Rabbit airway in-situ tumor image processing method

Abstract

The invention provides a rabbit airway in-situ tumor image processing method, relates to the field of image enhancement, and aims at improving the size of a traditional attention window by constructing a rabbit airway in-situ tumor image feature enhancement network and utilizing a dynamic window method, and enhancing the spatial feature extraction capability of an image by adaptively calculating the attention weight of the dynamic window. A BCDM model for extracting the global dependency relationship between the feature channels is provided, the global dependency relationship between the feature channels is improved and extracted, and the expression ability of channel features is enhanced; and in combination with the improved space attention weight and channel attention weight, carrying out weighted fusion on the space feature and the channel feature, thereby improving the enhancement effect of the image.

Description

Technical Field

[0001] The invention belongs to the field of image enhancement, and particularly relates to a method for processing rabbit airway in situ tumor images. Background Art

[0002] With the rapid development of medical imaging technology, especially the widespread application of small animal imaging equipment such as small animal CT and small animal MRI, early detection and treatment monitoring of tumors have become increasingly important. Rabbits, as an important experimental animal, have been widely used in tumor biology research, especially in the establishment of tumor models and drug screening. They play an irreplaceable role. The rabbit airway in situ tumor model is a commonly used model for studying lung diseases, tumor treatment methods and drug action mechanisms. However, the quality of rabbit airway in situ tumor images still faces some challenges. Although modern imaging technology can provide high-resolution images, it is often difficult to clearly display the details of the tumor and surrounding tissues due to the complexity of the tumor site, organ motion artifacts, and noise and low contrast problems during imaging. The lack of image quality seriously affects the accuracy of tumor diagnosis and the evaluation of treatment effects. Therefore, rabbit airway in situ tumor image enhancement technology is particularly important in medical image processing.

[0003] In terms of image feature extraction, spatial features and channel features are important components of the image. Spatial features determine the structural information of the image, while channel features reveal the image's properties such as color and texture. Traditional image processing methods are often unable to fully extract these two types of features and lack modeling of global dependencies between channels. In order to solve this problem, research in recent years has introduced adaptive mechanisms and attention mechanisms to optimize the feature extraction process. In particular, the combination of dynamic window methods and global dependency modeling methods has become an effective way to improve image feature expression capabilities and enhance image quality. Summary of the invention

[0004] The present invention provides a rabbit airway in situ tumor image processing method, which aims to enhance the spatial feature extraction capability of the image by constructing a rabbit airway in situ tumor image feature enhancement network, improving the size of the traditional attention window by using a dynamic window method, and adaptively calculating the attention weight of the dynamic window; a BCDM model for extracting the global dependency relationship between feature channels is proposed, the global dependency relationship between the extracted feature channels is improved, and the expression capability of the channel features is enhanced; and the spatial features and the channel features are weightedly fused in combination with the improved spatial attention weights and channel attention weights, so as to improve the image enhancement effect.

[0005] In order to achieve the above-mentioned object, the present invention provides the following technical solution: a method for processing rabbit airway in situ tumor images, comprising the following steps.

[0006] S1. Create a rabbit airway in situ tumor image dataset.

[0007] S2. Use the dynamic window method to improve the size of the traditional attention window, adaptively adjust the height and width of the window according to the feature gradient complexity, introduce the dynamic window self-attention in the self-attention calculation space, and combine multi-scale features to construct the tial-fea extraction block.

[0008] S3. A novel BCDM model is proposed to improve the extraction of global dependencies between feature channels. Specifically, the spatial position information is encoded into the channel relationship, and an adaptive adjustment factor is designed to dynamically adjust the importance of channel features.

[0009] S4. Dynamic channel weights are used to optimize the feature weighting method of traditional channel weights, and the cnel-acter extraction block is constructed in combination with the BCDM model.

[0010] S5. Introduce improved spatial attention weights and channel attention weights, and combine the tial-fea extraction block and the cnel-acter extraction block to build a Space-channel hybrid module.

