Alzheimer's disease prediction method and system based on biomics feature topological adjustment

Abstract

The invention relates to the technical field of deep learning and image data processing, in particular to an Alzheimer's disease prediction method and system based on dual-omics feature topological adjustment, and the method comprises the steps: dividing a brain region, respectively extracting radiomics features and morphological features, and correspondingly and respectively obtaining a radiomics feature matrix and a morphological node feature matrix; performing morphological correlation calculation among all brain regions based on the morphological node feature matrix to obtain an adjacent matrix; performing graph convolution processing on the adjacent matrix to obtain an enhanced topological matrix of each subject; performing sparse processing on the enhanced topological matrix to obtain a sparse topological matrix; and performing feature aggregation processing on the radiomics feature matrix for several times based on the sparse topology matrix to obtain a fusion feature matrix, performing flattening processing on the fusion feature matrix, and then performing feature extraction processing, layer normalization processing and classification processing in sequence to obtain an Alzheimer's disease prediction result. The accuracy of Alzheimer's disease prediction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of deep learning and image data processing, and specifically to an Alzheimer's disease prediction method and system based on bi-omics feature topological adjustment. Background Art

[0002] As an important research direction in the field of neuroscience, the study of structural brain networks aims to explore the anatomical connections of various brain regions and their roles in cognitive function, emotional regulation, motor control, and other aspects. In recent years, with the rapid development of neuroimaging technology, the study of structural brain networks has made significant progress. Structural magnetic resonance imaging (sMRI) has become an important tool for studying brain anatomical connections due to its high resolution and non-invasiveness. In particular, the brain structural connection network constructed using sMRI data has been widely used to explore the anatomical basis of the brain and provides an important theoretical basis for the prediction and evaluation of diseases such as Alzheimer's disease.

[0003] When constructing structural brain networks, morphological features, as an important biological feature, have become an important tool for studying brain anatomical structure. Morphological features directly reflect the size, shape, and local structural changes of brain regions. Providing a more intuitive representation of the anatomical connections between different brain regions helps to understand the structural associations between brain regions, thus providing a solid foundation for the construction of brain networks. In early brain network studies, the brain networks used were all constructed based on morphological features.

[0004] In recent years, with the introduction of imaging omics features, brain network construction based on imaging omics has gradually become an emerging research direction. By extracting a large number of quantitative features from sMRI images, such as texture, intensity and other information, imaging omics can reveal the microscopic changes and tissue heterogeneity of brain structure, which may be difficult to capture directly in morphological features. Therefore, imaging omics features have unique advantages in reflecting more subtle physiological connections between brain regions.

[0005] Morphological features and imaging omics features each have their own advantages in brain network construction, and there is a certain degree of complementarity. However, most of the current research on brain networks is based on a single omics, that is, only based on morphological features to build brain networks for optimization and research or only based on imaging omics features to build brain networks for optimization and research, without combining the advantages of the two omics to build and optimize brain networks. Summary of the invention

[0006] To solve the problems raised in the background technology, the present invention provides a method and system for predicting Alzheimer's disease based on topological adjustment of dual-omics features.

[0007] The technical solution of the present invention is as follows: A method for predicting Alzheimer's disease based on topological adjustment of bi-omics features, comprising the following steps: S1. Segmenting the acquired T1-weighted imaging data of the subject to obtain a gray matter volume map, dividing the gray matter volume map into several brain regions after preprocessing, extracting several imaging omics features of each brain region, and obtaining an imaging omics feature matrix based on the number of brain regions and imaging omics features; The T1-weighted imaging data were divided into brain regions, and several morphological features of each brain region were extracted. Based on the number of brain regions and morphological features, a morphological node feature matrix was obtained. S2. Calculate the morphological correlation between all brain regions based on the morphological node feature matrix to obtain an adjacency matrix; perform graph convolution on the adjacency matrix to obtain an enhanced topological matrix for each subject, where each element value in the enhanced topological matrix is ​​a connection strength value; S3, performing sparse processing on the enhanced topology matrix, the sparse processing process is: Calculate the average value of each connection strength value in the enhanced topological matrix of all subjects, obtain several distance values ​​according to the difference between different connection strength values ​​and the corresponding average value, set different contribution weights after sorting the several distance values, calculate the weighted average value of each connection strength value in the enhanced topological matrix based on the different contribution weights, perform sparse processing on the enhanced topological matrix based on the weighted average value, set the values ​​in the enhanced topological matrix that are less than the weighted average value to 0, and obtain a sparse topological matrix; S4. Based on the sparse topological matrix, the imaging genomics feature matrix is ​​subjected to several feature aggregation processes to obtain a fused feature matrix. The fused feature matrix is ​​flattened and then subjected to feature extraction, layer normalization and classification processes in sequence to obtain the prediction results of Alzheimer's disease.

