OCT image segmentation method based on fast Fourier transform and SKAttention mechanism

Abstract

The invention discloses an OCT (Optical Coherence Tomography) image segmentation method based on fast Fourier transform and an SKAttention mechanism. According to the method, the image preprocessing module is introduced into the UNet network, namely the initial stage of the backbone structure, so that the segmentation precision and efficiency of the oedema region of the OCT eye fundus image are remarkably improved. The image preprocessing module combines FFT and SKAttention mechanisms, and the module is designed to carry out preliminary transformation and feature weighting processing on input OCT image data so as to strengthen the recognition capability of the network on edema features. According to the technical scheme, the feature processing flow is optimized, the recognition capability of the model on fundus lesion features is enhanced, and particularly, high efficiency and high accuracy are shown when a complex or low-contrast edema region is processed. Besides, the method has good adaptability and expandability while keeping high processing speed, and is suitable for large-scale medical image processing application.

Description

Technical Field

[0001] The present invention belongs to the technical fields of medical image processing, deep learning, and signal processing, and specifically relates to an image processing technology based on deep learning. Background Art

[0002] Existing technologies mainly use traditional image processing technology and basic deep learning models in the field of fundus edema segmentation in optical coherence tomography (OCT). Traditional methods include image segmentation using techniques such as threshold processing and region growing. These methods rely on manually set parameters and thresholds and lack flexibility and adaptability. With the development of deep learning technology, the UNet network model has been widely used in medical image segmentation tasks because of its effective feature capture capability and good context information integration capability.

[0003] However, although UNet has demonstrated excellent performance in multiple medical image processing tasks, existing technologies still face several technical challenges when processing OCT fundus images. First, OCT images are usually noisy and have low contrast due to the influence of light scattering and light absorption, which makes it difficult for traditional UNet models to accurately identify and segment subtle edema areas. Second, the existing UNet model lacks an attention mechanism for local areas of the image when processing features, which limits the model's segmentation accuracy in details.

[0004] The present invention is intended to solve the main technical problems existing in the prior art, namely, to improve the segmentation accuracy and contrast of edema areas in OCT fundus images. By introducing fast Fourier transform (FFT) to preprocess the original input image, the present invention can enhance image features in the frequency domain, reduce the impact of noise, and restore the spatial structure of the image through inverse transformation. In addition, the introduction of the SKAttention selective kernel attention mechanism (SelectKernelAttention mechanism) can make the network pay more attention to key edema areas, thereby providing a more accurate segmentation effect. This method that combines FFT and SKAttention provides a new solution to the limitations of the prior art, which can significantly improve the accuracy and efficiency of segmentation. Summary of the invention

[0005] In the processing of fundus optical coherence tomography (OCT) images, accurate edema segmentation is of great significance for the diagnosis and treatment of ophthalmic diseases. The existing technology often encounters technical challenges such as inconsistent image quality and difficulty in capturing details when dealing with such tasks. These limitations affect the accuracy and reliability of segmentation. To solve these problems, the present invention proposes an OCT image segmentation method based on fast Fourier transform and SKAttention mechanism, which aims to improve the accuracy and efficiency of edema area segmentation in OCT images.

[0006] The core of the present invention is to integrate a specific image preprocessing module before the UNet network model. The module first uses FFT to convert the original image data to the frequency domain. This step utilizes the high efficiency of FFT in image analysis, can highlight the structural information in the image, and provide a richer data basis for subsequent processing. Next, the adaptive SKattention is applied to recalibrate the features of the transformed data, and the response to key features is enhanced by dynamically adjusting the weights of each channel. Finally, an inverse fast Fourier transform is performed to convert the processed data back to its original space, keeping the size and structure of the image unchanged. The key is that this preprocessing module proves that the best results can only be achieved when applied in the initial part of the network. This discovery provides important guidance for the design and application of modules.

[0007] The present invention can effectively solve the problem of accurate segmentation of edema areas in complex and low-contrast OCT images. Experimental results show that compared with traditional processing methods, the present invention can significantly improve the accuracy and robustness of segmentation, especially when processing image details and low-contrast areas. Therefore, the present invention not only improves the technical level of medical image processing, but also may provide more reliable support for clinical diagnosis and patient management.

[0008] An OCT image segmentation method based on fast Fourier transform and SKAttention mechanism. The specific implementation process is as follows:

[0009] Step 1: Data acquisition and preprocessing.