[0011] S6. Construct a feature enhancement network for rabbit airway in situ tumor images. The first stage performs multi-scale enhancement, the second stage performs feature interaction, and the third stage fuses the global and local features of rabbit airway in situ tumor images.

[0012] S7. Construct a rabbit airway in situ tumor image enhancement model. The model enhances the features of the rabbit airway in situ tumor image through the rabbit airway in situ tumor image feature enhancement network, and outputs the enhanced rabbit airway in situ tumor image.

[0013] Preferably, in step S1, an animal-specific micro-bronchoscope is used to directly capture images of the tumor inside the rabbit airway, and a high-quality rabbit airway in situ tumor image dataset covering the three growth stages of the tumor, namely, early, middle and late stages, is produced. High-resolution rabbit airway in situ tumor images are used to generate low-resolution rabbit airway in situ tumor images, and a high-resolution image is used to generate low-resolution images of different sampling factors. Different degrees of Gaussian noise are added to the low-resolution rabbit airway in situ tumor images to simulate the influence of rabbit micro-motion during acquisition, and the processed dataset is divided into three parts: training set, validation set and test set according to the ratio of 5:2:3.

[0014] Preferably, in step S2, the tial-fea extraction block extracts the spatial features of the first rabbit airway in situ tumor image, and the specific steps are: S21. Input the first rabbit airway in situ tumor image features , are the height, width, and channels of the input image respectively. Divide The two parts perform self-attention calculation and multi-scale feature extraction respectively; S22, spatial dynamic window self-attention calculation first uses a 3×3 convolution operation to generate position information , It is a 3×3 convolution, which combines the position information with Add the elements to get the position weighted feature , spatial dynamic window self-attention according to the feature gradient complexity Adaptively adjust the window size to get the width of the dynamic window and high ,calculate Spatial self-attention within a window ; The feature gradient complexity The mathematical model is: ; In the formula, and is the gradient of the feature map in height and width; The width of the dynamic window and the height of the dynamic window The mathematical model is: ; ; In the formula, is the minimum and maximum value of the window width, is the minimum and maximum value of the window height, is the Sigmoid function, and the control factor d adaptively controls the steepness of the Sigmoid curve. is the gradient threshold; The mathematical model of the control factor d is: ; In the formula, is the global mean, for The mean of the eigenvalues ​​within the window, are the offsets that control the steepness of the Sigmoid curve, To prevent a small constant with a zero denominator; S23, multi-scale convolution layer uses convolution kernels of 3, 5, and 7 for parallel processing, and then adds them together to obtain multi-scale features ; S24, the spatial dynamic window self-attention and multi-scale features are integrated, the GELU activation function is used, and the residual connection is introduced. Add the elements to get the spatial characteristics of the rabbit airway in situ tumor image .

[0015] Preferably, in step S2, by introducing a dynamic window adaptive mechanism, the width and height of the dynamic window are calculated according to the feature gradient complexity, and the window parameters can be adjusted according to different needs. A small window is used in a high-complexity area and a large window is used in a low-complexity area, thereby improving the flexibility of attention calculation; introducing a control factor in the calculation process of the width and height of the dynamic window to adaptively control the steepness of the Sigmoid curve helps to improve adaptability, stability and accuracy; using convolution kernels of different sizes can better capture features of different scales and improve the model's ability to process complex information; combining dynamic window self-attention with feature extraction of different scales helps to improve the accuracy and flexibility of spatial feature extraction, especially when processing complex and changeable spatial data, which can significantly improve the adaptability and computational efficiency of the model.

