Wavelet transform-based OCTA image segmentation method and coronary heart disease prediction system

Abstract

The application provides an OCTA image segmentation method based on wavelet transform and a coronary heart disease prediction system.In the aspect of OCTA image segmentation, the application adopts a deep learning network based on wavelet transform, which is composed of a DCNV4 encoder, a wavelet feature fusion path and a decoder.The network precisely captures image details through deformable convolution and multi-scale feature fusion of wavelet transform, and realizes high-precision segmentation of blood vessels and FAZ in the OCTA image.In the aspect of the prediction and diagnosis system, the system integrates various indicators of OCTA image blood vessels and FAZ and clinical text data to construct a coronary heart disease prediction and diagnosis model.The system can automatically and non-invasively complete early coronary heart disease prediction and diagnosis and generate a visual report, providing decision support for clinicians.The application solves the problems of insufficient segmentation details and low precision of existing OCTA image blood vessels and FAZ segmentation, and the problem of strong invasiveness in the coronary heart disease examination and prediction method.

Description

Technical Field

[0001] This invention relates to the field of smart medical systems technology, and in particular to a coronary heart disease prediction system based on vascular features of retinal OCTA images. Background Technology

[0002] Coronary artery disease (CAD) is a leading cause of death worldwide. Its pathogenesis is complex, and early, accurate prediction and diagnosis are crucial for prevention and treatment. While traditional diagnostic methods such as coronary angiography are considered the "gold standard," they are invasive, costly, and carry potential risks, making them unsuitable for large-scale screening. OCTA, as an emerging non-invasive imaging technology, can clearly present the three-dimensional structure of retinal vessels. Retinal vessels and the cardiovascular system share similarities in embryonic development, physiological function, and pathological changes, and vascular abnormalities can serve as early markers of systemic vascular diseases. However, the segmentation accuracy of vessels and the foveal avascular zone (FAZ) in OCTA images is relatively low, and currently, there is no technical solution to deeply integrate retinal vascular features from OCTA images with rich clinical textual data to systematically construct a CAD risk prediction model. Summary of the Invention

[0003] The technical problem to be solved by the present invention is as follows: The present invention provides a method and system for predicting and diagnosing coronary heart disease based on retinal OCTA image segmentation, which is used to solve the problems of insufficient detail and low accuracy of vascular and FAZ segmentation in current OCTA images, as well as the problems of high cost, long time consumption and strong invasiveness in coronary heart disease examination and prediction methods.

[0004] In order to solve the technical problems proposed by the present invention, the present invention proposes the following technical solutions:

[0005] In a first aspect, this invention proposes a method for segmenting blood vessels and FAZ in OCTA images based on wavelet transform, comprising:

[0006] Obtain the OCTA dataset and annotate the blood vessel information in the images;

[0007] Denoising and data augmentation preprocessing are performed on the dataset;

[0008] The preprocessed OCTA image can be input into a wavelet transform-based segmentation model to obtain the OCTA image blood vessel and FAZ segmentation results;

[0009] A deep learning network based on wavelet transform for OCTA image segmentation was constructed to improve the segmentation accuracy of blood vessels and FAZ in OCTA images. It consists of a DCNV4 encoder, a wavelet feature fusion module, and a decoder, as detailed below:

[0010] a) A DCNV4 encoder is used to extract multi-scale feature representations from the input OCTA image. By progressively increasing the number of channels in the convolutional layers, multi-scale features of the image are extracted, generating feature maps at different levels. The encoder includes a DCNV4 encoding module and a patch embedding module. Each DCNV4 encoding module is responsible for further feature extraction and refinement. Each layer contains:

[0011] ① Wavelet global feature extraction module: Comprehensive learning of global features to provide overall semantic information for subsequent feature extraction.

[0012] ②DCNv4 operation: Utilizing deformable convolution technology, the position of the convolution kernel can be adaptively adjusted to better capture local features in the image.

[0013] ③MLPLayer: Multilayer Perceptron (MLP) layer, used to further process feature maps and increase the non-linearity of the model.

[0014] ④ Normalization and activation functions: Use normalization and activation functions such as LayerNorm and GELU to stabilize the training process and improve model performance.