[0008] Specifically, in S3, different contribution weights are set respectively after sorting a number of distance values, specifically: after sorting a number of distance values ​​from small to large, a first data set is constructed based on a first number of distance values ​​sorted first, and a first contribution weight is set for all distance values ​​in the first data set; a second data set is constructed based on the remaining number of distance values, and a second contribution weight is set for all distance values ​​in the second data set.

[0009] Furthermore, in S3, the weighted average value of each connection strength value in the enhanced topology matrix is ​​calculated based on different contribution weights, and the formula is expressed as: , in, is the weighted average, For the nth subject, in the enhanced topology matrix The connection strength value on For the first data set, For the second data set, is the first contribution weight, is the second contribution weight.

[0010] In S3, the enhanced topology matrix is ​​sparsely processed based on the weighted average value, and the values ​​in the enhanced topology matrix that are less than the weighted average value are set to 0 to obtain a sparse topology matrix, which is expressed as follows: , in, is the connection strength value in the sparse topological moment, is the weighted average.

[0011] Specifically, in S4, the imaging omics feature matrix is ​​subjected to several feature aggregation processes based on the sparse topological matrix to obtain a fused feature matrix. The process of feature aggregation process is as follows: Based on the connection strength value in the sparse topological matrix, the adjacent node features of each node feature in the radiomics feature matrix are aggregated to obtain the updated node features. The updated node features are sequentially subjected to feature normalization, nonlinear activation and loss processing to obtain the first fused feature matrix as the input for the next feature aggregation processing.

[0012] Specifically, in S2, the adjacency matrix is ​​subjected to graph convolution processing, and the process of graph convolution processing is as follows: The adjacency matrix is ​​normalized to obtain a first adjacency matrix. Based on the first adjacency matrix and the set weight matrix, the adjacency matrix is ​​sequentially processed with an activation function and a loss layer to obtain an enhanced topology matrix.

[0013] Furthermore, in S2, the morphological correlation between all brain regions is calculated based on the morphological node feature matrix, and the formula is expressed as: , in, is the morphological correlation between brain regions i and j, k is the sequence number of the morphological feature, ; is the value of the kth morphological feature in the ith brain region, is the average value of all morphological features in the i-th brain region, is the value of the kth morphological feature in the jth brain region, is the average value of all morphological features in the jth brain region.

[0014] Specifically, the several imaging genomics features in S1 include: brain area texture, intensity, and shape.

[0015] Specifically, the several morphological features in S1 include: brain region volume, surface area, thickness, and curvature.

