasible implementation, assume the input feature map size is (C × H × W), i.e., height is H, width is W, and number of channels is C. The model performs parallel processing on the input feature map in three branches. The first branch pools the feature map along the channel dimension, preserving the spatial information of height × width, generating a channel-compressed feature map. A 7×7 convolution is then used to process the pooled feature map to obtain the first attention feature map. After passing through the sigmoid function, the first channel attention weights are generated.

[0115] The second branch pools the feature map in the width dimension, retaining the height × channel information, and generates a feature map with compressed width. The pooled feature map is then processed using a 7×7 convolution to obtain the second attention feature map. After passing through the Sigmoid function, the second channel attention weights are generated.

[0116] The third branch pools the feature map in the height dimension, retaining the width × channel information, and generates a highly compressed feature map. The pooled feature map is then processed using a 7×7 convolution to obtain the third attention feature map. After passing through the Sigmoid function, the third channel attention weights are generated.

[0117] The attention weights of the second and third branches are dimension-replaced, and the attention weights obtained from the three branches are multiplied element-wise with the input feature map to obtain the overall feature map. The feature maps of the three branches are then fused and averaged to obtain the final output feature map. The output feature map is calculated as follows: (10), (11), (12), (13):

[0118] (10)

[0119] (11)

[0120] (12)

[0121] (13)

[0122] Among them, T a Here, σ is the input feature map, σ is the Sigmoid function, p represents dimensionality permutation, and A... channel For the attention weights of the first branch, A width For the attention weights of the second branch, A height is the attention weight for the third branch, Avg represents averaging, and ﹒ represents element-wise multiplication.

[0123] This invention proposes a forearm vessel image segmentation method based on the HAMU-Net deep learning network. Through multiple effective segmentation network architectures and modules, including an encoder-decoder structure, a U-Net backbone network, HybridAttention, Triplet Attention, and Spatial Attention, as well as a Deeplabv3 segmentation head and an FCN auxiliary segmentation head, the HAMU-Net vessel segmentation network model possesses significant advantages such as multi-level feature extraction, multi-scale receptive fields, and high segmentation accuracy, thus improving the model's robustness and generalization ability in forearm vessel image segmentation tasks. This invention is an efficient and accurate near-infrared forearm vessel image segmentation method based on hybrid attention modules.

[0124] Figure 6 This is a block diagram of a forearm blood vessel image segmentation device based on a HAMU-Net deep learning network, according to an exemplary embodiment. The device is used in a forearm blood vessel image segmentation method based on a HAMU-Net deep learning network. (Refer to...) Figure 6 The device includes an image acquisition module 610, an image preprocessing module 620, an image enhancement module 630, a model building module 640, and an image segmentation module 650. Among them:

[0125] The image acquisition module 610 is used to acquire images of forearm blood vessels through a near-infrared acquisition device to obtain raw images.

[0126] The image preprocessing module 620 is used to preprocess the original image based on a Gaussian filter and according to a preset background grayscale threshold to obtain a processed image.

[0127] The image enhancement module 630 is used to perform multi-stage image enhancement processing based on the Reinforcement-CLAHE technology, according to the processed image, using a residual convolutional autoencoder to obtain an enhanced image;

[0128] Model building module 640 is used to build the HAMU-Net blood vessel segmentation model based on the Hybrid Attention module, Triplet Attention module, Spatial attention module and U-Net network structure;

[0129] The image segmentation module 650 is used to input the enhanced image into the HAMU-Net blood vessel segmentation model to obtain a blood vessel segmentation image.

[0130] Optionally, the image preprocessing module 620 is further used for:

[0131] Based on a preset background grayscale threshold, background noise is removed from the original image to obtain the first image;

[0132] The first image is input into a Gaussian filter to remove noise, thus obtaining the second image;

[0133] The second image is contrast-stretched to obtain the processed image.

[0134] Optionally, the image enhancement module 630 is further used for:

[0135] Based on the processed image, image enhancement processing is performed using a residual convolutional autoencoder to obtain the first enhanced image;

[0136] Based on the Reinforcement-CLAHE technique, the first enhanced image is subjected to image enhancement processing to obtain the second enhanced image;

[0137] The first enhanced image and the second enhanced image are weighted and fused to obtain the enhanced image.

[0138] The HAMU-Net vessel segmentation model includes the HAMU-Net backbone network, skip connection part, bottleneck layer and output layer.

[0139] The HAMU-Net backbone network includes an encoder and a decoder; the encoder includes nine 3×3 convolutional layers and four 2×2 downsampling layers; the decoder includes nine 3×3 convolutional layers, four 2×2 upsampling layers and one Spatialattention module.