[0015] Finally, at the end of each DCNV4 encoding module, a patch embedding module is used to reduce the spatial resolution of the feature map using a downsampling layer, while increasing the number of channels, providing a higher-level representation for subsequent feature extraction.

[0016] b) A multi-scale wavelet feature fusion module enhances and fuses the feature maps extracted by the DCNV4 encoder at multiple scales through wavelet transform, thereby enhancing the detailed information of the features and improving segmentation accuracy. Specifically, it includes a wavelet feature enhancement module and a wavelet feature fusion module:

[0017] The wavelet feature enhancement module combines global and local feature enhancement to enhance the features at low and high frequencies respectively by using wavelet transform technology on the feature maps of each level extracted by the DCNV4 encoder.

[0018] The wavelet feature fusion module learns vascular features in OCTA images at multiple scales by fusing feature maps from two stages. Through upsampling and convolution operations, feature maps at different scales at each level are aligned and fused to generate an enhanced fused feature map. During feature fusion, convolutional layers are used to further extract and refine features from the fused feature map, enhancing its expressive power.

[0019] c) Decoder: The decoder progressively upsamples the fused feature map to restore it to the resolution of the original image, generating the final segmentation result.

[0020] Secondly, this invention proposes a coronary heart disease prediction and diagnosis system based on retinal OCTA image segmentation, comprising:

[0021] OCTA Image Acquisition Module: Acquires retinal OCTA images of the patient using professional OCTA scanning equipment, ensuring that the image quality meets the analysis requirements.

[0022] Blood vessel and FAZ segmentation module: Using the wavelet transform-based OCTA image blood vessel and FAZ segmentation method proposed in this invention, the wavelet transform-based image segmentation algorithm automatically segments the blood vessels and FAZ in the preprocessed OCTA image to generate blood vessel segmentation images and FAZ segmentation images.

[0023] The indicator analysis module extracts vascular indicators such as vessel diameter, vessel density, tortuosity, and fractal dimension from vessel segmentation images; and extracts indicators such as perimeter, area, and non-circularity index of the FAZ from FAZ segmentation images.

[0024] Clinical text data collection module: It connects to the hospital information system (HIS) and electronic medical record system (EMR) to automatically extract patients' basic physiological information (gender, age, height, weight, etc.), medical history records (history of hypertension, history of diabetes, smoking history, drinking history, etc.), and laboratory test data (four lipid items, blood glucose, creatinine, etc.). It also supports manual supplementary entry to ensure the comprehensiveness and accuracy of the data.

[0025] Data fusion module: It integrates the extracted vascular indicators, FAZ indicators and clinical text data to construct a feature vector containing multi-source information.

[0026] Predictive model building and training module: Integrates the processed data to construct a fused feature vector. Machine learning or deep learning algorithms such as random forest, support vector machine, and deep neural network are selected. Using massive historical data from diagnosed coronary heart disease patients and healthy individuals as the training set, the model is trained, validated, and optimized to determine the optimal model parameters, enabling the model to possess powerful coronary heart disease risk prediction capabilities.

[0027] Risk assessment and output module: Input the patient data to be tested into the trained prediction model, and output the probability value of the risk of coronary heart disease after model calculation. According to the preset threshold, it is divided into low risk, medium risk, high risk and confirmed coronary heart disease level, and generates a visual report containing risk assessment results, risk factor analysis and personalized prevention suggestions, which is displayed to clinicians or patients through the system interface.

[0028] Results Display and Output Module: Displays patients' coronary heart disease risk prediction results in an intuitive way (such as charts, text reports, etc.) to provide decision support for clinicians.

[0029] Thirdly, the present invention provides an OCTA image segmentation device based on wavelet transform, comprising:

[0030] The data acquisition module is used to acquire OCTA image datasets labeled with blood vessels and avascular areas of the macula;

[0031] The preprocessing module is used to perform noise reduction and data augmentation preprocessing on the dataset;

[0032] An encoding module is used to extract multi-scale feature representations from the preprocessed OCTA image using a DCNV4 encoder. The DCNV4 encoder includes multiple DCNV4 encoding modules connected in sequence. Each DCNV4 encoding module contains a wavelet global feature extraction unit, a DCNv4 deformable convolutional unit, a multilayer perceptron unit, and a normalization and activation function unit. The DCNV4 encoder also includes a patch embedding module for downsampling the feature map output by the DCNV4 encoding module.