[0010] Step 2: Preprocess the original input image by fast Fourier transform (FFT);

[0011] Step 3: Apply adaptive SKAttention to recalibrate the features of the transformed data, and enhance the response to key features by dynamically adjusting the weights of each channel.

[0012] Step 4: Inverse Fast Fourier Transform.

[0013] The frequency domain data processed by SKAttention, i.e. the final feature map Vc, is remapped to the spatial domain through inverse fast Fourier transform to restore the original spatial structure of the image.

[0014] Step 5: Network encoding and decoding processing.

[0015] Step 6: Train the OCT image segmentation model consisting of the image preprocessing module and the backbone structure.

[0016] Step 7: Use the trained model for inference, that is, predict the edema area on new fundus OCT images.

[0017] The image to be segmented is preprocessed in step 1, and then segmented using the trained OCT image segmentation model, and finally a segmentation result is generated, and an image predicted to be the edema part is output.

[0018] Furthermore, the specific method of step 1 is as follows:

[0019] The original OCT images obtained through the OCT system are first standardized and resized to meet the needs of subsequent processing. All images are uniformly processed to 512x512 pixels, RGB three channels. The fundus edema part on the image is annotated to generate a labeled image; all labeled images are stored in a 512x512 pixel, single-channel format, and each original OCT image corresponds to the annotated labeled image one by one.

[0020] Furthermore, the specific method of step 2 is as follows:

[0021] The standardized image data is converted to the frequency domain using the fast Fourier transform. The conversion formula is as follows:

[0022]

[0023] The symbols represent the following:

[0024] F(u,v): represents the image in the frequency domain, where u and v are the horizontal and vertical coordinates in the frequency domain respectively;

[0025] f(x,y): is the pixel value at position (x,y) in the spatial domain;

[0026] M and N: are the image dimensions, i.e. the width and height of the image;

[0027] e: is the base of natural logarithm, approximately equal to 2.71828;

[0028] i: is an imaginary unit, satisfying i 2 = -1;

[0029] ∑: represents the summation symbol, which is used to accumulate the transformation results corresponding to all spatial domain pixels;

[0030] This part expresses the relationship between frequency and spatial position, where and Represent the frequency components in the horizontal and vertical directions respectively.

[0031] The application of Fast Fourier Transform allows the model to operate in the frequency domain in order to capture both high- and low-frequency features in the image.

[0032] Furthermore, the specific method of step 3 is as follows:

[0033] After the image is converted to the frequency domain through the fast Fourier transform, the input feature map X is processed by two different convolution branches. In the split stage, the two convolution branches use 3x3 depth convolution and 3x3 void convolution respectively (the expansion rate is 2 to replace the 5x5 kernel convolution), and then generate feature maps through batch normalization and ReLU function (ReLU Activation) and

[0034] Then in the fusion stage, the outputs of the two branches are fused by element-wise addition to obtain the fused feature map U, and the channel-level statistics are generated by global average pooling. The cth element of is the eigenvector Then through the dimension reduction fully connected layer F fc Generate compact feature vectors

[0035]

[0036] The meanings of the symbols are as follows:

[0037] F gp : represents the global average pooling operation;

[0038] U c : represents the cth channel in the fused feature map;

[0039] H, W: represents the length and width of the image;

[0040] F fc : represents the function of the fully connected layer, which is a common layer in neural networks and is usually used to learn nonlinear combinations of input data;

[0041] represents channel-level statistics, which is used as the input feature vector of the dimension reduction fully connected layer. The cth element of is the eigenvector This can be the output of any layer from the previous part of the network, or some transformation of the original input data;

[0042] W: is the weight matrix in the fully connected layer. This matrix is ​​used to linearly transform the input feature s to a new feature space, usually to reduce the dimension or adjust the feature expression to better suit the task;

[0043] Bn: represents the batch normalization operation applied after the fully connected layer. Batch normalization is a technique used to improve the speed and stability of neural network training by normalizing the input of the layer to reduce internal covariate shift.

