Magnetic resonance image analysis method for Alzheimer's disease
By constructing a gender-specific gray matter structure brain network through multivariate regression and graph theory analysis, we have solved the problem of gender differences affecting the accuracy of Alzheimer's disease magnetic resonance image analysis in existing technologies, and achieved higher analytical accuracy and reliability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG UNIV
- Filing Date
- 2026-01-04
- Publication Date
- 2026-04-14
AI Technical Summary
Existing machine learning models struggle to effectively distinguish gender differences when analyzing complex and variable brain network data from different patients, impacting the accuracy and reliability of Alzheimer's disease magnetic resonance imaging analysis.
Through multivariate regression analysis and graph theory analysis, we extracted gender-related and gender-specific gray matter structures from the brain network, constructed gender-mixed, female, and male magnetic resonance imaging analysis models, and used gender-specific gray matter structures as input features for machine learning models to perform image recognition and analysis.
It significantly improves the accuracy and reliability of magnetic resonance image analysis, reduces the impact of individual differences on the analysis results, and enhances the effectiveness of gender-specific analysis.
Smart Images
Figure CN121861002A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image analysis technology, and in particular to a magnetic resonance imaging analysis method for Alzheimer's disease. Background Technology
[0002] Alzheimer's disease exhibits a sex difference, with a higher incidence rate in women than men. This difference may be attributed to a number of complex physiological, pathological, and psychosocial factors, including women's longer life expectancy, hormonal fluctuations, cardiovascular risk factors, depression, sleep quality, insufficient physical activity, social roles, and social isolation. These factors may be related to different brain connectivity patterns. Specifically, the male brain tends to show greater intrahemispheric connectivity, while the female brain shows more interhemispheric connectivity. This difference in connectivity patterns may explain women's superior performance in social cognition and men's proficiency in sensorimotor skills and spatial information processing.
[0003] In the analysis of magnetic resonance imaging (MRI) images, various machine learning models play a crucial role by utilizing imaging data. Models such as logistic regression, support vector machines (SVM), k-nearest neighbor (KNN) algorithms, neural networks, and Naive Bayes analyze brain structure and functional information from neuroimaging techniques like MRI and PET. Logistic regression and SVM can identify biomarkers associated with Alzheimer's disease, such as brain atrophy, through classification algorithms, enabling early detection. The KNN algorithm analyzes cases by calculating similarity to known instances based on imaging features. Neural network models, especially deep learning models like convolutional neural networks (CNNs), can automatically learn and extract meaningful features from large datasets, demonstrating exceptional performance in handling complex medical imaging data. Naive Bayes models apply Bayes' theorem to combine prior knowledge with new imaging data to estimate the probability of disease presence.
[0004] Existing machine learning models primarily rely on in-depth analysis of patients' brain network data to provide analytical results. However, the brain network structures of different patients are complex and varied, exhibiting significant individual differences. Faced with such complex and variable brain network data, existing machine learning models often fall short in their analysis, affecting the accuracy and reliability of the results. Summary of the Invention
[0005] Therefore, it is necessary to provide a magnetic resonance imaging analysis method for Alzheimer's disease to address the aforementioned technical problems.
[0006] This invention provides a magnetic resonance imaging analysis method for Alzheimer's disease, comprising: Obtain a set of magnetic resonance imaging (MRI) images of Alzheimer's disease, consisting of MRI images of healthy men, healthy women, men with the disease, and women with the disease. Multivariate regression analysis was used to perform gender regression analysis on magnetic resonance images of different sexes but all in a healthy state, in order to extract multiple gray matter clusters related to gender differences and obtain a gender-related gray matter structure brain network. All gray matter clusters were converted into corresponding gray matter masks. Paired-samples T-tests were performed on magnetic resonance images of the same sex but different disease states within the gray matter masks to obtain the absolute gray matter volume of each gray matter mask. Pearson correlation analysis was performed on the absolute gray matter volume of all gray matter masks using graph theory analysis to obtain female-specific gray matter structure brain networks and male-specific gray matter structure brain networks. The machine learning model was trained by gender-related gray matter structure brain network, female-specific gray matter structure brain network and male-specific gray matter structure brain network respectively, resulting in gender-mixed magnetic resonance image analysis model, female magnetic resonance image analysis model and male magnetic resonance image analysis model. Magnetic resonance imaging (MRI) images of the patients to be tested were acquired. Image recognition was performed on the MRI images of the patients to be tested using a gender-mixed MRI image analysis model and a gender-specific MRI image analysis model, respectively, to obtain gender-related gray matter structure brain networks and gender-specific gray matter structure brain networks.
