Mental disease classification method and system based on individual difference structure covariant network and machine learning

Abstract

The invention provides a mental disease classification method and system based on an individual difference structure covariant network and machine learning, and belongs to the field of mental disease classification. The problem of classification performance bottleneck caused by heterogeneity of mental diseases in the prior art is solved. The method comprises the following steps: acquiring a structural magnetic resonance T1 weighted image, and preprocessing the image; performing brain region segmentation on the pre-processed T1 image based on an AAL brain map, and extracting the gray matter volume of each brain region; constructing an IDSCN network by calculating the Pearson's correlation coefficient of the brain grey matter volume of the paired brain regions; calculating the area under a node topological attribute curve of the IDSCN network; screening node attribute indexes with statistical differences between the patient group and the healthy control group through double-sample t test; and taking the screened node attribute indexes as feature vectors, and inputting the feature vectors into a support vector machine classification model for disease classification. The method is mainly used in the medical image processing field.

Description

Technical Field

[0001] The present invention belongs to the field of mental illness classification, and in particular relates to a mental illness classification method based on individual difference structural covariation network and machine learning. Background Art

[0002] Schizophrenia (SZ), obsessive-compulsive disorder (OCD), and bipolar disorder (BD) are three psychiatric disorders with complex clinical manifestations and highly overlapping genetic and environmental risk factors. Current diagnoses rely primarily on clinical symptom assessment and lack objective biological markers, resulting in high misdiagnosis rates and difficulty in achieving accurate classification. In recent years, neuroimaging studies have demonstrated that these disorders are closely associated with abnormal connectivity in large-scale brain networks. Structural magnetic resonance imaging (sMRI) can capture the structural covariation of brain networks through morphological metrics such as gray matter volume and cortical thickness, providing a new approach to uncovering disease pathology.

[0003] Traditional structural covariance networks (SCNs) are constructed by calculating correlations between group-level brain morphological metrics, but this approach ignores the significant individual heterogeneity of psychiatric disorders. For example, patients with SZ, OCD, and BD exhibit wide variations in symptom dimensions, treatment responses, and prognosis, and group-level SCNs struggle to capture these individual-specific brain network alterations. This limitation results in low reproducibility of imaging findings and hinders the discovery of stable biomarkers. To overcome this issue, the individual differential structural covariance network (IDSCN) was proposed. By quantifying differences in brain region coordination between individuals relative to healthy controls, it can more sensitively capture disease-related brain network remodeling. However, existing research is largely limited to the construction and analysis of IDSCNs, and their application in psychiatric classification has not been fully explored.

[0004] While machine learning methods such as support vector machines (SVMs) are widely used for neuroimaging classification, their direct application to brain network classification faces significant challenges. First, traditional SVMs are susceptible to the curse of dimensionality when processing high-dimensional brain network features, leading to model overfitting. Second, conventional methods fail to consider the dynamic range of individual network characteristics, such as the impact of sparsity threshold selection on network topology. Furthermore, existing techniques often use single node or edge features, failing to effectively integrate the multi-scale topological properties of IDSCNs, limiting further improvements in classification performance. Summary of the Invention

[0005] In view of this, the present invention aims to propose a method and system for classifying mental illnesses based on individual difference structural covariation networks and machine learning, so as to solve the classification performance bottleneck problem caused by the heterogeneity of existing mental illnesses.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for classifying mental illnesses based on individual difference structural covariation networks and machine learning, the method comprising: Step 1: Acquire structural magnetic resonance T1-weighted images and perform image preprocessing, including data format conversion, bias field correction, and skull stripping; Step 2: Segment the preprocessed T1 image into brain regions based on the AAL brain atlas and extract the gray matter volume of each brain region; Step 3: Construct the IDSCN network by calculating the Pearson correlation coefficient of the gray matter volume of paired brain regions; Step 4: Calculate the area under the node topology attribute curve of the IDSCN network within the sparsity threshold range of 0.05-0.4; Step 5: Use the two-sample t-test to screen the node attribute indicators with statistical differences between the patient group and the healthy control group; Step 6: Use the selected node attribute indicators as feature vectors and input them into the support vector machine classification model for disease classification.

[0007] Furthermore, a preferred embodiment is proposed, wherein the pretreatment in step 1 further comprises: The DARTEL algorithm was used for brain region registration and gray matter volume normalization; Gray matter images were spatially smoothed using a 6-mm full-width-at-half-maximum Gaussian kernel; Gray matter volume was corrected using total brain volume and age as covariates.