[0044] Finally, in the selection stage, cross-channel soft attention is used to adaptively select information of different spatial scales, which is achieved by using compact feature vectors. To guide the implementation. Where A, B∈R C×d , is a matrix defined in the real number field, used to guide the selection of different spatial scales of information. a and b represent and The soft attention vector of the feature map U e and U b The number of channels, d represents the compact feature vector For each channel c, the calculation is as follows:

[0045]

[0046] Here, A c and B c are the cth rows of matrices A and B respectively, and a c and b c are the cth elements of the soft attention vectors a and b respectively. The outputs of the two branches are weightedly fused using these weights to obtain the final feature map:

[0047]

[0048] In order to maintain the consistency and standardization of feature weighting, the a of each channel is designed to be c and b c The sum is 1. This ensures that the merging of information from the two branches is balanced, preventing one branch from contributing too much to the final output.

[0049] This attention mechanism allows the model to adaptively adjust its receptive field size according to the input content, optimize information flow and improve feature processing. The SKattention mechanism dynamically evaluates and adjusts the importance of frequency components, optimizes feature representation, and enhances the model's sensitivity to key areas such as edema.

[0050] Further, step 5 is specifically put as follows:

[0051] First, the image data after inverse fast Fourier transformation is subjected to basic preprocessing, including normalization and denoising, to ensure the quality and consistency of the input data. The preprocessed image will be input into the subsequent backbone structure, first passing through a contraction path, which consists of a four-layer encoder structure, and gradually reduces the spatial dimension through convolutional layers and pooling layers, while increasing the depth of the features. Then, the image is gradually restored to the size of the original image through an expansion path consisting of a four-layer decoder structure, which includes upsampling and convolution operations. In the expansion path, the features in the contraction path are combined with the features in the expansion path through jump connections to maintain the key information of the image. At the output of the network, the probability of each pixel belonging to the edema area can be obtained. The output is a segmentation map of the same size as the input image, and the value of each pixel represents the probability that the pixel belongs to the edema area.

[0052] Furthermore, the specific method of step 6 is as follows: During the training process, the OCT image segmentation model composed of the image preprocessing module and the backbone structure is trained by the training data annotated in step 1 to identify and predict the edema area in the unseen fundus OCT image. The training process selects a two-dimensional cross entropy loss function to evaluate the difference between the model output and the actual annotation, and continuously optimizes the model weights through back propagation, gradient descent, dynamic learning rate adjustment and regularization techniques. The image preprocessing module includes fast Fourier transform (FFT), SKAttention and inverse fast Fourier transform.

[0053] The beneficial effects of the present invention are as follows:

[0054] The present invention significantly improves the segmentation accuracy and efficiency of edema areas in OCT fundus images by introducing an image preprocessing module in the initial stage of the UNet network, i.e., the backbone structure. This technical solution optimizes the feature processing flow and enhances the model's ability to recognize fundus lesion features, especially when processing complex or low-contrast edema areas, showing high efficiency and accuracy. In addition, while maintaining a high processing speed, the method also has good adaptability and scalability, and is suitable for large-scale medical image processing applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a diagram illustrating the network location applicable to the module.

[0056] Figure 2 This is the workflow of the OCT image segmentation model according to an embodiment of the present invention.

[0057] Figure 3 The following is a specific working process of the image preprocessing module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0059] Glossary

[0060] OCT (Optical Coherence Tomography) is an imaging technology that uses the principle of light reflection to obtain high-resolution three-dimensional images of the inside of tissues by measuring the time delay and intensity of light reflected from tissues. OCT is widely used in ophthalmology, dermatology and other fields, especially in fundus examination, and can perform fine imaging of eye structures such as the retina.

[0061] FFT (Fast Fourier Transform): is an efficient algorithm for converting time or space signals into frequency or wave number signals. FFT is a fast implementation of Fourier transform and is commonly used in signal processing, image analysis, speech processing and other fields to analyze the existence and characteristics of different frequency components.

[0062] UNet: is a convolutional neural network structure designed specifically for biomedical image segmentation. The UNet structure adopts a symmetrical "U"-shaped design, which includes a contraction path for capturing contextual information and a symmetrical expansion path for precise positioning. Through this design, UNet can effectively process images of different scales while maintaining image details. It is widely used in the field of medical image processing, especially in the precise segmentation of tissues and cells.

[0063] Specifically, the present invention provides an improved OCT image segmentation model, which integrates a special preprocessing module at the beginning to improve the automatic segmentation effect of edema areas in fundus OCT images. This technology is particularly suitable for medical image processing tasks that require high-precision and high-efficiency fundus edema detection.