[0007] Optionally, multivariate regression analysis is used to perform gender regression analysis on magnetic resonance images of different sexes but all in a healthy state to extract multiple gray matter clusters associated with gender differences, thereby obtaining a gender-related gray matter structure brain network, specifically including: Regression calculations were performed on magnetic resonance images of different sexes but both in a healthy state to quantify the influence of sex factors on the analysis results of magnetic resonance images and obtain the feature values of each brain region in the magnetic resonance images. The familial error rate of the feature values of each brain region is corrected to obtain the corrected feature values; and the brain regions corresponding to the corrected feature values that are greater than the preset threshold are identified as gray matter clusters.
[0008] Optionally, all gray matter clusters are converted into corresponding gray matter masks. Paired-samples t-tests are then performed on magnetic resonance images of the same sex but different disease states within the gray matter masks to obtain the absolute gray matter volume of each mask, specifically including: Convert all gray matter clusters in the gender-related gray matter structure brain network into a binary gray matter mask; In each binarized gray matter mask, the absolute gray matter volume of each subject who is of the same gender but has different disease status is determined; By performing a paired-samples t-test on the gray matter volumes of healthy and diseased subjects within the same mask, the absolute gray matter volume representing sex-specific atrophy in each gray matter mask was obtained.
[0009] Optionally, Pearson correlation analysis is performed on the absolute gray matter volume of all gray matter masks using graph theory analysis to obtain female-specific and male-specific gray matter structure brain networks, specifically including: Based on the absolute gray volume data corresponding to all gray masks, construct the feature vector for each subject; The Pearson correlation coefficient between eigenvectors of the same gender was determined by graph theory analysis, resulting in a gender correlation coefficient matrix, which includes a female correlation coefficient matrix and a male correlation coefficient matrix. The gender correlation coefficient matrix is converted into a binary adjacency matrix, and the commonalities of subjects of the same gender are determined within the binary adjacency matrix according to a preset coefficient threshold, thus obtaining the threshold screening results of the gender correlation coefficient matrix. Based on the threshold screening results of the gender correlation coefficient matrix, female-specific gray matter structure brain network and male-specific gray matter structure brain network diagrams were constructed respectively.
[0010] Optionally, when acquiring the magnetic resonance image set, the magnetic resonance images are preprocessed by downsampling, skull image stripping, template registration, spatial normalization and gray matter segmentation in sequence, and the segmented gray matter is spatially smoothed to obtain the processed magnetic resonance images.
[0011] Optionally, the machine learning model includes, but is not limited to, logistic regression models, support vector machine models, k-nearest neighbor algorithms, neural network models, and Naive Bayes classification models.