[0016] The present invention also provides an Alzheimer's disease prediction system based on bi-omics feature topology adjustment, comprising: Feature extraction module: used to segment the acquired T1-weighted imaging data of the subject to obtain a gray matter volume map, divide the gray matter volume map into several brain regions after preprocessing, extract several imaging features of each brain region, and obtain an imaging feature matrix based on the number of brain regions and imaging features; The T1-weighted imaging data were divided into brain regions, and several morphological features of each brain region were extracted. Based on the number of brain regions and morphological features, a morphological node feature matrix was obtained. Feature enhancement module: used to calculate the morphological correlation between all brain regions based on the morphological node feature matrix to obtain the adjacency matrix; perform graph convolution processing on the adjacency matrix to obtain the enhanced topological matrix of each subject, and each element value in the enhanced topological matrix is ​​the connection strength value; Sparse processing module: used to perform sparse processing on the enhanced topology matrix, and the sparse processing process is: Calculate the average value of each connection strength value in the enhanced topological matrix of all subjects, obtain several distance values ​​according to the difference between different connection strength values ​​and the corresponding average value, set different contribution weights after sorting the several distance values, calculate the weighted average value of each connection strength value in the enhanced topological matrix based on the different contribution weights, perform sparse processing on the enhanced topological matrix based on the weighted average value, set the values ​​in the enhanced topological matrix that are less than the weighted average value to 0, and obtain a sparse topological matrix; Prediction module: It is used to perform several feature aggregation processes on the imaging genomics feature matrix based on the sparse topological matrix to obtain a fused feature matrix. After flattening the fused feature matrix, feature extraction, layer normalization and classification are performed in sequence to obtain the Alzheimer's disease prediction results.

[0017] The beneficial effects of the present invention are: 1. The present invention obtains an imaging omics feature matrix by extracting imaging omics features of different brain regions, extracts morphological features to obtain a morphological node feature matrix, obtains an adjacency matrix based on the morphological node feature matrix, and then performs graph convolution processing to obtain an enhanced topological matrix, further performs sparse processing on the enhanced topological matrix, calculates the weighted average of each connection strength value in the enhanced topological matrix based on different contribution weights, and sets the values ​​in the enhanced topological matrix that are less than the weighted average to 0, which helps to highlight important connections between brain regions and weaken irrelevant connections, thereby capturing key connections between brain regions, making the topological structure of the brain network clearer, and further improving the accuracy of predicting Alzheimer's disease.

[0018] 2. The present invention performs several feature aggregation processes on the imaging genomics feature matrix based on the sparse topological matrix to obtain a fused feature matrix. By combining morphological features with imaging genomics features, the deficiencies of morphological features in refined representation can be supplemented, and more comprehensive brain region connections and functional characteristics can be provided. By fusing the imaging genomics feature matrix with the topological relationship between brain regions, the complex relationship between brain regions can be captured, high-level feature representations can be extracted, and the prediction results of Alzheimer's disease can be further obtained, which significantly improves the ability to analyze the complex structure and functional connections of the brain, and has important scientific research and clinical application prospects. DETAILED DESCRIPTION

[0019] Exemplary embodiments of the present disclosure are described in more detail below.

[0020] This embodiment provides a method for predicting Alzheimer's disease based on bi-omics feature topology adjustment, comprising the following steps: S1. Segmenting the acquired T1-weighted imaging data of the subject to obtain a gray matter volume map, dividing the gray matter volume map into several brain regions after preprocessing, extracting several imaging omics features of each brain region, and obtaining an imaging omics feature matrix based on the number of brain regions and imaging omics features; The T1-weighted imaging data were divided into brain regions, and several morphological features of each brain region were extracted. Based on the number of brain regions and morphological features, a morphological node feature matrix was obtained.

[0021] In step S1, the acquired T1-weighted imaging data of the subject is segmented to obtain a gray matter volume map. After the gray matter volume map is subjected to deviation correction and registration, the registered gray matter volume map is divided into several brain regions according to automatic anatomical markers, and is divided into 68 brain regions in this embodiment. Then, several imaging genomics features of each brain region are extracted, for example, 25 imaging genomics features are extracted. Several imaging genomics features include: brain region texture, intensity, and shape.

[0022] Further based on the number of brain regions and the radiomics features, an radiomics feature matrix is ​​obtained. In this embodiment, a 68*25 radiomics feature matrix is ​​obtained.

[0023] It also includes dividing the brain into 68 brain regions based on the automatic anatomical labeling of T1-weighted imaging data. After a series of processes such as skull removal and motion correction, several morphological features of each brain region are extracted, for example, 9 morphological features are extracted. Several morphological features include: brain region volume, surface area, thickness, and curvature.