[0064] The core of the present invention is to introduce a module that combines FFT and SKAttention mechanisms. The module is designed to perform preliminary transformation and feature weighting processing on the input OCT image data to enhance the network's ability to recognize edema features. The specific implementation process of the embodiment is as follows:

[0065] Preprocessing stage: The public dataset OIMHS is used, which contains 10,908 pairs of image data as training data and 2,700 pairs of image data as validation test data. The original OCT images are first standardized and resized to adapt to the subsequent processing requirements. All images are uniformly processed to 512x512 pixels and RGB three channels.

[0066] Fast Fourier Transform: In this step, the normalized image data is converted to the frequency domain. The transformation formula is as follows: The symbols are as follows: F(u,v): represents the image in the frequency domain, where u and v are the horizontal and vertical coordinates in the frequency domain respectively; f(x,y): is the pixel value at position (x,y)(x,y)(x,y) in the spatial domain; M and N: are the dimensions of the image, i.e., the width and height of the image; e: is the base of the natural logarithm, approximately equal to 2.71828; i: is the imaginary unit, satisfying i 2 =-1; ∑: represents the summation symbol, which is used to accumulate the transformation results corresponding to all spatial domain pixels; This part expresses the relationship between frequency and spatial position, where and Represent the frequency components in the horizontal and vertical directions respectively. The application of fast Fourier transform enables the model to operate in the frequency domain in order to capture the high-frequency and low-frequency features in the image.

[0067] SKAttention mechanism: After the image is converted to the frequency domain through the fast Fourier transform, the input feature map X is processed by two different convolution branches. In the Split stage, 3x3 group or depth convolution and 3x3 dilated convolution (with a dilation rate of 2 instead of 5x5 convolution) are used to generate feature maps U and

[0068] Then in the Fuse stage, the outputs of the two branches are fused by element-wise addition. And through global average pooling, a single feature vector is obtained Then through the dimension reduction fully connected layer F fc Generate compact feature vector Z = F fc (s)Bn(σ(Ws)), where the symbols represent the following meanings: F gp : represents the global average pooling operation; U c : represents the cth channel in the feature map; H, W: represents the length and width of the image; F fc: represents the function of the fully connected layer, which is a common layer in neural networks and is usually used to learn nonlinear combinations of input data; s: represents the input feature vector of the fully connected layer, which can be the output of any layer from the previous part of the network, or some transformation of the original input data; W: is the weight matrix in the fully connected layer. This matrix is ​​used to linearly transform the input feature s to a new feature space, usually for dimensionality reduction or to adjust the feature expression to better suit the task; Bn: represents the batch normalization operation applied after the fully connected layer. Batch normalization is a technique used to improve the speed and stability of neural network training by normalizing the input of the layer to reduce internal covariate shift.

[0069] Finally, in the Select stage, the compact feature vector Calculate the soft attention weights for each branch and The meanings of the symbols are as follows: A, B∈R C×d , is a matrix defined in the real number field, used to guide the selection of different spatial scales of information. a and b represent and The soft attention vector of the feature map U e and U b The number of channels, d represents the compact feature vector Dimensions. c and B c are the cth rows of matrices A and B respectively, and a c and b c are the cth elements of the soft attention vectors a and b respectively. The outputs of the two branches are weightedly fused through these weights to obtain the final feature map This attention mechanism allows the model to adaptively adjust its receptive field size according to the input content, optimize information flow and improve feature processing. The SKAttention mechanism dynamically evaluates and adjusts the importance of frequency components, optimizes feature representation, and enhances the model's sensitivity to key areas such as edema.

[0070] Inverse Fast Fourier Transform: The frequency domain data processed by SKAttention is remapped to the spatial domain through inverse transformation to restore the original spatial structure of the image.

[0071] Network encoding and decoding processing: The transformed image data first undergoes basic preprocessing, including normalization and denoising, to ensure the quality and consistency of the input data. The preprocessed image is input into the backbone structure and first passes through a contraction path consisting of a four-layer encoder structure, which gradually reduces the spatial dimension through convolutional layers and pooling layers while increasing the depth of the features. Then, the image is gradually restored to the size of the original image through an expansion path consisting of a four-layer decoder structure, which includes upsampling and convolution operations. In the expansion path, the features in the contraction path are combined with the features in the expansion path through jump connections to maintain the key information of the image. At the output of the network, the probability of each pixel belonging to the edema area can be obtained. The output is usually a segmentation map of the same size as the input image, and the value of each pixel represents the probability that the pixel belongs to the edema area.