[0012] The magnetic resonance imaging analysis method for Alzheimer's disease provided in this embodiment of the invention has the following advantages compared with the prior art: This invention uses a gender-related gray matter structure brain network to specifically characterize the brain structure degeneration patterns of patients of different genders. By using a gender-specific gray matter structure brain network, it captures the inherent differences between genders at the neuroanatomical level, using gender factors as a dividing point. In the image analysis process, it retains common features while taking into account gender-differentiated expressions, effectively reducing the impact of individual differences on the analysis results and significantly improving the reliability and accuracy of magnetic resonance image analysis. Attached Figure Description
[0013] Figure 1 This is a multivariate regression analysis plot of a magnetic resonance image analysis method for Alzheimer's disease provided in one embodiment. Figure 1In the diagram, A represents the brain region activation map obtained from multivariate regression analysis of healthy female and male subjects. Figure 1 In the table, B represents the gray matter clusters related to gender. Figure 1 In the diagram, C represents the brain region activation map obtained from a paired-samples t-test between healthy women and women with AD. Figure 1 The "D" in the text refers to the gray matter cladding characteristic of female AD. Figure 1 In the diagram, E represents the brain region activation map obtained from a paired-samples t-test between healthy males and male subjects with AD. Figure 1 F in the figure represents the gray matter clumps characteristic of male AD; Figure 2 This is a heatmap of the center coordinates and brain network connectivity of a magnetic resonance imaging analysis method for Alzheimer's disease provided in one embodiment. Figure 2 In the diagram, A represents the center coordinates of the gray matter clusters related to gender. Figure 2 In the diagram, B represents the coordinates of the center point of the gender-specific gray matter cluster in female Alzheimer's patients. Figure 2 In the diagram, C represents the coordinates of the center point of the gender-specific gray matter cluster in male Alzheimer's patients. Figure 2 D in the image is a heatmap of brain network connectivity in female-specific gray matter structures. The left side represents healthy subjects, and the right side represents Alzheimer's patients. Figure 2 E in the image is a heatmap of brain network connectivity in male-specific gray matter structures. The left side represents healthy subjects, and the right side represents Alzheimer's patients. Figure 2 In the figure, F represents the threshold map of the connectivity strength of female-specific gray matter structural brain networks. Figure 2 G in the figure represents the threshold map of the connectivity strength of male-specific gray matter structural brain networks. Figure 3 This is a network connection pattern diagram of a magnetic resonance image analysis method for Alzheimer's disease provided in one embodiment. Figure 3 In the diagram, A represents the network connectivity pattern of the female-specific gray matter structure in the brain of healthy female subjects. Figure 3 In the diagram, B represents the network connectivity pattern of the female-specific gray matter structure in the brains of female AD subjects. Figure 3 C in the diagram represents the network connectivity pattern of the male-specific gray matter structure in the brain of healthy male subjects. Figure 3 In the diagram, D represents the network connectivity pattern of the male-specific gray matter structure in the brain of male AD subjects. Figure 3 In the graph, E represents the p-value of a local graph theory indicator that shows significant changes in female Alzheimer's patients. Figure 3 F in the figure represents the mean difference plot of a local graph theory indices that show significant variations in female Alzheimer's patients. Figure 3 In the graph, G represents the p-value of a local graph theory indicative value that shows significant changes in male Alzheimer's patients. Figure 3H in the figure represents the mean difference plot of local graph theory indices that show significant variations in male Alzheimer's disease patients; Figure 4 This is a schematic diagram illustrating model training for a magnetic resonance imaging analysis method for Alzheimer's disease provided in one embodiment. Figure 4 In the diagram, A represents the construction process of the gender-specific magnetic resonance imaging analysis model. Figure 4 In this context, B represents the confusion matrix of the gender-mixed magnetic resonance imaging analysis model. Figure 4 In this context, C represents the confusion matrix of the female magnetic resonance imaging analysis model. Figure 4 The confusion matrix of the D male magnetic resonance image analysis model. Figure 4 In the figure, E represents the ROC curve of the gender-mixed magnetic resonance imaging analysis model for all subjects. Figure 4 In this context, F represents the ROC curve of the female magnetic resonance imaging analysis model for female subjects. Figure 4 In this context, G represents the ROC curve of the male magnetic resonance imaging analysis model for male subjects. Figure 5 This is a flowchart illustrating a magnetic resonance imaging analysis method for Alzheimer's disease provided in one embodiment. Detailed Implementation
[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0015] Studies have shown that in the early stages of Alzheimer's Disease (AD), the rate of gray matter atrophy in the female brain is generally faster than in the male brain. However, men experience accelerated atrophy in later stages, potentially leading to a considerable degree of shrinkage over time. This difference may be related to the APOE4 gene, which is more prevalent in women, increasing their susceptibility to AD and resulting in more pronounced hippocampal atrophy and memory decline. In summary, the sex differences in brain structure in AD patients involve multiple factors, including genetic susceptibility and connectivity patterns. These differences provide valuable insights into the pathogenesis of AD and pave the way for the development of targeted therapies.