[0140] The skip connection part includes three first skip connections and two second skip connections; the first skip connections introduce the Hybrid Attention module;

[0141] The bottleneck layer consists of one Triplet Attention module; the Triplet Attention module includes a first branch, a second branch, and a third branch;

[0142] The output layer includes a main segmentation head and an auxiliary segmentation head; the main segmentation head includes one 3×3 convolutional layer, one Deeplabv3 Head, and one 1×1 convolutional layer; the auxiliary segmentation head includes one 3×3 convolutional layer and one FCN Head.

[0143] Optionally, the image segmentation module 650 is further used for:

[0144] The enhanced image is input to the encoder for feature extraction to obtain the feature map of the input image;

[0145] Based on the input image feature map, multi-dimensional feature capture is performed through the bottleneck layer to obtain the output feature map;

[0146] Based on the skip connection part, the output feature map is input into the decoder to generate an image and obtain an image of the forearm blood vessels;

[0147] Based on the forearm blood vessel image, image segmentation is performed through the output layer to obtain a blood vessel segmentation image.

[0148] Optionally, the image segmentation module 650 is further used for:

[0149] The input image feature map is fed into the first branch for convolution to obtain the first attention feature map; based on the first attention feature map, the sigmoid function is used for data processing to obtain the first channel attention weights;

[0150] The input image feature map is fed into the second branch for convolution to obtain the second attention feature map; based on the second attention feature map, the sigmoid function is used for data processing to obtain the second channel attention weights;

[0151] The input image feature map is fed into the third branch for convolution to obtain the third attention feature map; based on the third attention feature map, the sigmoid function is used for data processing to obtain the third channel attention weights;

[0152] Based on the attention weights of the first, second, and third channels, the first, second, and third attention feature maps are weighted and fused to obtain the output feature map.

[0153] This invention proposes a forearm vessel image segmentation method based on the HAMU-Net deep learning network. Through multiple effective segmentation network architectures and modules, including an encoder-decoder structure, a U-Net backbone network, HybridAttention, Triplet Attention, and Spatial Attention, as well as a Deeplabv3 segmentation head and an FCN auxiliary segmentation head, the HAMU-Net vessel segmentation network model possesses significant advantages such as multi-level feature extraction, multi-scale receptive fields, and high segmentation accuracy, thus improving the model's robustness and generalization ability in forearm vessel image segmentation tasks. This invention is an efficient and accurate near-infrared forearm vessel image segmentation method based on hybrid attention modules.

[0154] Figure 7 This is a schematic diagram of the structure of a forearm blood vessel image segmentation device provided in an embodiment of the present invention, as shown below. Figure 7 As shown, the forearm blood vessel image segmentation device may include the above-mentioned Figure 6 The forearm blood vessel image segmentation device shown is based on the HAMU-Net deep learning network. Optionally, the forearm blood vessel image segmentation device 710 may include a first processor 2001.

[0155] Optionally, the forearm blood vessel image segmentation device 710 may also include a memory 2002 and a transceiver 2003.

[0156] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0157] The following is combined Figure 7 A detailed description of each component of the forearm blood vessel image segmentation device 710 is provided below:

[0158] The first processor 2001 is the control center of the forearm blood vessel image segmentation device 710. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0159] Optionally, the first processor 2001 can perform various functions of the forearm vascular image segmentation device 710 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0160] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 7 CPU0 and CPU1 are shown in the diagram.

[0161] In a specific implementation, as one embodiment, the forearm blood vessel image segmentation device 710 may also include multiple processors, for example... Figure 7 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0162] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0163] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the forearm blood vessel image segmentation device 710. Figure 7 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0164] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0165] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 7 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0166] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected to the interface circuit of the forearm blood vessel image segmentation device 710. Figure 7 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0167] It should be noted that, Figure 7 The structure of the forearm blood vessel image segmentation device 710 shown does not constitute a limitation on this router. Actual knowledge structure recognition devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0168] Furthermore, the technical effect of the forearm blood vessel image segmentation device 710 can be referred to the technical effect of the forearm blood vessel image segmentation method based on the HAMU-Net deep learning network described in the above method embodiments, and will not be repeated here.