[0024] Further based on the number of brain regions and morphological features, a morphological node feature matrix is ​​obtained. In this embodiment, a 68*9 morphological node feature matrix is ​​obtained.

[0025] S2. Calculate the morphological correlations between all brain regions based on the morphological node feature matrix to obtain the adjacency matrix; perform graph convolution on the adjacency matrix to obtain an enhanced topological matrix for each subject, where each element value in the enhanced topological matrix is ​​a connection strength value.

[0026] The morphological correlation between all brain regions is calculated based on the morphological node feature matrix, and the formula is expressed as: , in, is the morphological correlation between brain regions i and j, k is the sequence number of the morphological feature, ; is the value of the kth morphological feature in the ith brain region, is the average value of all morphological features in the i-th brain region, is the value of the kth morphological feature in the jth brain region, is the average value of all morphological features in the jth brain region.

[0027] The adjacency matrix is ​​subjected to graph convolution. The process of graph convolution is as follows: The adjacency matrix is ​​normalized to obtain a first adjacency matrix, and based on the first adjacency matrix and the set weight matrix, the adjacency matrix is ​​sequentially processed with an activation function and a loss layer to obtain an enhanced topology matrix. The process formula is expressed as: , in, is the output of graph convolution processing, is the first adjacency matrix, is the adjacency matrix, is the weight matrix set, is the activation function used, Processing for loss layer.

[0028] S3, performing sparse processing on the enhanced topology matrix, the sparse processing process is: The average value of each connection strength value in the enhanced topological matrix of all subjects was calculated, and several distance values ​​were obtained according to the difference between different connection strength values ​​and the corresponding average values. After sorting the several distance values, different contribution weights were set respectively, and the weighted average value of each connection strength value in the enhanced topological matrix was calculated based on the different contribution weights. The enhanced topological matrix was sparsely processed based on the weighted average value, and the values ​​in the enhanced topological matrix that were less than the weighted average value were set to 0 to obtain a sparse topological matrix.

[0029] In step S3, the average value of each connection strength value in the enhanced topology matrix of all subjects is first calculated. , the formula is: , in, is the number of subjects, For the n subjects on the enhanced topological matrix The connection strength value on .

[0030] Then, according to the difference between different connection strength values ​​and the corresponding average value, several distance values ​​are obtained. , the formula is: , in, For the n subjects on the enhanced topological matrix The connection strength value on is the average of each connection strength value in the augmented topology matrix of all subjects.

[0031] After further sorting the distance values, different contribution weights are set respectively. Specifically, after sorting the distance values ​​from small to large, a first data set is constructed based on the first number of distance values ​​sorted first, and the contribution weights are set respectively. Indicates that a certain number of distance values ​​in the first order can be set to the distance values ​​in the first 20% after ordering. The first contribution weight is set for all distance values ​​in the first data set, using express; Construct a second dataset based on the remaining number of distance values, using Indicates that the second contribution weight is set for all distance values ​​in the second data set, using express.

[0032] The weighted average of each connection strength value in the enhanced topology matrix is ​​calculated based on different contribution weights. The formula is expressed as: , in, is the weighted average, For the nth subject, in the enhanced topology matrix The connection strength value on For the first data set, For the second data set, is the first contribution weight, is the second contribution weight.

[0033] With this weighting method, 20% of the distance values ​​in the first data set contribute 80% of the weight, and the remaining distance values ​​account for 80% and contribute 20% of the weight.

[0034] The enhanced topology matrix is ​​sparsely processed based on the weighted average value, and the values ​​in the enhanced topology matrix that are less than the weighted average value are set to 0 to obtain a sparse topology matrix. The formula is expressed as: , in, is the connection strength value in the sparse topological moment, is the weighted average.

[0035] S4. Based on the sparse topological matrix, the imaging genomics feature matrix is ​​subjected to several feature aggregation processes to obtain a fused feature matrix. The fused feature matrix is ​​flattened and then subjected to feature extraction, layer normalization and classification processes in sequence to obtain the prediction results of Alzheimer's disease.