[0072] During the training process, at this stage, the labeled training data obtained in the preprocessing stage is used. These data contain fundus OCT images and their corresponding edema area annotations. With this data, you can train the model to identify and predict edema areas in unseen fundus OCT images. Training usually selects a two-dimensional cross entropy loss function to evaluate the difference between the model output and the actual annotation, and continuously optimizes the model weights through back propagation, gradient descent, dynamic learning rate adjustment, and regularization techniques. Specifically, the pixel value of the background pixel in this dataset is 0, and the pixel value of the edema part is 192. At the beginning of training, each edema pixel is assigned a probability value of 1, and the background part is assigned a probability value of 0. The corresponding probability value is output through the forward propagation of the model, and then the loss for each pixel is obtained through the two-dimensional cross entropy loss function. The loss calculated for all pixels of the entire image or batch is accumulated to obtain the final loss value. Through back propagation, the gradient of the loss relative to each model parameter is calculated. This gradient represents the sensitivity of the loss function to each parameter under the current model parameters. The optimizer adjusts the value of each parameter in the model according to the calculated gradient to optimize the model weight. Finally, by continuously iteratively updating the parameters, the minimum value of the loss function is gradually approached to achieve the optimization of the segmentation effect.

[0073] The trained model can be used for inference, that is, predicting the edema area on a new fundus OCT image. This process includes passing the new image through the same preprocessing, then segmenting it using the trained model, and finally generating a segmentation result, outputting an image predicted to be the edema part.

[0074] The core technology of the present invention is to introduce the fast Fourier transform to process the original input data. Through this transformation, the image data is converted to the frequency domain, so that the model can analyze and process image features at different frequency levels. This processing step is particularly important for improving the network's sensitivity to image details and segmentation accuracy, especially when processing medical imaging data with complex textures and noise. Compared with the prior art, this method uses information in the frequency domain to help better capture and utilize structural features in the image, thereby improving segmentation performance. At the same time, through experimental verification, using this technology at the beginning can achieve more obvious results than using it in other parts of the network.

[0075] The present invention applies the SKAttention mechanism after FFT processing, which is an adaptive attention regulation mechanism that can automatically adjust the attention weight according to different frequency responses. In deep learning, the general attention mechanism is directly added to the model to directly process the input data. The application of the method mentioned in the present invention enables the model to better mine the information in the frequency domain of the image, increasing the dimension of model processing. At the same time, compared with the ordinary attention mechanism, the selection of SKAttention not only focuses on the detailed information of the local area, but also can be adjusted and optimized according to the global information. This strategy significantly enhances the model's ability to recognize important features in the image, especially in the identification of edema areas in the image. After applying this module, the segmentation model can more accurately locate and identify the lesion area.

[0076] The above contents are further detailed descriptions of the present invention in combination with specific / preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, they can also make several substitutions or modifications to these described embodiments without departing from the concept of the present invention, and these substitutions or modifications should be regarded as belonging to the protection scope of the present invention.

[0077] Parts of the present invention that are not described in detail belong to the well-known technologies of those skilled in the art.

Claims

1. An OCT image segmentation method based on fast Fourier transform and SKAttention mechanism, characterized in that: The specific implementation process is as follows: Step 1: Data acquisition and preprocessing; Step 2: Preprocess the original input image by fast Fourier transform; Step 3: Apply adaptive SKAttention to recalibrate the features of the transformed data, and enhance the response to key features by dynamically adjusting the weights of each channel; Step 4: Inverse Fast Fourier Transform; The frequency domain data processed by SKAttention is the final feature map V c Re-mapping to the spatial domain through inverse fast Fourier transform to restore the original spatial structure of the image; Step 5: Network coding and decoding processing; Step 6: Train the OCT image segmentation model consisting of the image preprocessing module and the backbone structure; Step 7: Use the trained model for inference, that is, predict the edema area on the new fundus OCT image; The image to be segmented is preprocessed in step 1, and then segmented using the trained OCT image segmentation model, and finally a segmentation result is generated, and an image predicted to be the edema part is output.