[0008] Furthermore, a preferred embodiment is proposed, wherein step 2 comprises: The Pearson correlation coefficient matrix PCCn of the gray matter volumes of paired brain regions of n healthy subjects was calculated as the reference network; The target subject is added to the reference group, and the Pearson correlation coefficient matrix PCCn+1 of n+1 subjects is calculated as the perturbation network; The difference between the perturbation network and the reference network was calculated and Z-score normalized to obtain the individual difference network.

[0009] Furthermore, a preferred method is proposed, wherein the Z-score normalization includes: .

[0010] Furthermore, a preferred embodiment is proposed, wherein the method for determining the sparsity threshold in step 4 includes: Calculate the small world coefficient in the range of 0.05-0.5; The maximum sparsity range corresponding to the small-world coefficient > 1.1 is selected as the calculation threshold interval.

[0011] Furthermore, a preferred method is proposed, in which the SVM classification model adopts an optimized combination of a linear kernel function and a regularization parameter C, and the result is verified by a leave-one-out cross-validation strategy.

[0012] Furthermore, a preferred embodiment is proposed, wherein the node attributes include: The area under the curve of betweenness centrality in the bilateral middle frontal gyrus; The area under the node efficiency curve of the left calcarine groove; At least one combination of the area under the degree centrality curve of the right putamen.

[0013] Based on the same inventive concept, the present invention also proposes a mental illness classification system based on individual difference structural covariation network and machine learning, the system comprising: The preprocessing unit is used to acquire structural magnetic resonance T1-weighted images and perform image preprocessing, including data format conversion, bias field correction, and skull stripping; The segmentation unit is used to segment the preprocessed T1 image into brain regions based on the AAL brain atlas and extract the gray matter volume of each brain region; IDSCN network construction unit, used to construct the IDSCN network by calculating the Pearson correlation coefficient of the gray matter volume of paired brain regions; A calculation unit, used for calculating the area under the curve of the node topology attribute of the IDSCN network within a sparsity threshold range of 0.05-0.4; A screening unit is used to screen the node attribute indicators with statistical differences between the patient group and the healthy control group through a two-sample t test; The classification unit is used to input the filtered node attribute indicators as feature vectors into the support vector machine classification model for disease classification.

[0014] Based on the same inventive concept, the present invention also proposes a computer device, including a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes a mental illness classification method based on individual difference structural covariation network and machine learning as described in any one of the above items.

[0015] Based on the same inventive concept, the present invention also proposes a computer-readable storage medium, which stores a computer program. When the computer program is run by a processor, it executes the steps of a mental illness classification method based on individual difference structural covariation network and machine learning as described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The clinical manifestations and prognosis of patients with mental illnesses such as schizophrenia (SZ), obsessive-compulsive disorder (OCD), and bipolar disorder (BD) vary significantly. Existing classification methods are mostly based on group-level brain network analysis, which is difficult to accurately reflect individual differences. The method proposed in this paper introduces the individual difference structural covariation network (IDSCN) to quantify the differences in brain region coordination changes between individuals and healthy controls, thereby more sensitively capturing the characteristics of disease-related brain network remodeling and effectively addressing the individual differences in mental illness. Compared with the traditional group SCN: (1) it eliminates sample selection bias and improves the stability of network construction (experiments show that the network similarity correlation coefficient r ≥ 0.85 within the healthy control group); (2) it captures disease-specific topological attribute changes, such as the abnormal increase in bilateral middle frontal gyrus betweenness centrality in SZ patients, providing a new dimension for the analysis of heterogeneous pathological mechanisms.

[0017] The proposed method for mental illness classification uses the area under the IDSCN node attribute curve (AUC) as a classification feature and combines it with an SVM parameter optimization strategy (regularization parameter C = 1.5, linear kernel function) to achieve high-precision cross-disease identification: (1) The classification accuracy of patients and healthy controls reached 86.54%-95.74% (e.g., SZ vs HCs accuracy of 86.54%, AUC = 0.89); (2) The performance of disease subtype differentiation broke through the bottleneck of existing technologies (e.g., OCD vs BD accuracy of 61.90%, an increase of 23.6% compared with traditional SCN+SVM), and the permutation test confirmed the significant results (p < 0.001).