[0036] Specifically, the imaging omics feature matrix is ​​subjected to several feature aggregation processes based on the sparse topological matrix to obtain a fused feature matrix. The process of feature aggregation processing is as follows: Based on the connection strength value in the sparse topological matrix, the adjacent node features of each node feature in the radiomics feature matrix are aggregated to obtain the updated node features. The updated node features are sequentially subjected to feature normalization, nonlinear activation and loss processing to obtain the first fused feature matrix as the input for the next feature aggregation processing.

[0037] The above process can be expressed as follows: , in, is the output after feature aggregation processing, is a sparse topological matrix, is the radiomics feature matrix, , are different weight matrices, is the nonlinear activation function used, Indicates lost processing.

[0038] The present invention also provides an Alzheimer's disease prediction system based on bi-omics feature topology adjustment, comprising: Feature extraction module: used to segment the acquired T1-weighted imaging data of the subject to obtain a gray matter volume map, divide the gray matter volume map into several brain regions after preprocessing, extract several imaging features of each brain region, and obtain an imaging feature matrix based on the number of brain regions and imaging features; The T1-weighted imaging data were divided into brain regions, and several morphological features of each brain region were extracted. Based on the number of brain regions and morphological features, a morphological node feature matrix was obtained. Feature enhancement module: used to calculate the morphological correlation between all brain regions based on the morphological node feature matrix to obtain the adjacency matrix; perform graph convolution processing on the adjacency matrix to obtain the enhanced topological matrix of each subject, and each element value in the enhanced topological matrix is ​​the connection strength value; Sparse processing module: used to perform sparse processing on the enhanced topology matrix, and the sparse processing process is: Calculate the average value of each connection strength value in the enhanced topological matrix of all subjects, obtain several distance values ​​according to the difference between different connection strength values ​​and the corresponding average value, set different contribution weights after sorting the several distance values, calculate the weighted average value of each connection strength value in the enhanced topological matrix based on the different contribution weights, perform sparse processing on the enhanced topological matrix based on the weighted average value, set the values ​​in the enhanced topological matrix that are less than the weighted average value to 0, and obtain a sparse topological matrix; Prediction module: It is used to perform several feature aggregation processes on the imaging genomics feature matrix based on the sparse topological matrix to obtain a fused feature matrix. After flattening the fused feature matrix, feature extraction, layer normalization and classification are performed in sequence to obtain the Alzheimer's disease prediction results.

[0039] Experimental content Experimental settings: The dataset was randomly divided into training set, validation set, and test set in a ratio of 8:1:1. The experimental development environment used was pytorch2.0.1. Training was performed on an NVIDIA GeForce RTX 4060 Laptop Gpu using the Adam optimizer with an initial learning rate of 3e. -4 , the batch size is set to 4.

[0040] The experimental results are shown in Table 1.

[0041] Table 1 Experimental results

[0042] Among them, TP represents the correctly predicted diseased individuals, TN represents the correctly predicted disease-free individuals, FP represents the incorrectly predicted diseased individuals (actually disease-free), and FN represents the incorrectly predicted disease-free individuals (actually diseased).

[0043] The accuracy in Table 1 is the proportion of individuals with and without disease that are correctly predicted, and the formula is as follows: , Sensitivity is the proportion of individuals with the disease that are correctly predicted, also known as the true positive rate, and the formula is as follows: , Specificity is the proportion of disease-free individuals correctly predicted, also known as the true negative rate, and the formula is as follows: .

[0044] The specific operation process is as follows: When only morphological features are considered, the T1-weighted imaging data of the subject is divided into brain regions as described in step S1 according to the automatic anatomical labeling, and the brain is divided into 68 brain regions. 9 morphological features are extracted from each brain region to obtain a 68*9 morphological node feature matrix, and the morphological correlation between every two brain regions is calculated to obtain a 68*68 adjacency matrix. Using the 68*9 morphological node feature matrix and the 68*68 adjacency matrix, several feature fusion processes are performed to obtain a fused feature matrix, and the fused feature matrix is ​​further flattened and then subjected to feature extraction, layer normalization, and classification processes in sequence to obtain the Alzheimer's disease prediction results. At the same time, the best performing parameters on the validation set are saved for actual Alzheimer's disease prediction.