[0016] Preferably, in step S3, the weighted rabbit airway tumor in situ image features are input ,The BCDM model is used to extract the global dependencies between feature channels and is defined as: ; In the formula, is the global dependency between feature channels, is the weighted feature input into the BCDM model, is the fusion weight, Q, K and V are the query, key and value of the image feature, is the weight matrix, is the flattened matrix, is the normalization factor, is the adaptive adjustment factor; The adaptive adjustment factor The mathematical model is: ; In the formula, is the fast Fourier transform; The fusion weight The mathematical model is: ; In the formula, is the eigenvalue of channel i at position (h,w), is the largest eigenvalue in channel i.

[0017] Preferably, in step S3, the BCDM model is used to extract the global dependencies between feature channels and encode the spatial position information into the channel relationship to capture the interaction between the spatial structure and the global dependencies; the adaptive adjustment factor uses fast Fourier transform to extract low-frequency information and strengthen the interaction between low-frequency and high-frequency features; and a learnable fusion weight is introduced to balance the contribution of the adaptive adjustment factor.

[0018] Preferably, in step S4, the cnel-acter extraction block extracts the channel features of the second rabbit airway in situ tumor image, and the specific steps are: S41. Input the second rabbit airway in situ tumor image features ,according to Dynamically generate channel weights , the channel weight and Multiply element by element to get the weighted features ; The channel weight The mathematical model is: ; In the formula, represents the weight matrix of the linear transformation, represents the global pooling layer, represents the bias vector; S42. The weighted features are input into the BCDM model. The output of the BCDM model captures local channel details through deep separable convolution, and then batch normalization and GELU activation function operations are performed to extract channel features. ; ; In the formula, is the batch normalization layer, is a depth-wise separable convolution, For the BCDM model, is the GELU activation function.

[0019] Preferably, in step S4, the cnel-acter extraction block is used to extract the channel features of the second rabbit airway in situ tumor image. The expression ability of the image features is effectively improved by introducing dynamic channel weights. The dynamically generated channel weights enable the channel features of each image to be adaptively adjusted according to changes in the input, thereby enhancing the correlation between the channel features and the image content. By using the BCDM model and the deep convolutional layer, the details of the channel features can be finely extracted. By combining batch normalization and GELU activation functions, the nonlinear modeling capability and training stability are further improved.

[0020] Preferably, in step S5, a Space-channel hybrid module is used to fuse the spatial features and the channel features, and the specific steps are as follows: Input spatial features of rabbit airway in situ tumor images extracted by tial-fea extraction block , calculate the improved spatial attention weights , input the channel features of the rabbit airway in situ tumor image extracted by the cnel-acter extraction block , calculate the improved channel attention weight , combined with the improved spatial attention weight and channel attention weight and Weighted and fused, the fused features are batch normalized and convolved to obtain the initial enhanced rabbit airway in situ tumor image features ; ; The improved spatial attention weights The mathematical model is: ; In the formula, and are average pooling and maximum pooling operations respectively, For feature fusion; The improved channel attention weight The mathematical model is: ; In the formula, is the eigenvalue of channel i.

[0021] Preferably, in step S5, the Space-channel hybrid module is mainly composed of a tial-fea extraction block and a cnel-acter extraction block, which is used to extract the image features of rabbit airway in situ tumors. By combining spatial features and channel features, the expression and enhancement effects of image features are improved; in terms of spatial feature extraction, the spatial attention weight is calculated by splicing the results after average pooling and maximum pooling with the original feature map, and a convolution operation is performed on it, thereby enhancing the expression ability of spatial features; in terms of channel feature extraction, the correlation of feature channels is further improved by combining the channel attention weight. Through feature weighting, the focus of the image can be adjusted more accurately, thereby highlighting key areas in image enhancement and reducing the interference of background noise, thereby improving image details and overall expressiveness.