[0033] The feature fusion module is used to perform wavelet transform enhancement and cross-scale fusion on the multi-scale feature map extracted by the coding module through the multi-scale wavelet feature fusion module.

[0034] The decoding and segmentation module is used to upsample the fused feature map to the original image resolution through the decoder, generating segmentation results for blood vessels and avascular regions of the macula.

[0035] Secondly, the present invention proposes an electronic system comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, characterized in that the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the present invention.

[0036] Meanwhile, the present invention proposes a computer-readable storage medium storing computer instructions for causing the computer to execute the method described in the present invention.

[0037] The present invention adopts the above technical solution and has the following technical effects compared with the prior art:

[0038] 1. This invention achieves higher-precision segmentation of blood vessels and FAZ in OCTA images: By constructing a deep learning segmentation network that integrates wavelet transform and deformable convolution, it effectively overcomes the problem of insufficient segmentation accuracy of traditional methods at complex vascular structures and weak boundaries. The deformable convolutional units in the DCNV4 encoder can adaptively focus on vascular details, while the multi-scale wavelet feature fusion module significantly improves the accuracy of identification and segmentation of small vascular branches and blurred boundaries of avascular areas in the macula by enhancing and fusing low-frequency global information and high-frequency detail information respectively, providing a more reliable foundation for subsequent quantitative analysis.

[0039] 2. This invention constructs a non-invasive and efficient early risk prediction system for coronary heart disease: This invention deeply integrates high-precision segmented retinal vascular morphological indicators (such as vessel density, tortuosity, and fractal dimension) and FAZ morphological indicators with multi-dimensional clinical text data from patients. This approach overcomes the limitations of traditional coronary heart disease prediction, which relies on invasive examinations or single information sources. By using machine learning / deep learning models to uncover the deep correlation between changes in fundus microvessels and overall cardiovascular health, it achieves truly non-invasive, rapid, and low-cost early screening and assessment of coronary heart disease risk.

[0040] 3. This invention enhances the robustness and discriminative power of feature representation: In terms of network structure, the synergy between wavelet global feature extraction units and DCNv4 deformable convolutional units effectively unifies global semantic context and local geometric deformation features. At the system level, a data fusion module complementarily integrates radiomics features and clinical phenotypic data, constructing a more comprehensive and discriminative integrated feature vector. This results in a prediction model with stronger generalization ability and higher reliability for clinical auxiliary diagnosis.

[0041] In summary, this invention integrates high-precision image segmentation, multi-source information fusion, intelligent risk prediction, and visual report generation into a complete system. This system can automatically execute the entire process from image input to risk assessment report output, greatly improving clinical work efficiency and providing intuitive, quantitative decision support tools, thus contributing to the popularization and application of OCTA technology in the early screening of coronary heart disease. Attached Figure Description

[0042] Figure 1 This is a diagram of the system framework for the prediction and diagnosis of coronary heart disease based on wavelet transform OCTA image segmentation.

[0043] Figure 2 This is a diagram of the wavelet transform OCTA image segmentation model of the present invention.

[0044] Figure 3 This is a flowchart of the training process for the segmentation model and the prediction and diagnosis model of this invention.

[0045] Figure 4 This is a diagram showing the OCTA segmentation effect based on wavelet transform in this invention. Detailed Implementation

[0046] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0047] like Figure 1 The diagram shows a flowchart of a method for predicting and diagnosing coronary artery disease based on wavelet transform OCTA image segmentation. It specifically includes the following steps:

[0048] Step 1: Data Acquisition: Upon patient visit, the OCTA image acquisition and clinical text data collection process is initiated simultaneously. OCTA scanning is performed according to standardized operating procedures to ensure image quality; clinical text data collection covers relevant information from the patient's past and current visits, ensuring data timeliness.