[0045] When only imaging features are considered, the acquired T1-weighted imaging is segmented into gray matter volume maps. After deviation correction and registration, the registered gray matter volume maps are divided into 68 brain regions according to automatic anatomical markers. 25 corresponding imaging features are extracted from each brain region to obtain a 68*25 imaging feature matrix. Then, the imaging correlation between every two brain regions is calculated based on the imaging feature matrix to obtain a 68*68 adjacency matrix. The 68*25 imaging feature matrix and the corresponding 68*68 adjacency matrix are used to perform feature fusion several times to obtain a fused feature matrix. The fused feature matrix is ​​further flattened and then feature extraction, layer normalization and classification are performed in sequence to obtain the prediction results of Alzheimer's disease, and the best performing parameters of the validation set are saved.

[0046] It can be seen that the Alzheimer's disease prediction method with adaptive topological adjustment of dual-omics features provided by the present invention has high accuracy, sensitivity and specificity in the prediction of Alzheimer's disease.

Claims

1. A method for predicting Alzheimer's disease based on topological adjustment of bi-omics features, characterized in that: The following steps are involved: S1. Segmenting the acquired T1-weighted imaging data of the subject to obtain a gray matter volume map, dividing the gray matter volume map into several brain regions after preprocessing, extracting several imaging omics features of each brain region, and obtaining an imaging omics feature matrix based on the number of brain regions and imaging omics features; The T1-weighted imaging data were divided into brain regions, and several morphological features of each brain region were extracted. Based on the number of brain regions and morphological features, a morphological node feature matrix was obtained. S2, calculating the morphological correlation between all brain regions based on the morphological node feature matrix to obtain an adjacency matrix; Perform graph convolution on the adjacency matrix to obtain an enhanced topological matrix for each subject, where each element in the enhanced topological matrix is ​​a connection strength value. S3, performing sparse processing on the enhanced topology matrix, the sparse processing process is: Calculate the average value of each connection strength value in the enhanced topological matrix of all subjects, obtain several distance values ​​according to the difference between different connection strength values ​​and the corresponding average value, set different contribution weights after sorting the several distance values, calculate the weighted average value of each connection strength value in the enhanced topological matrix based on the different contribution weights, perform sparse processing on the enhanced topological matrix based on the weighted average value, set the values ​​in the enhanced topological matrix that are less than the weighted average value to 0, and obtain a sparse topological matrix; S4. Based on the sparse topological matrix, the imaging genomics feature matrix is ​​subjected to several feature aggregation processes to obtain a fused feature matrix. The fused feature matrix is ​​flattened and then subjected to feature extraction, layer normalization and classification processes in sequence to obtain the prediction results of Alzheimer's disease.

2. The Alzheimer's disease prediction method based on dual-omics feature topology adjustment according to claim 1, characterized in that: In S3, different contribution weights are set respectively after sorting a plurality of distance values, specifically: after sorting a plurality of distance values ​​from small to large, a first data set is constructed based on a first number of distance values ​​sorted first, and a first contribution weight is set for all distance values ​​in the first data set; a second data set is constructed based on the remaining number of distance values, and a second contribution weight is set for all distance values ​​in the second data set.

3. The Alzheimer's disease prediction method based on dual-omics feature topology adjustment according to claim 2, characterized in that: In S3, the weighted average value of each connection strength value in the enhanced topology matrix is ​​calculated based on different contribution weights, and the formula is expressed as: ， in, is the weighted average, For the nth subject, in the enhanced topological matrix The connection strength value on For the first data set, For the second data set, is the first contribution weight, is the second contribution weight.