[0018] The proposed method for classifying mental illnesses innovatively uses node attribute integrals (0.05-0.4, with a step size of 0.01) under multiple sparsity thresholds as feature input, addressing the feature sensitivity issues caused by selecting a single threshold. Experimental results show that: (1) the combined feature combination of betweenness centrality and node efficiency improves the classification specificity of SZ and OCD to 82.61% (compared to a 17.3% increase for single attribute features); (2) by using Bonferroni correction to select key brain regions (such as the right superior parietal gyrus aEloc), the feature dimensionality is effectively reduced (the number of features is reduced from 90 brain regions to 3-5), and the model computational efficiency is improved by 40 times.

[0019] The present invention is applied to the field of medical image processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings: Figure 1 This is a flow chart of a method for classifying mental illness based on individual difference structural covariation networks and machine learning according to the present invention; Figure 2 Schematic diagram of the ROC curve for the SZ patient-specific node attribute classification according to the present invention, wherein: Figure 2 (a) Schematic diagram of ROC curve for SZ and HCs classification. Figure 2 (b) Schematic diagram of the ROC curve for SZ and OCD classification; Figure 3 Schematic diagram of the ROC curve for the OCD patient-specific node attribute classification according to the present invention; Figure 4 Schematic diagram of the ROC curve for the BD patient-specific node attribute classification of the present invention, wherein: Figure 4 (a) Schematic diagram of the ROC curve for the classification of BD patients and healthy controls. Figure 4 (b) Schematic diagram of the ROC curve for the classification of BD patients and SZ patients. False Positive Rate represents the false positive rate, and True Positive Rate represents the true positive rate. DETAILED DESCRIPTION

[0021] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely explain the technical solutions in the embodiments of the present invention. It should be noted that the embodiments of the present invention and the features therein can be combined with each other in the absence of conflict, and the embodiments described are only part of the embodiments of the present invention, not all of the embodiments.

[0022] Implementation method 1, see Figure 1 This embodiment describes a method for classifying mental illnesses based on individual difference structural covariation networks and machine learning, comprising: Step 1: Acquire structural magnetic resonance T1-weighted images and perform image preprocessing, including data format conversion, bias field correction, and skull stripping; Step 2: Segment the preprocessed T1 image into brain regions based on the AAL brain atlas and extract the gray matter volume of each brain region; Step 3: Construct the IDSCN network by calculating the Pearson correlation coefficient of the gray matter volume of paired brain regions; Step 4: Calculate the area under the node topology attribute curve of the IDSCN network within the sparsity threshold range of 0.05-0.4; Step 5: Use the two-sample t-test to screen the node attribute indicators with statistical differences between the patient group and the healthy control group; Step 6: Use the selected node attribute indicators as feature vectors and input them into the support vector machine classification model for disease classification.

[0023] Patients with psychiatric disorders such as schizophrenia (SZ), obsessive-compulsive disorder (OCD), and bipolar disorder (BD) exhibit significant differences in clinical presentation and prognosis. Existing classification methods, which are mostly based on group-level brain network analysis, struggle to accurately reflect individual differences. The method proposed in this embodiment, by introducing the Individual Difference Structural Covariance Network (IDSCN), can quantify differences in brain region coordination between individuals relative to healthy controls. This allows for more sensitive capture of disease-related brain network remodeling, effectively addressing the individual variability of psychiatric disorders.

[0024] The method proposed in this implementation introduces personalized network characteristics based on traditional methods by constructing an individual difference-based structural covariation network (IDSCN). This allows for a full representation of each patient's brain network characteristics, rather than simply based on group averages. This feature can make disease classification results more personalized, improving classification accuracy and reliability.

[0025] Unlike traditional methods, the method proposed in this implementation further calculates the area under the curve of the network's node topology properties after constructing the IDSCN network, exploring different network topological characteristics at different sparsity thresholds. This multi-scale analysis method comprehensively considers the organizational and functional characteristics of brain networks at different levels, avoiding the single-scale limitations of traditional methods and improving the comprehensiveness of feature representation.