[0049] Step 2, Data Preprocessing: The acquired OCTA images are transmitted to the image processing submodule, where image denoising, grayscale correction, and other preprocessing operations are performed sequentially to highlight the main vascular structures and remove interfering information. Clinical text data undergoes text cleaning to remove irrelevant characters and duplicate information, and to unify the data format and encoding rules, providing clean data for subsequent processing.

[0050] Step 3: OCTA Image Segmentation: Input the patient's OCTA image into a trained wavelet transform-based segmentation model to segment the blood vessels and FAZ regions. Calculate indicators such as blood vessel diameter, blood vessel density, tortuosity, fractal dimension, perimeter, area, and non-circularity index of the FAZ based on the segmentation results.

[0051] Step 4: Model Prediction: Load the pre-trained coronary heart disease risk prediction model, standardize the collected indicators and input them into the model. Based on the complex internal mathematical mapping relationship, the model quickly calculates the probability value of the patient's risk of developing coronary heart disease, ensuring real-time performance and efficiency.

[0052] Step 5: Result Assessment and Feedback: The risk probability value is converted into an easily understandable risk level by the risk assessment rule engine. Combined with the patient's risk factor profile, personalized prevention recommendations are generated, such as lifestyle adjustments and drug interventions. The assessment results and recommendations are displayed intuitively through the system interface in the form of charts and text, and a PDF report can be generated for download and printing, assisting clinical decision-making and doctor-patient communication.

[0053] Example 1: As Figure 3 The diagram shows the flowchart for training the segmentation model and the prediction / diagnosis model during the preparation phase, including:

[0054] Step 1: Dataset Creation: OCTA images and clinical text data from 100 patients with coronary artery disease and an equivalent number of healthy individuals were collected. Experienced retinal specialists manually annotated the blood vessels and FAZ in the OCTA images using LabelMe software, ensuring the accuracy and reliability of the data annotation.

[0055] Step 2: Dataset Preprocessing: Denoising and data augmentation are performed on the OCTA image dataset. The clinical text dataset is cleaned, structured, and standardized. The dataset is randomly divided into training, validation, and test sets in an 8:1:1 ratio.

[0056] Step 3: Segmentation Model Construction: Construct a deep learning network for OCTA image segmentation based on wavelet transform, referring to... Figure 2 As shown, it consists of a DCNV4 encoder, a wavelet feature fusion path, and a decoder, and is designed to improve the segmentation accuracy of blood vessels and FAZ in OCTA images.

[0057] Step 4: Training the OCTA image segmentation model based on wavelet transform: This invention was performed on a workstation equipped with an NVIDIA-RTX4070 GPU, using Python as the programming language and the PyTorch MMSegmentation deep learning framework. Adam was used as the optimizer for model training, with a batch size of 4 and a base learning rate of 0.005. OCTA images from the training data were input into the segmentation network for training, with 80,000 training iterations and validation performed every 2,000 iterations. The model parameters with the best training performance were saved. To verify the model's performance, it was tested on a test set. The fundus photography and OCTA images in the test set underwent the same preprocessing operations before being fed into the optimal segmentation weight network, outputting the predicted segmentation results, thus completing model training. Figure 4 The image shows the OCTA segmentation results based on wavelet transform. It can be seen that the segmentation method of this invention significantly improves the accuracy of identification and segmentation of small vascular branches and blurred boundaries of the avascular area of ​​the macula, providing a more reliable basis for subsequent quantitative analysis.

[0058] Step 5: Calculate vessel and FAZ indices in OCTA images: Use libraries such as OpenCV and NumPy to extract various indices from the segmented images of vessels and FAZ, including: vessel diameter, vessel density, tortuosity, fractal dimension and other vessel indices, and perimeter, area, non-circularity index, etc. of FAZ.

[0059] Step 6: Train the predictive diagnostic model: Train the extracted features using a Support Vector Machine (SVM) machine learning algorithm to construct a predictive model for the risk of adverse events in coronary heart disease. To further optimize the model, PSO (Particle Swarm Optimization) combined with ANN (Artificial Neural Networks) can be used. PSO optimizes the weights of the ANN by simulating the social behavior of flocks of birds or schools of fish, thereby improving the model's performance. To ensure the model's generalization ability, 10-fold cross-validation can be used to evaluate the results.