4. The Alzheimer's disease prediction method based on bi-omics feature topology adjustment according to claim 1, characterized in that: In S3, the enhanced topology matrix is ​​sparsely processed based on the weighted average value, and the values ​​in the enhanced topology matrix that are less than the weighted average value are set to 0 to obtain a sparse topology matrix, which is expressed as follows: ， in, is the connection strength value in the sparse topological moment, is the weighted average.

5. The Alzheimer's disease prediction method based on bi-omics feature topology adjustment according to claim 1, characterized in that: In S4, the imaging omics feature matrix is ​​subjected to several feature aggregation processes based on the sparse topological matrix to obtain a fused feature matrix. The process of feature aggregation process is as follows: Based on the connection strength value in the sparse topological matrix, the adjacent node features of each node feature in the radiomics feature matrix are aggregated to obtain the updated node features. The updated node features are sequentially subjected to feature normalization, nonlinear activation and loss processing to obtain the first fused feature matrix as the input for the next feature aggregation processing.

6. The Alzheimer's disease prediction method based on bi-omics feature topology adjustment according to claim 1, characterized in that: In S2, the adjacency matrix is ​​subjected to graph convolution processing, and the process of graph convolution processing is as follows: The adjacency matrix is ​​normalized to obtain a first adjacency matrix. Based on the first adjacency matrix and the set weight matrix, the adjacency matrix is ​​sequentially processed with an activation function and a loss layer to obtain an enhanced topology matrix.

7. The Alzheimer's disease prediction method based on bi-omics feature topology adjustment according to claim 1, characterized in that: In S2, the morphological correlation between all brain regions is calculated based on the morphological node feature matrix, and the formula is expressed as: ， in, is the morphological correlation between brain regions i and j, k is the sequence number of the morphological feature, ; is the value of the kth morphological feature in the ith brain region, is the average value of all morphological features in the i-th brain region, is the value of the kth morphological feature in the jth brain region, is the average value of all morphological features in the jth brain region.

8. The method for predicting Alzheimer's disease based on bi-omics feature topology adjustment according to claim 1, characterized in that: The several imaging features in S1 include: brain area texture, intensity, and shape.

9. The method for predicting Alzheimer's disease based on bi-omics feature topology adjustment according to claim 1, characterized in that: The several morphological features in S1 include: brain region volume, surface area, thickness, and curvature.

10. An Alzheimer's disease prediction system based on bi-omics feature topology adjustment, characterized in that: include: Feature extraction module: used to segment the acquired T1-weighted imaging data of the subject to obtain a gray matter volume map, divide the gray matter volume map into several brain regions after preprocessing, extract several imaging features of each brain region, and obtain an imaging feature matrix based on the number of brain regions and imaging features; The T1-weighted imaging data were divided into brain regions, and several morphological features of each brain region were extracted. Based on the number of brain regions and morphological features, a morphological node feature matrix was obtained. Feature enhancement module: used to calculate the morphological correlation between all brain regions based on the morphological node feature matrix to obtain the adjacency matrix; Perform graph convolution on the adjacency matrix to obtain an enhanced topological matrix for each subject, where each element in the enhanced topological matrix is ​​a connection strength value. Sparse processing module: used to perform sparse processing on the enhanced topology matrix, and the sparse processing process is: Calculate the average value of each connection strength value in the enhanced topological matrix of all subjects, obtain several distance values ​​according to the difference between different connection strength values ​​and the corresponding average value, set different contribution weights after sorting the several distance values, calculate the weighted average value of each connection strength value in the enhanced topological matrix based on the different contribution weights, perform sparse processing on the enhanced topological matrix based on the weighted average value, set the values ​​in the enhanced topological matrix that are less than the weighted average value to 0, and obtain a sparse topological matrix; Prediction module: It is used to perform several feature aggregation processes on the imaging genomics feature matrix based on the sparse topological matrix to obtain a fused feature matrix. After flattening the fused feature matrix, feature extraction, layer normalization and classification are performed in sequence to obtain the Alzheimer's disease prediction results.

Images