[0016] The purpose of this invention is to identify gender-related gray matter regions and investigate whether these regions exhibit significant atrophy in Alzheimer's disease. Specifically, gray matter atrophy patterns in female and male Alzheimer's patients were analyzed to construct a gender-specific gray matter structural brain network. Graph theory analysis was used to assess the differences in atrophy between female and male Alzheimer's patients. Finally, the gender-specific gray matter structure was used as input features for machine learning to develop a magnetic resonance imaging (MRI) model.
[0017] In one embodiment, a magnetic resonance imaging analysis method for Alzheimer's disease is provided, such as... Figure 5 As shown, the method includes: Obtain a set of magnetic resonance imaging (MRI) images of Alzheimer's disease, consisting of MRI images of healthy men, healthy women, men with the disease, and women with the disease.
[0018] Multivariate regression analysis was used to perform gender regression analysis on magnetic resonance images of different sexes but all in a healthy state, in order to extract multiple gray matter clusters associated with gender differences and obtain a gender-related gray matter structure brain network.
[0019] All gray matter clusters were converted into corresponding gray matter masks. Paired-samples t-tests were performed on MRI images of the same sex but different disease states within the gray matter masks to obtain the absolute gray matter volume of each mask. Pearson correlation analysis was performed on the absolute gray matter volumes of all gray matter masks using graph theory to obtain female-specific and male-specific gray matter structure brain networks.
[0020] The machine learning model was trained by gender-related gray matter structure brain network, female-specific gray matter structure brain network and male-specific gray matter structure brain network, respectively, to obtain gender-mixed magnetic resonance imaging analysis model, female magnetic resonance imaging analysis model and male magnetic resonance imaging analysis model.
[0021] Magnetic resonance imaging (MRI) images of the patients to be tested were acquired. Image recognition was performed on the MRI images of the patients to be tested using a gender-mixed MRI image analysis model and a gender-specific MRI image analysis model, respectively, to obtain gender-related gray matter structure brain networks and gender-specific gray matter structure brain networks.
[0022] Among them, the gender-related gray matter structure brain network was obtained by processing the magnetic resonance images of the test patients through a gender-mixed magnetic resonance image analysis model, and the gender-specific gray matter structure brain network was obtained by processing the magnetic resonance images of the test patients through a magnetic resonance image analysis model corresponding to the gender.
[0023] Specifically, multivariate regression analysis was used to perform gender regression analysis on magnetic resonance images of different sexes but all in a healthy state to extract multiple gray matter clusters associated with gender differences, thereby obtaining a gender-related gray matter structure brain network, which specifically includes: Regression calculations were performed on magnetic resonance images of different sexes but both in a healthy state to quantify the influence of sex on the analysis results of magnetic resonance images and obtain the feature values of each brain region in the magnetic resonance images.
[0024] The familial error rate of the feature values of each brain region is corrected to obtain the corrected feature values; and the brain regions corresponding to the corrected feature values that are greater than the preset threshold are identified as gray matter clusters.
[0025] This process involves converting all gray matter clusters into corresponding gray matter masks. Paired-samples t-tests are then performed on magnetic resonance images of the same sex but different disease states within the gray matter masks to obtain the absolute gray matter volume of each mask. Specifically, this includes: All gray matter clusters in the gender-related gray matter structure brain network were converted into binary gray matter masks. Within each binary gray matter mask, the absolute gray matter volume of each subject, who was of the same sex but had different disease states, was determined. Paired-samples t-tests were performed on the gray matter volumes of healthy and diseased subjects within the same mask to obtain the absolute gray matter volume representing gender-specific atrophy in each gray matter mask.