[0060] Example 2: This example proposes a system for predicting and diagnosing coronary heart disease based on retinal OCTA image segmentation, including:

[0061] OCTA Image Acquisition Module: Acquires retinal OCTA images of the patient using professional OCTA scanning equipment, ensuring that the image quality meets the analysis requirements.

[0062] Blood vessel and FAZ segmentation module: Using a wavelet transform-based image segmentation algorithm, the blood vessels and FAZ in the preprocessed OCTA image are automatically segmented to generate blood vessel segmentation images and FAZ segmentation images.

[0063] The indicator analysis module extracts vascular indicators such as vessel diameter, vessel density, tortuosity, and fractal dimension from vessel segmentation images; and extracts indicators such as perimeter, area, and non-circularity index of the FAZ from FAZ segmentation images.

[0064] Clinical text data collection module: It connects to the hospital information system (HIS) and electronic medical record system (EMR) to automatically extract patients' basic physiological information (gender, age, height, weight, etc.), medical history records (history of hypertension, history of diabetes, smoking history, drinking history, etc.), and laboratory test data (four lipid items, blood glucose, creatinine, etc.). It also supports manual supplementary entry to ensure the comprehensiveness and accuracy of the data.

[0065] Data fusion module: It integrates the extracted vascular indicators, FAZ indicators and clinical text data to construct a feature vector containing multi-source information.

[0066] Predictive model building and training module: Integrates the processed data to construct a fused feature vector. Machine learning or deep learning algorithms such as random forest, support vector machine, and deep neural network are selected. Using massive historical data from diagnosed coronary heart disease patients and healthy individuals as the training set, the model is trained, validated, and optimized to determine the optimal model parameters, enabling the model to possess powerful coronary heart disease risk prediction capabilities.

[0067] Risk assessment and output module: Input the patient data to be tested into the trained prediction model, and output the probability value of the risk of coronary heart disease after model calculation. According to the preset threshold, it is divided into low risk, medium risk, high risk and confirmed coronary heart disease level, and generates a visual report containing risk assessment results, risk factor analysis and personalized prevention suggestions, which is displayed to clinicians or patients through the system interface.

[0068] Results Display and Output Module: Displays patients' coronary heart disease risk prediction results in an intuitive way (such as charts, text reports, etc.) to provide decision support for clinicians.

[0069] Example 3: This example proposes an OCTA image segmentation device based on wavelet transform, comprising:

[0070] The data acquisition module is used to acquire OCTA image datasets labeled with blood vessels and avascular areas of the macula;

[0071] The preprocessing module is used to perform noise reduction and data augmentation preprocessing on the dataset;

[0072] An encoding module is used to extract multi-scale feature representations from the preprocessed OCTA image using a DCNV4 encoder. The DCNV4 encoder includes multiple DCNV4 encoding modules connected in sequence. Each DCNV4 encoding module contains a wavelet global feature extraction unit, a DCNv4 deformable convolutional unit, a multilayer perceptron unit, and a normalization and activation function unit. The DCNV4 encoder also includes a patch embedding module for downsampling the feature map output by the DCNV4 encoding module.

[0073] The feature fusion module is used to perform wavelet transform enhancement and cross-scale fusion on the multi-scale feature map extracted by the coding module through the multi-scale wavelet feature fusion module.

[0074] The decoding and segmentation module is used to upsample the fused feature map to the original image resolution through the decoder, generating segmentation results for blood vessels and avascular regions of the macula.

[0075] Example 4: This example proposes an electronic system, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method steps proposed in this invention.

[0076] Example 5: A computer-readable storage medium storing computer instructions for causing the computer to perform the method steps proposed in this invention.

[0077] It should be noted that the processing flow of embodiments 2-4 corresponds to the specific steps of the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0078] The computer program can be stored in a computer-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The computer-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the computer-readable medium includes, but is not limited to, the above-mentioned components.

[0079] It should be further explained that any one of the methods in the above embodiments can be stored in the computer-readable storage medium and loaded and executed on the processor to facilitate the storage and application of the above methods.