[0022] Preferably, in step S6, a rabbit airway tumor in situ image feature enhancement network is used to achieve feature optimization, and the specific steps are: S61 input initial rabbit airway in situ tumor image features In the first stage, the input features are divided into three feature streams, and different convolution operations are performed in each feature stream to obtain , and : , , ; In the formula, For the Space-channel mixing module, is a 5×5 convolution, It is a 7×7 convolution; The enhanced multi-scale features are fused and the first-stage features are obtained using the Space-channel mixing module: ; S62, the second stage converts the output features of the first stage Divided into four feature blocks, , For feature partitioning, dynamic convolution is applied to each group of feature blocks separately; ; Where t is the convolution kernel dynamically generated by the global context information of the output features of the first stage, ; Introduce an interaction mechanism to connect different groups of features: , ; The cross-group interaction features are fused and the second-stage features are obtained using the Space-channel mixing module: ; S63. The third stage captures the global and local features of the output features of the second stage through deep feature expression, combines the global and local features and uses the Space-channel hybrid module to form the final enhanced features. : .

[0023] Preferably, in step S6, the rabbit airway tumor in situ image feature enhancement network gradually extracts, fuses and enhances image features in a multi-stage manner. The network is divided into three stages. In the first stage, the initial rabbit airway tumor in situ image features are divided into three feature streams, and multi-scale enhancement is performed in each feature stream to preliminarily enhance the features of the rabbit airway tumor in situ image. In the second stage, the features output by the first stage are further enhanced, and the features output by the first stage are divided into four feature blocks. An interaction mechanism is introduced to connect different groups of features, and the features of different groups are interacted to capture more details. In the third stage, the global pattern and local details of the image are captured through deep feature expression capabilities to obtain the final enhanced rabbit airway tumor in situ image features.

[0024] Compared with the prior art, the present invention has the following beneficial effects: The technical solution provided by the present invention is a rabbit airway in situ tumor image feature enhancement network, which uses a dynamic window method to improve the size of the traditional attention window, and enhances the spatial feature extraction ability of the image by adaptively calculating the attention weight of the dynamic window; proposes a BCDM model for extracting the global dependency relationship between feature channels, improves the global dependency relationship between extracted feature channels, and enhances the expression ability of channel features; combines the improved spatial attention weight and channel attention weight to perform weighted fusion of spatial features and channel features to improve the image enhancement effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of a rabbit airway in situ tumor image processing method provided by the present invention.

[0026] Figure 2 This is a structural diagram of the tial-fea extraction block provided by the present invention.

[0027] Figure 3 It is a structural diagram of the Space-channel mixing module provided by the present invention.

[0028] Figure 4 This is a network structure diagram of rabbit airway in situ tumor image feature enhancement provided by the present invention.

[0029] Figure 5 This is a diagram showing the image enhancement effect achieved by using a low-resolution rabbit airway in situ tumor image provided by the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the accompanying drawings of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] See also Figures 1 to 5 The present invention provides a method for processing rabbit airway in situ tumor images, which uses a dynamic window method to improve the size of a traditional attention window, and enhances the spatial feature extraction capability of an image by adaptively calculating the attention weight of a dynamic window; proposes a BCDM model for extracting global dependencies between feature channels, improves the global dependencies between extracted feature channels, and enhances the expression capability of channel features; combines the improved spatial attention weight and channel attention weight to perform weighted fusion of spatial features and channel features, thereby improving the image enhancement effect.

[0032] See also Figure 1 As shown, a method for processing rabbit airway in situ tumor images in an embodiment of the present application.

[0033] S1. Create a rabbit airway in situ tumor image dataset.

[0034] Furthermore, a micro bronchoscope for animals was used to capture 1,000 images of tumors inside rabbit airways, and a high-quality rabbit airway in situ tumor image dataset covering the early, middle, and late growth stages of the tumor was created. 1,000 high-resolution rabbit airway in situ tumor images were used to generate 3,000 low-resolution rabbit airway in situ tumor images, and one high-resolution image was used to generate three low-resolution images with different sampling factors. Different degrees of Gaussian noise were added to the low-resolution rabbit airway in situ tumor images to simulate the influence of rabbit micro-motion during acquisition. The processed dataset was divided into three parts: training set, validation set, and test set according to the ratio of 5:2:3, with 1,500 images in the training set, 600 images in the validation set, and 900 images in the test set.