[0026] Specifically, Pearson correlation analysis was performed on the absolute gray matter volume of all gray matter masks using graph theory analysis to obtain female-specific and male-specific gray matter structure brain networks, including: Based on the absolute gray matter volume data corresponding to all gray matter masks, a feature vector for each subject is constructed. Pearson correlation coefficients between feature vectors of the same gender are determined using graph theory analysis, resulting in a gender correlation coefficient matrix, which includes both female and male correlation coefficient matrices. The gender correlation coefficient matrix is then converted into a binary adjacency matrix, and common relationships among subjects of the same gender are determined within this matrix based on preset coefficient thresholds, yielding a threshold selection result for the gender correlation coefficient matrix. Based on the threshold selection result of the gender correlation coefficient matrix, female-specific gray matter structure brain networks and male-specific gray matter structure brain network diagrams are constructed respectively.
[0027] In the process of acquiring the magnetic resonance image set, the magnetic resonance images are preprocessed by downsampling, skull image stripping, template registration, spatial normalization and gray matter segmentation in sequence, and the segmented gray matter is spatially smoothed to obtain the processed magnetic resonance images.
[0028] Machine learning models include, but are not limited to, logistic regression models, support vector machine models, k-nearest neighbor algorithms, neural network models, and Naive Bayes classification models.
[0029] The specific implementation includes: 1. Acquisition and preprocessing of magnetic resonance data.
[0030] The 3.0 Tesla structural MRI data from 247 healthy elderly female subjects, 299 healthy elderly male subjects, 135 elderly female patients, and 226 elderly male subjects were all obtained from the ADNI database. First, the subjects were downsampled to 1.5 mm voxel images in NIFTI format. Then, CAT12 was used for skull dissection, template registration, spatial normalization, and gray matter segmentation, and the gray matter was spatially smoothed using an 8 mm Gaussian smoothing kernel.
[0031] 2. Multivariate regression analysis and paired-samples t-test.
[0032] Multivariate regression analysis was used to perform gender regression analysis on the smooth gray matter structure of females and males. During the regression process, the influence of variables such as age and total intracranial volume was excluded. In this invention, the scanning parameters of all images were consistent, and the image quality score was greater than 75%. Therefore, other factors such as surgical location, scanning parameters, and image quality were not considered. Similarly, the influence of factors such as age and total intracranial volume was eliminated when performing paired-samples t-tests on healthy subjects and Alzheimer's disease subjects. All image analyses required family-wise error rate correction (FWE correction), and a p-value less than 0.05 and a k-value greater than 30 after correction were considered statistically significant.
[0033] 3. Construction and graph theory analysis of gender-specific gray matter brain networks.
[0034] Using the Pearson correlation coefficient between gender-specific gray matter structures, female-specific and male-specific gray matter structures were constructed. A Pearson correlation coefficient greater than 0.3 and a p-value less than 0.05 were considered to indicate connections between network nodes. Then, analytical methods based on graph theory were used to measure parameters such as proximity centrality, clustering, eccentricity, global efficiency nodes, local efficiency nodes, path length, strength, and triangle formation of nodes in the brain networks.
[0035] Measurement process: The Pearson correlation coefficient was obtained by extracting the gray matter volume parameters of gray matter clusters in the image using the ROI gray matter volume extraction module in the CAT12 software package, and then correlation analysis was performed on the gray matter volume parameters of each gray matter cluster. The parameters of the brain network nodes were measured using the BRAPH software package to measure the network node information of each gray matter cluster.
[0036] Then, a t-test was performed on healthy subjects and Alzheimer's disease subjects. A p-value less than 0.05 was considered statistically significant.
[0037] 4. Machine learning models.
[0038] Five relatively common machine learning algorithms were selected, including logistic regression, support vector machine, k-nearest neighbor, neural network, and Naive Bayes classification. The models in this invention are divided into three groups. The input features of the gender-mixed magnetic resonance imaging (MRI) model are gender-related gray matter structures; the input features of the female MRI model are female-specific gray matter structures; and the input features of the male MRI model are male-specific gray matter structures. The training set to test set ratio for all models is 8:2. In the model evaluation process, this invention uses accuracy, precision, recall, F1 score, AUC value, confusion matrix, and ROC curve to comprehensively evaluate the models.
[0039] 5. Results.
[0040] 5.1 Gender-specific gray matter clusters.