2. The OCT image segmentation method based on fast Fourier transform and SKAttention mechanism according to claim 1, characterized in that: Step 1: Through the OCT system, the original OCT images are first standardized and resized to adapt to subsequent processing requirements; all images are uniformly processed to 512x512 pixels, RGB three channels; and the fundus edema part on the image is annotated to generate a labeled image; all labeled images are stored in 512x512 pixels, single-channel format, and each original OCT image corresponds one-to-one to the annotated labeled image.

3. The OCT image segmentation method based on fast Fourier transform and SKAttention mechanism according to claim 1, characterized in that: Step 2: The standardized image data is converted to the frequency domain using the fast Fourier transform. The conversion formula is as follows: The symbols are as follows: F(u,v): represents the image in the frequency domain, where u and v are the horizontal and vertical coordinates in the frequency domain respectively; f(x,y): is the pixel value at position (x,y) in the spatial domain; M and N: are the image dimensions, i.e. the width and height of the image; e: is the base of natural logarithm; i: is an imaginary unit, satisfying i 2 = -1; ∑: represents the summation symbol, which is used to accumulate the transformation results corresponding to all spatial domain pixels; This part expresses the relationship between frequency and spatial position, where and Represent the frequency components in the horizontal and vertical directions respectively.

4. The OCT image segmentation method based on fast Fourier transform and SKAttention mechanism according to claim 3, characterized in that: Step 3: After the image is converted to the frequency domain through the fast Fourier transform, the input feature map X is processed by two different convolution branches. In the segmentation stage, the two convolution branches use 3x3 depth convolution and 3x3 hole convolution respectively, and then generate feature maps through batch normalization and ReLU function. and Then in the fusion stage, the outputs of the two branches are fused by element-wise addition to obtain the fused feature map U, and channel-level statistics are generated by global average pooling. The cth element of is the eigenvector Then through the dimension reduction fully connected layer F fc Generate compact feature vectors The meanings of the symbols are as follows: F gp : represents the global average pooling operation; U c : represents the cth channel in the fused feature map; H, W: represents the length and width of the image; F fc : represents the function of the fully connected layer; represents channel-level statistics, which is used as the input feature vector of the dimension reduction fully connected layer. The cth element of is the eigenvector W: is the weight matrix in the fully connected layer; Bn: represents the batch normalization operation applied after the fully connected layer; Finally, in the selection stage, cross-channel soft attention is used to adaptively select information at different spatial scales, which is guided by the compact feature vector z; where A,B∈R C×d , is a matrix defined in the real number field, used to guide the selection of different spatial scales of information; a and b represent and The soft attention vector of e and U b The number of channels, d represents the compact feature vector For each channel c, the calculation is as follows: and Here, A c and B c are the cth rows of matrices A and B respectively, and a c and b c are the cth elements of the soft attention vectors a and b respectively; the outputs of the two branches are weightedly fused through these weights to obtain the final feature map: In order to maintain the consistency and standardization of feature weighting, the a of each channel is designed to be c and b c The sum is 1.

5. The OCT image segmentation method based on fast Fourier transform and SKAttention mechanism according to claim 1 or 4, characterized in that: Step 5 is as follows: First, the image data after inverse fast Fourier transformation is preprocessed, including normalization and denoising, to ensure the quality and consistency of the input data. The preprocessed image will be input into the subsequent backbone structure, first passing through a contraction path, which consists of a four-layer encoder structure, and gradually reduces the spatial dimension through convolutional layers and pooling layers, while increasing the depth of features. Then, the image is gradually restored to the size of the original image through an expansion path consisting of a four-layer decoder structure, which includes upsampling and convolution operations; in the expansion path, the features in the contraction path are combined with the features in the expansion path through jump connections to maintain the key information of the image; at the output of the network, the probability of each pixel belonging to the edema area can be obtained; the output is a segmentation map of the same size as the input image, and the value of each pixel represents the probability that the pixel belongs to the edema area.

6. The OCT image segmentation method based on fast Fourier transform and SKAttention mechanism according to claim 5, characterized in that: The specific method of step 6 is as follows: during the training process, the OCT image segmentation model composed of the image preprocessing module and the backbone structure is trained through the training data labeled in step 1 to identify and predict the edema area in the unseen fundus OCT image; the training process selects a two-dimensional cross entropy loss function to evaluate the difference between the model output and the actual annotation, and continuously optimizes the model weight through back propagation, gradient descent, dynamic learning rate adjustment and regularization technology; the image preprocessing module includes fast Fourier transform, SKAttention and inverse fast Fourier transform.

Images