[0035] S2. Use the dynamic window method to improve the size of the traditional attention window, adaptively adjust the height and width of the window according to the feature gradient complexity, introduce the dynamic window self-attention in the self-attention calculation space, and combine multi-scale features to construct the tial-fea extraction block.

[0036] Furthermore, if Figure 2 As shown, the tial-fea extraction block extracts the spatial features of the first rabbit airway in situ tumor image, and the specific steps are as follows.

[0037] S21. Input the first rabbit airway in situ tumor image features , The height, width, and channels of the input image are 640, 320, and 3 respectively. Divide The two parts are used for self-attention calculation and multi-scale feature extraction respectively.

[0038] S22, spatial dynamic window self-attention calculation first uses a 3×3 convolution operation to generate position information , It is a 3×3 convolution, which combines the position information with Add the elements to get the position weighted feature , spatial dynamic window self-attention according to the feature gradient complexity Adaptive adjustment of window size and feature gradient complexity The value range is 0.1 to 0.6, and the initial value is set to 0.4, which is the width of the dynamic window. and high , The value range is from 2 to 8. The value range is 4 to 16. Spatial self-attention within a window ; The feature gradient complexity The mathematical model is: ; In the formula, and is the gradient of the feature map in height and width, ranging from 0 to 255; The width of the dynamic window and the height of the dynamic window The mathematical model is: ; ; In the formula, is the minimum and maximum value of the window width, the minimum value is 2, and the maximum value is 8. The minimum and maximum values ​​of the window height are 4 and 16, respectively. The control factor d is a Sigmoid function that adaptively controls the steepness of the Sigmoid curve. The control factor d ranges from 0.3 to 0.5, and the initial value is set to 0.3. is the gradient threshold, set to 0.4; The mathematical model of the control factor d is: ; In the formula, is the global mean , for The mean of the eigenvalues ​​within the window , for The eigenvalue at the i-th row and j-th column in are the offsets that control the steepness of the Sigmoid curve, Set to 0.8, To prevent small constants with zero denominators, Set to .

[0039] S23, multi-scale convolution layer uses convolution kernels of 3, 5, and 7 for parallel processing, and then adds them together to obtain multi-scale features .

[0040] S24, the spatial dynamic window self-attention and multi-scale features are integrated, the GELU activation function is used, and the residual connection is introduced. Add the elements to get the spatial characteristics of the rabbit airway in situ tumor image .

[0041] S3. A novel BCDM model is proposed to improve the global dependency between feature channels. Specifically, the spatial position information is encoded into the channel relationship, and an adaptive adjustment factor is designed to dynamically adjust the importance of channel features.

[0042] Furthermore, the weighted rabbit airway tumor in situ image features are input ,The BCDM model is used for the global dependencies between feature channels and is defined as: ; In the formula, is the global dependency between feature channels, is the weighted feature input into the BCDM model, The fusion weight is set to 0.25 to 0.75 to balance the contribution of the adaptive adjustment factor. Q, K, and V are the query, key, and value of the image feature. is the weight matrix, is the flattened matrix, is the normalization factor, The value of is set to 64, is the adaptive adjustment factor, the value range is 0 to 1, and the initial value is set to 0.5; The adaptive adjustment factor The mathematical model is: ; In the formula, is the fast Fourier transform; The fusion weight The mathematical model is: ; In the formula, is the eigenvalue of channel i at position (h,w), is the largest eigenvalue in channel i.

[0043] S4. Dynamic channel weights are used to optimize the feature weighting method of traditional channel weights, and the cnel-acter extraction block is constructed in combination with the BCDM model.

[0044] Furthermore, the cnel-acter extraction block extracts the channel features of the second rabbit airway in situ tumor image, and the specific steps are as follows.