[0041] Multivariate regression analysis was performed on preprocessed 3.0 T structural magnetic resonance images of healthy female and male subjects, identifying 11 gray matter clusters associated with sex differences (…). Figure 1 (A) In this invention, 11 gray matter clusters are used as masks to perform paired-samples t-tests on gray matter images of healthy female subjects and female subjects with Alzheimer's disease.
[0042] The calculation process of the paired-samples t-test: The gray volume parameters of the gray mass clusters in the image are obtained through the ROI gray volume extraction module in the CAT12 software package, and then correlation analysis is performed on the gray volume parameters of each gray mass cluster. This could actually be excluded from the claims.
[0043] Analysis of this invention showed that 8 out of 11 gray matter clusters exhibited significant atrophy in the brains of female Alzheimer's patients. Figure 1 (C in the text). Similarly, in the male cohort, 8 out of 11 gray matter clusters also showed significant atrophy (C in the text). Figure 1 (E in the text).
[0044] 5.2 Gender-specific gray matter structural brain networks.
[0045] In 11 sex-related gray matter clusters, female Alzheimer's patients did not show significant atrophy in groups 3, 6, and 11. Figure 2 (B in the text). On the other hand, male Alzheimer's patients did not show significant atrophy in groups 3, 10, and 11 ( Figure 2 Then, the absolute gray matter volumes of eight different gray matter clusters from female and male Alzheimer's patients were obtained, and a gender-specific gray matter structural brain network was constructed. Figure 2 D and Figure 2E in the text). After setting the node correlation threshold of the gray matter structural brain network to 0.3, a significant reduction in the overall connectivity of the network was observed in both female and male Alzheimer's patients. Figure 2 F and Figure 2 (G in the middle).
[0046] 5.3 Graph theory analysis of gender-specific gray matter brain networks.
[0047] This study uses graph theory to investigate global and local changes in female- and male-specific gray matter structural brain networks in Alzheimer's disease. The analysis shows that the connectivity complexity in the gray matter structural brain networks of both female and male Alzheimer's patients is significantly reduced. Figure 3 A in Figure 3 B in Figure 3 C and Figure 3 (D in the text). However, the changes in male-specific gray matter structural brain networks were more pronounced in male Alzheimer's patients. In female-specific gray matter brain networks, six local graph theory indices from the second cluster and seven indices from the ninth cluster showed significant changes in female Alzheimer's patients (D in the text). Figure 3 E and Figure 3 In contrast, in the male-specific gray matter brain network, 33 local graph theory indices across groups 1, 2, 4, 5, 7, 8, and 9 showed significant changes in male Alzheimer's patients (F). Figure 3 G and Figure 3 (H in the figure). The graph theory analysis results above indicate that there are significant differences in gray matter atrophy patterns between female and male Alzheimer's patients.
[0048] 5.4 Magnetic Resonance Image Analysis Model.
[0049] Five machine learning algorithms (Logistic Regression, Support Vector Machine, k-Nearest Neighbor, Neural Network, and Naive Bayes) were used to develop three different magnetic resonance imaging (MRI) models: a gender-mixed MRI model, a female MRI model, and a male MRI model. Figure 4 (A in Table 1).
[0050] The data processing procedure of the model: The first step is to input the volume size index of each gray matter cluster for each subject corresponding to each model.
[0051] There are five types of machine learning models: logistic regression, support vector machine, k-nearest neighbor algorithm, neural network, and Naive Bayes classification. For machine learning models, we use 80% of the sample size for training and 20% for testing.
[0052] The model output is: magnetic resonance image analysis results.
[0053] The evaluation metrics for the model include five: accuracy, precision, recall, F1 score, and AUC.
[0054] The performance of these models was evaluated using accuracy, precision, recall, F1 score, and AUC. Combining the five evaluation metrics, the gender-mixed MRI image analysis model had the highest number of input feature values, but its performance was the lowest (AUC 0.9006). Figure 4 In contrast, the female MRI image analysis model (AUC 0.9506, E). Figure 4 The F in the model and the male magnetic resonance imaging analysis model (AUC 0.9156, Figure 4 All of the G algorithms in the evaluation showed relatively high performance. Among the five machine learning models evaluated, the k-nearest neighbor algorithm demonstrated superior performance.