[0169] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0170] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0171] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0172] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0173] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0174] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0175] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0176] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0177] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0178] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0179] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0180] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0181] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for segmenting forearm blood vessel images based on the HAMU-Net deep learning network, characterized in that, The method includes: Forearm vascular images are acquired using a near-infrared acquisition device to obtain raw images; Based on a Gaussian filter, the original image is preprocessed according to a preset background grayscale threshold to obtain a processed image; Based on the Reinforcement-CLAHE technique, a multi-stage image enhancement process is performed using a residual convolutional autoencoder on the processed image to obtain an enhanced image. The HAMU-Net vessel segmentation model is constructed based on the Hybrid Attention module, Triplet Attention module, Spatial attention module, and U-Net network structure. The HAMU-Net vessel segmentation model includes a HAMU-Net backbone network, a skip connection part, a bottleneck layer, and an output layer. The HAMU-Net backbone network includes an encoder and a decoder; the encoder includes nine 3×3 convolutional layers and four 2×2 downsampling layers; the decoder includes nine 3×3 convolutional layers, four 2×2 upsampling layers, and one spatial attention module. The skip connection portion includes three first skip connections and two second skip connections; the first skip connections introduce a Hybrid Attention module; The bottleneck layer is a Triplet Attention module; the Triplet Attention module includes a first branch, a second branch, and a third branch; The output layer includes a main segmentation head and an auxiliary segmentation head; the main segmentation head includes a 3×3 convolutional layer, a Deeplabv3 Head, and a 1×1 convolutional layer; the auxiliary segmentation head includes a 3×3 convolutional layer and an FCN Head. The enhanced image is input into the HAMU-Net blood vessel segmentation model to obtain a blood vessel segmentation image.

2. The forearm blood vessel image segmentation method based on HAMU-Net deep learning network according to claim 1, characterized in that, The method of preprocessing the original image based on a Gaussian filter and according to a preset background grayscale threshold to obtain a processed image includes: Based on a preset background grayscale threshold, background noise is removed from the original image to obtain a first image; The first image is input into a Gaussian filter for noise removal to obtain the second image; The second image is contrast-stretched to obtain the processed image.

3. The forearm blood vessel image segmentation method based on HAMU-Net deep learning network according to claim 1, characterized in that, The Reinforcement-CLAHE technique involves performing multi-stage image enhancement processing using a residual convolutional autoencoder on the processed image to obtain an enhanced image, including: Based on the processed image, image enhancement processing is performed using a residual convolutional autoencoder to obtain a first enhanced image; Based on the Reinforcement-CLAHE technique, the first enhanced image is subjected to image enhancement processing to obtain the second enhanced image; The first enhanced image and the second enhanced image are weighted and fused to obtain an enhanced image.

4. The forearm blood vessel image segmentation method based on HAMU-Net deep learning network according to claim 1, characterized in that, The step of inputting the enhanced image into the HAMU-Net vessel segmentation model to obtain a vessel segmentation image includes: The enhanced image is input into the encoder for feature extraction to obtain an input image feature map; Based on the input image feature map, multi-dimensional feature capture is performed through the bottleneck layer to obtain the output feature map; Based on the skip connection portion, the output feature map is input into the decoder to generate an image and obtain a forearm blood vessel image; Based on the forearm blood vessel image, image segmentation is performed using the output layer to obtain a blood vessel segmentation image.

5. The forearm blood vessel image segmentation method based on HAMU-Net deep learning network according to claim 4, characterized in that, The step of obtaining an output feature map by performing multi-dimensional feature capture through the bottleneck layer based on the input image feature map includes: The input image feature map is input into the first branch for convolution to obtain the first attention feature map; based on the first attention feature map, the sigmoid function is used for data processing to obtain the first channel attention weights; The input image feature map is input into the second branch for convolution to obtain the second attention feature map; based on the second attention feature map, the sigmoid function is used for data processing to obtain the second channel attention weights; The input image feature map is input into the third branch for convolution to obtain the third attention feature map; based on the third attention feature map, the sigmoid function is used for data processing to obtain the third channel attention weights; Based on the first channel attention weight, the second channel attention weight, and the third channel attention weight, the first attention feature map, the second attention feature map, and the third attention feature map are weighted and fused to obtain the output feature map.

6. A forearm blood vessel image segmentation device based on a HAMU-Net deep learning network, wherein the forearm blood vessel image segmentation device based on a HAMU-Net deep learning network is used to implement the forearm blood vessel image segmentation method based on a HAMU-Net deep learning network as described in any one of claims 1-5, characterized in that, The device includes: The image acquisition module is used to acquire images of forearm blood vessels using a near-infrared acquisition device to obtain raw images. The image preprocessing module is used to preprocess the original image based on a Gaussian filter and according to a preset background grayscale threshold to obtain a processed image. The image enhancement module is used to perform multi-stage image enhancement processing based on the Reinforcement-CLAHE technology, using a residual convolutional autoencoder to obtain an enhanced image; The model building module is used to build the HAMU-Net blood vessel segmentation model based on the Hybrid Attention module, Triplet Attention module, Spatialattention module and U-Net network structure; The image segmentation module is used to input the enhanced image into the HAMU-Net blood vessel segmentation model to obtain a blood vessel segmentation image.

7. A forearm blood vessel image segmentation device, characterized in that, The forearm blood vessel image segmentation device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 5.

Images