[0045] S41. Input the second rabbit airway in situ tumor image features ,according to Dynamically generate channel weights , the channel weight and Multiply element by element to get the weighted features ; The channel weight The mathematical model is: ; In the formula, represents the weight matrix of the linear transformation, represents the global pooling layer, Indicates that the bias vector ranges from -0.1 to 0.1, and the initial value is set to 0.

[0046] S42. The weighted features are input into the BCDM model. The output of the BCDM model captures local channel details through deep separable convolution, and then batch normalization and GELU activation function operations are performed to extract channel features. ; ; In the formula, is the batch normalization layer, is a depth-wise separable convolution, For the BCDM model, is the GELU activation function.

[0047] S5. Introduce improved spatial attention weights and channel attention weights, and combine the tial-fea extraction block and the cnel-acter extraction block to build a Space-channel hybrid module.

[0048] Furthermore, if Figure 3 As shown in the figure, the Space-channel hybrid module is used to fuse the spatial features and channel features. The specific steps are as follows: Input spatial features of rabbit airway in situ tumor images extracted by tial-fea extraction block , calculate the improved spatial attention weights , The value range is 0 to 1, and the channel features of the rabbit airway in situ tumor image extracted by the cnel-acter extraction block are input , calculate the improved channel attention weight , The value range is 0 to 1. and Added to 1, combined with the improved spatial attention weight and channel attention weight and Weighted and fused, the fused features are batch normalized and convolved to obtain the initial enhanced rabbit airway in situ tumor image features ; ; The improved spatial attention weights The mathematical model is: ; In the formula, and are average pooling and maximum pooling operations respectively, For feature fusion; The improved channel attention weight The mathematical model is: ; In the formula, is the eigenvalue of channel i.

[0049] S6. Construct a feature enhancement network for rabbit airway in situ tumor images. The first stage performs multi-scale enhancement, the second stage performs feature interaction, and the third stage fuses the global and local features of rabbit airway in situ tumor images.

[0050] Furthermore, if Figure 4 As shown, the rabbit airway in situ tumor image feature enhancement network is used to achieve feature optimization, and the specific steps are as follows.

[0051] S61 input initial rabbit airway in situ tumor image features In the first stage, the input features are divided into three feature streams, and different convolution operations are performed in each feature stream to obtain , and : , , ; In the formula, For the Space-channel mixing module, is a 5×5 convolution, It is a 7×7 convolution; The enhanced multi-scale features are fused and the first-stage features are obtained using the Space-channel mixing module: .

[0052] S62, the second stage converts the output features of the first stage Divided into four feature blocks, , For feature partitioning, dynamic convolution is applied to each group of feature blocks separately; ; Where t is the convolution kernel dynamically generated by the global context information of the input feature, and the initial values ​​of t are set to 3, 5, and 9 respectively. ; Introduce an interaction mechanism to connect different groups of features: , ; The cross-group interaction features are fused and the second-stage features are obtained using the Space-channel mixing module: .

[0053] S63. The third stage captures the global and local features of the output features of the second stage through deep feature expression, combines the global and local features and uses the Space-channel hybrid module to form the final enhanced features. : .

[0054] S7. Construct a rabbit airway in situ tumor image enhancement model. The model enhances the features of the rabbit airway in situ tumor image through the rabbit airway in situ tumor image feature enhancement network, and outputs the enhanced rabbit airway in situ tumor image.

[0055] Further, in step S7, a rabbit airway in situ tumor image enhancement model is provided, the model including input, initial feature extraction, rabbit airway in situ tumor image feature enhancement network, image reconstruction module and output, the rabbit airway in situ tumor image is input into the model, the model enhances the rabbit airway in situ tumor image features through the rabbit airway in situ tumor image feature enhancement network, the enhanced rabbit airway in situ tumor image features are used to reconstruct the rabbit airway in situ tumor image through the image reconstruction module, and the enhanced rabbit airway in situ tumor image is output.