[0055] Table 1 Model Evaluation Table Results Analysis: Multivariate regression analysis of preprocessed 3.0T structural MRI images of healthy elderly women and men revealed 11 gray matter clusters significantly associated with sex. Specifically, the first and second gray matter clusters were primarily located in the left and right limbic lobes, respectively. The third, sixth, and eighth gray matter clusters were primarily located in the left and right occipital lobes. The fourth gray matter cluster was primarily distributed in the inferior lobe. The fifth gray matter cluster was primarily located in the left inferior temporal gyrus and temporal pole. The seventh gray matter cluster was primarily located in the right parietal lobe. The ninth gray matter cluster was primarily located in the right middle temporal gyrus. Notably, the tenth and eleventh gray matter clusters... Overall, these sex-related brain regions were primarily located in the limbic and temporal lobes. Significant progress has been made in research on sex differences in the limbic and temporal lobes. Studies have shown that male brain structures tend to be larger than female brain structures, particularly in the temporal lobe. This structural difference may contribute to differences in cognitive and behavioral characteristics between sexes. For example, men typically perform better in spatial orientation, mathematical reasoning, and mechanical skills, while women typically perform better in language processing, social cognition, and emotional tasks. These gender-related differences in brain structure are likely the result of a complex interplay of genetic, hormonal, and sociocultural factors.
[0056] Of the 11 gray matter clusters, several showed significant atrophy in both female and male Alzheimer's patients, particularly clusters 1, 2, 4, 5, 7, 8, and 9. However, the degree of atrophy varied among these clusters. Furthermore, female Alzheimer's patients showed significant atrophy in group 10, while male Alzheimer's patients showed significant atrophy in group 6. Despite the relatively small size of these two gray matter clusters, the stability of each node within the network is fundamental to maintaining the stability of the entire network. This invention constructs gender-specific gray matter structural brain networks for women and men based on gender-specific gray matter clusters. Comparing the gray matter structural brain networks of healthy and Alzheimer's patients, this invention found that the complexity of gray matter structural brain network connections was significantly reduced in both female and male Alzheimer's patients. The study indicates that, compared to healthy brains, Alzheimer's patients exhibit increased brain network path length and reduced overall efficiency, suggesting significant structural damage. This damage is closely related to the pathological features of AD, including amyloid plaques, neurofibrillary tangles, neuronal loss, and synaptic dysfunction.
[0057] While both female and male Alzheimer's patients showed a similar decreasing trend in gray matter structural brain networks in terms of complexity, significant differences existed in changes in network nodes. Graph theory analysis revealed that in female-specific gray matter structural brain networks, significant changes were primarily limited to proximity centrality, clustering coefficient, path length, and node strength within clusters 2 and 9. In contrast, when comparing male Alzheimer's patients with healthy controls, male-specific gray matter structural brain networks showed significant differences across seven clusters. It was observed that gender-specific changes in gray matter structural brain networks were more pronounced in male patients than in female patients.
[0058] This invention develops three different magnetic resonance imaging (MRI) models: a gender-mixed MRI model utilizing a gender-related gray matter structure brain network, a female MRI model using a female-specific gray matter structure brain network, and a male MRI model using a male-specific gray matter structure brain network.
[0059] Model evaluation shows that, compared with the gender-mixed MRI image analysis model, the gender-specific MRI image analysis model has higher accuracy despite having fewer input features. This invention employs five machine learning algorithms: logistic regression, support vector machine (SVM), k-nearest neighbor (KNN), neural network, and Naive Bayes.
[0060] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.