[0080] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0081] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0082] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An OCTA image segmentation method based on wavelet transform, characterized in that, Includes the following steps: Obtain the OCTA dataset with annotated blood vessels and avascular regions of the macula; The dataset is preprocessed with denoising and data augmentation. The preprocessed OCTA image is input into a wavelet transform-based segmentation model to obtain segmentation results for blood vessels and the avascular region of the macula. The segmentation model includes: A DCNV4 encoder is used to extract multi-scale feature representations from an input OCTA image. The DCNV4 encoder includes multiple DCNV4 encoding modules connected in sequence. Each DCNV4 encoding module includes a wavelet global feature extraction unit, a DCNv4 deformable convolution unit, a multilayer perceptron unit, and a normalization and activation function unit. The DCNV4 encoder also includes a patch embedding module for downsampling the feature map output by the DCNV4 encoding modules. A multi-scale wavelet feature fusion module is connected to the DCNV4 encoder and is used to perform wavelet transform enhancement and cross-scale fusion on the multi-scale feature maps extracted by the DCNV4 encoder. The decoder, connected to the multi-scale wavelet feature fusion module, is used to progressively upsample the fused feature map to restore it to the resolution of the original image and generate the final segmentation result.

2. The method according to claim 1, characterized in that, The wavelet global feature extraction unit is used to perform global context modeling on the input features, providing semantic information for subsequent feature extraction.

3. The method according to claim 1, characterized in that, The DCNv4 deformable convolutional unit can adaptively adjust the sampling position of the convolutional kernel according to the image content in order to capture local detail features in the image.

4. The method according to claim 1, characterized in that, The multi-scale wavelet feature fusion module includes: The wavelet feature enhancement submodule is used to decompose the feature map at each scale into low-frequency and high-frequency sub-bands using wavelet transform, and then enhance them separately. The wavelet feature fusion submodule is used to align and fuse enhanced feature maps of different scales.

5. The method according to claim 4, characterized in that, The wavelet feature fusion submodule aligns feature maps of different scales to the same resolution through upsampling and convolution operations before fusing them.

6. A coronary heart disease prediction system based on retinal OCTA image segmentation, characterized in that, include: The image acquisition module is used to acquire retinal OCTA images of the patient; The blood vessel and FAZ segmentation module is used to segment blood vessels and avascular areas of the macula in OCTA images, and it adopts the OCTA image segmentation method based on wavelet transform as described in any one of claims 1 to 5. The indicator analysis module is used to extract vascular morphology indicators and FAZ morphology indicators from the segmentation results; The clinical data collection module is used to acquire patients' clinical text data; The data fusion module is used to fuse the vascular morphology indicators, FAZ morphology indicators and clinical text data into a multi-source feature vector; The prediction module is used to output the risk assessment result of coronary heart disease based on the multi-source feature vector and the trained coronary heart disease risk prediction model. The report generation module is used to generate a visual report based on the risk assessment results.

7. The system according to claim 6, characterized in that, The prediction module employs at least one of the following model algorithms: random forest, support vector machine, or deep neural network.

8. An OCTA image segmentation device based on wavelet transform, characterized in that, include: The data acquisition module is used to acquire OCTA image datasets labeled with blood vessels and avascular areas of the macula; The preprocessing module is used to perform noise reduction and data augmentation preprocessing on the dataset; The encoding module is used to extract multi-scale feature representations from the preprocessed OCTA image using the DCNV4 encoder; The DCNV4 encoder includes multiple DCNV4 encoding modules connected in sequence. Each DCNV4 encoding module includes a wavelet global feature extraction unit, a DCNv4 deformable convolutional unit, a multilayer perceptron unit, and a normalization and activation function unit. The DCNV4 encoder also includes a patch embedding module for downsampling the feature map output by the DCNV4 encoding module. The feature fusion module is used to perform wavelet transform enhancement and cross-scale fusion on the multi-scale feature map extracted by the coding module through the multi-scale wavelet feature fusion module. The decoding and segmentation module is used to upsample the fused feature map to the original image resolution through the decoder, generating segmentation results for blood vessels and avascular regions of the macula.

9. An electronic system comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, characterized in that the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1-8.

10. A computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-8.

Images