[0056] Further, in step S7, a rabbit airway tumor in situ image enhancement model is developed using the Python language through the Pycharm application and implemented based on the Pytorch framework. The model is trained using 1500 training sets in the rabbit airway tumor in situ image dataset, and the training is started using a self-supervised learning method. The training results are compared with the validation set, and the model is tested with the test set. The model extracts spatial features and channel features in the image, enhances the features of the rabbit airway tumor in situ image from the spatial and channel directions, and constructs an enhanced rabbit airway tumor in situ image through the enhanced features.

[0057] Furthermore, if Figure 5 As shown, Figure 5 The middle left half is a low-resolution image of a rabbit airway tumor in situ. Figure 5 The middle right half is the enhanced rabbit airway in situ tumor image obtained after being processed by the rabbit airway in situ tumor image enhancement model, which significantly enhances the clarity and recognizability of the rabbit airway in situ tumor image.

[0058] The above are only preferred embodiments of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A rabbit airway in situ tumor image processing method, characterized in that: The following steps are involved: S1. Create a rabbit airway in situ tumor image dataset; S2. Use dynamic window method to improve the size of traditional attention window, adaptively adjust the height and width of the window according to the complexity of feature gradient, introduce dynamic window self-attention in self-attention calculation space, and construct tial-fea extraction block by combining multi-scale features; S3. A novel BCDM model is proposed to improve the global dependency between feature channels. Specifically, the spatial position information is encoded into the channel relationship, and an adaptive adjustment factor is designed to dynamically adjust the importance of channel features. S4, using dynamic channel weights to optimize the feature weighting method of traditional channel weights, and combining the BCDM model to build the cnel-acter extraction block; S5, introduce improved spatial attention weights and channel attention weights, and combine the tial-fea extraction block and the cnel-acter extraction block to build a Space-channel hybrid module; S6. Construct a rabbit airway tumor in situ image feature enhancement network. The first stage performs multi-scale enhancement, the second stage performs feature interaction, and the third stage fuses the global and local features of the rabbit airway tumor in situ image. S7. Construct a rabbit airway in situ tumor image enhancement model. The model enhances the features of the rabbit airway in situ tumor image through the rabbit airway in situ tumor image feature enhancement network, and outputs the enhanced rabbit airway in situ tumor image.

2. The rabbit airway in situ tumor image processing method according to claim 1, characterized in that: In the step S1, a high-quality rabbit airway in situ tumor image dataset covering the three growth stages of the tumor, namely, early stage, middle stage and late stage, is produced, a low-resolution rabbit airway in situ tumor image is generated using a high-resolution rabbit airway in situ tumor image, a high-resolution image is used to generate a low-resolution image with different sampling factors, different degrees of Gaussian noise are added to the low-resolution rabbit airway in situ tumor image to simulate the influence of the rabbit's micro-motion during acquisition, and the processed dataset is divided into three parts: a training set, a validation set and a test set.

3. The rabbit airway in situ tumor image processing method according to claim 2, characterized in that: In the step S2, the specific steps of the tial-fea extraction block are: S21. Input the first rabbit airway in situ tumor image features , are the height, width, and channels of the input image respectively. Divide The two parts perform spatial dynamic window self-attention calculation and multi-scale feature extraction respectively; S22, spatial dynamic window self-attention calculation first uses a 3×3 convolution operation to generate position information , It is a 3×3 convolution, which combines the position information with Adding elements , spatial dynamic window self-attention according to the feature gradient complexity Adaptively adjust the window size to get the width of the dynamic window and high ,calculate Spatial self-attention within a window ; The feature gradient complexity The mathematical model is: ； In the formula, and is the gradient of the feature map in height and width; The width of the dynamic window and the height of the dynamic window The mathematical model is: ； ； In the formula, is the minimum and maximum value of the window width, is the minimum and maximum value of the window height, is the Sigmoid function, and the control factor d adaptively controls the steepness of the Sigmoid curve. is the gradient threshold; The mathematical model of the control factor d is: ； In the formula, is the global mean, for The mean of the eigenvalues ​​within the window, To control the offset of the steepness of the Sigmoid curve, To prevent a small constant with a zero denominator; S23, multi-scale convolution layer uses convolution kernels of 3, 5, and 7 for parallel processing, and then adds them together to obtain multi-scale features ; S24, the spatial dynamic window self-attention and multi-scale features are integrated, the GELU activation function is used, and the residual connection is introduced. Add the elements to get the spatial characteristics of the rabbit airway in situ tumor image .