Claims
1. A magnetic resonance imaging analysis method for Alzheimer's disease, characterized in that, include: Obtain a set of magnetic resonance imaging (MRI) images of Alzheimer's disease, consisting of MRI images of healthy men, healthy women, men with the disease, and women with the disease. Multivariate regression analysis was used to perform gender regression analysis on magnetic resonance images of different sexes but all in a healthy state, in order to extract multiple gray matter clusters related to gender differences and obtain a gender-related gray matter structure brain network. All gray matter clusters are converted into corresponding gray matter masks. Paired-samples T-tests are performed on magnetic resonance images of the same sex but different disease states within the gray matter masks to obtain the absolute gray matter volume of each gray matter mask. Pearson correlation analysis was performed on the absolute gray matter volume of all gray matter masks using graph theory analysis to obtain female-specific gray matter structure brain networks and male-specific gray matter structure brain networks. The machine learning model was trained by gender-related gray matter structure brain network, female-specific gray matter structure brain network and male-specific gray matter structure brain network respectively, resulting in gender-mixed magnetic resonance image analysis model, female magnetic resonance image analysis model and male magnetic resonance image analysis model. Magnetic resonance imaging (MRI) images of the patients to be tested were acquired. Image recognition was performed on the MRI images of the patients to be tested using a gender-mixed MRI image analysis model and a gender-specific MRI image analysis model, respectively, to obtain gender-related gray matter structure brain networks and gender-specific gray matter structure brain networks.
2. The magnetic resonance imaging analysis method for Alzheimer's disease as described in claim 1, characterized in that, The method involves performing gender regression analysis on magnetic resonance images of individuals of different sexes but all in a healthy state to extract multiple gray matter clusters associated with gender differences, thereby obtaining a gender-related gray matter structure brain network. Specifically, this includes: Regression calculations were performed on magnetic resonance images of different sexes but both in a healthy state to quantify the influence of sex factors on the analysis results of magnetic resonance images and obtain the feature values of each brain region in the magnetic resonance images. The familial error rate of the feature values of each brain region is corrected to obtain the corrected feature values; and the brain regions corresponding to the corrected feature values that are greater than the preset threshold are identified as gray matter clusters.
3. The magnetic resonance imaging analysis method for Alzheimer's disease as described in claim 1, characterized in that, The process involves converting all gray matter clusters into corresponding gray matter masks, performing paired-samples T-tests on magnetic resonance images of the same sex but different disease states within the gray matter masks, and obtaining the absolute gray matter volume of each gray matter mask. Specifically, this includes: Convert all gray matter clusters in the gender-related gray matter structure brain network into a binary gray matter mask; In each binarized gray matter mask, the absolute gray matter volume of each subject who is of the same gender but has different disease status is determined; By performing a paired-samples t-test on the gray matter volumes of healthy and diseased subjects within the same mask, the absolute gray matter volume representing sex-specific atrophy in each gray matter mask was obtained.
4. The magnetic resonance imaging analysis method for Alzheimer's disease as described in claim 1, characterized in that, The graph theory analysis method was used to perform Pearson correlation analysis on the absolute gray matter volume of all gray matter masks to obtain female-specific and male-specific gray matter structure brain networks, specifically including: Based on the absolute gray volume data corresponding to all gray masks, construct the feature vector for each subject; The Pearson correlation coefficient between eigenvectors of the same gender is determined by graph theory analysis, resulting in a gender correlation coefficient matrix, which includes a female correlation coefficient matrix and a male correlation coefficient matrix. The gender correlation coefficient matrix is converted into a binary adjacency matrix, and the commonalities of subjects of the same gender are determined within the binary adjacency matrix according to a preset coefficient threshold, thus obtaining the threshold screening results of the gender correlation coefficient matrix. Based on the threshold screening results of the gender correlation coefficient matrix, female-specific gray matter structure brain network and male-specific gray matter structure brain network diagrams were constructed respectively.
5. The magnetic resonance imaging analysis method for Alzheimer's disease as described in claim 1, characterized in that, When acquiring the magnetic resonance image set, the magnetic resonance images are preprocessed by downsampling, skull image stripping, template registration, spatial normalization and gray matter segmentation in sequence, and the segmented gray matter is spatially smoothed to obtain the processed magnetic resonance images.
6. The magnetic resonance imaging analysis method for Alzheimer's disease as described in claim 1, characterized in that, The machine learning models include, but are not limited to, logistic regression models, support vector machine models, k-nearest neighbor algorithms, neural network models, and Naive Bayes classification models.