4. The rabbit airway in situ tumor image processing method according to claim 3, characterized in that: In the step S3, the weighted rabbit airway tumor in situ image features are input. , They are The height, width, and channel of the BCDM model are defined as: ； In the formula, is the global dependency between feature channels, is the fusion weight, Q, K and V are the weighted query, key and value of the rabbit airway in situ tumor image features, is the weight matrix, is the flattened matrix, is the normalization factor, is the adaptive adjustment factor; The adaptive adjustment factor The mathematical model is: ； In the formula, is the fast Fourier transform; The fusion weight The mathematical model is: ； In the formula, is the eigenvalue of channel i at position (h,w), is the largest eigenvalue in the channel.

5. The rabbit airway in situ tumor image processing method according to claim 4, characterized in that: In the step S4, the specific steps of cnel-acter extraction block are: S41. Input the second rabbit airway in situ tumor image features ,according to Dynamically generate channel weights , the channel weight and Multiply element by element to get the weighted features ; The channel weight The mathematical model is: ； In the formula, represents the weight matrix of the linear transformation, represents the global pooling layer, represents the bias vector; S42. The weighted features are input into the BCDM model. The output of the BCDM model captures local channel details through deep separable convolution, and then batch normalization and GELU activation function operations are performed to extract channel features. ; ； In the formula, is the batch normalization layer, is a depth-wise separable convolution, For the BCDM model, is the GELU activation function.

6. A rabbit airway tumor in situ image processing method according to claim 5, characterized in that: In the step S5, the specific steps of the Space-channel mixing module are: Input spatial features of rabbit airway in situ tumor images extracted by tial-fea extraction block , calculate the improved spatial attention weights , input the channel features of the rabbit airway in situ tumor image extracted by the cnel-acter extraction block , calculate the improved channel attention weight , combined with the improved spatial attention weight and channel attention weight and Weighted and fused, the fused features are batch normalized and convolved to obtain the initial enhanced rabbit airway in situ tumor image features ; ； The improved spatial attention weights The mathematical model is: ； In the formula, and are average pooling and maximum pooling operations respectively, is feature fusion; The improved channel attention weight The mathematical model is: ； In the formula, is the eigenvalue of channel i.

7. The rabbit airway in situ tumor image processing method according to claim 6, characterized in that: In step S6, the specific steps of the rabbit airway in situ tumor image feature enhancement network are as follows: S61 input initial rabbit airway in situ tumor image features In the first stage, the input features are divided into three feature streams, and different convolution operations are performed in each feature stream to obtain , and : ， ， ； In the formula, For the Space-channel mixing module, is a 5×5 convolution, It is a 7×7 convolution; The enhanced multi-scale features are fused and the Space-channel hybrid module is used to obtain the first-stage multi-scale enhanced features: ； S62, the second stage converts the output features of the first stage Divided into four feature blocks, , For feature partitioning, dynamic convolution is applied to each group of feature blocks separately; ； Where t is the convolution kernel dynamically generated by the global context information of the output features of the first stage, ; Introduce an interaction mechanism to connect different groups of features: ， ； The cross-group interaction features are fused and the second-stage features are obtained using the Space-channel mixing module: ； S63. The third stage captures the global and local features of the output features of the second stage through deep feature expression, combines the global and local features and uses the Space-channel hybrid module to obtain the final enhanced features. ; 。

Images