[0026] This embodiment uses a sparsity threshold adjusted within a range of 0.05-0.4 to calculate the area under the curve of the node topology attributes of the IDSCN network. This processing method, through flexible sparsity selection, can better adapt to the expression of different brain network features, reduce the impact of different sparsities on the network topology structure, and further optimize the feature extraction process. Through a two-sample t-test, node attribute indicators with statistical differences between the patient group and the healthy control group are screened, allowing the classification model to focus on the most discriminatory features for disease diagnosis, avoiding the interference of redundant and irrelevant features on classification performance, and effectively improving the accuracy of the classification model. At the same time, the method proposed in this embodiment combines support vector machines (SVMs) and inputs the screened node attributes as feature vectors into the SVM classification model. This can effectively capture subtle differences between brain networks in different groups, thereby improving classification accuracy. This combined approach helps overcome the shortcomings of traditional methods in the classification of mental illnesses and achieve more accurate classification.

[0027] Implementation 2: This implementation further limits the mental illness classification method based on individual difference structural covariation network and machine learning described in Implementation 1. The preprocessing in step 1 further includes: The DARTEL algorithm was used for brain region registration and gray matter volume normalization; Gray matter images were spatially smoothed using a 6-mm full-width-at-half-maximum Gaussian kernel; Gray matter volume was corrected using total brain volume and age as covariates.

[0028] In this implementation, deformable registration of brain images allows for mapping the brain structures of different individuals to the same standard space, minimizing the impact of individual differences. Gray matter volume normalization ensures that potential differences between individuals due to varying brain volumes do not affect the final classification results. A 6mm full-width-at-half-maximum Gaussian kernel is used to spatially smooth the gray matter images, eliminating noise and improving image quality. In gray matter volume analysis, taking into account covariates such as total brain volume and age effectively eliminates the influence of these factors, preventing them from interfering with the detection of pathological changes and preventing misleading disease classification due to differences in age or brain volume.

[0029] Implementation 3: This implementation further limits the mental illness classification method based on individual difference structural covariation network and machine learning described in Implementation 1. Step 2 includes: The Pearson correlation coefficient matrix PCCn of the gray matter volumes of paired brain regions of n healthy subjects was calculated as the reference network; The target subject is added to the reference group, and the Pearson correlation coefficient matrix PCCn+1 of n+1 subjects is calculated as the perturbation network; The difference between the perturbation network and the reference network was calculated and Z-score normalized to obtain the individual difference network.

[0030] In this implementation, the target subject is compared with a group of healthy subjects to calculate the Pearson correlation coefficient matrix (PCC) for each subject, effectively capturing inter-individual differences in brain gray matter volume. By calculating the difference between the target subject and the reference group and applying Z-score normalization, potential biases between individuals are eliminated, ensuring data consistency and reliability, and avoiding the limitations of relying solely on absolute data. The calculation of a perturbation network further enhances the flexibility and adaptability of the analysis. By adding the target subject to a reference group of healthy subjects, a "perturbation network" is formed, which better captures the changes in the target subject's overall network structure.

[0031] Embodiment 4: This embodiment further limits the mental illness classification method based on individual difference structural covariation network and machine learning described in embodiment 3, wherein the Z-score standardization includes: .

[0032] Implementation 5: This implementation further limits the mental illness classification method based on individual difference structural covariation network and machine learning described in Implementation 1. The method for determining the sparsity threshold in step 4 includes: Calculate the small world coefficient in the range of 0.05-0.5; The maximum sparsity range corresponding to the small-world coefficient > 1.1 is selected as the calculation threshold interval.

[0033] Implementation method six. This implementation method further limits the mental illness classification method based on individual difference structural covariation network and machine learning described in implementation method one. The SVM classification model adopts an optimized combination of linear kernel function and regularization parameter C, and verifies the results through leave-one-out cross-validation strategy.

[0034] Implementation 7: This implementation further limits the mental illness classification method based on individual difference structural covariation network and machine learning described in Implementation 1. The node attributes include: The area under the curve of betweenness centrality in the bilateral middle frontal gyrus; The area under the node efficiency curve of the left calcarine groove; At least one combination of the area under the degree centrality curve of the right putamen.

[0035] Embodiment 8: A mental illness classification system based on individual difference structural covariation network and machine learning, characterized in that the system includes: The preprocessing unit is used to acquire structural magnetic resonance T1-weighted images and perform image preprocessing, including data format conversion, bias field correction, and skull stripping; The segmentation unit is used to segment the preprocessed T1 image into brain regions based on the AAL brain atlas and extract the gray matter volume of each brain region; IDSCN network construction unit, used to construct the IDSCN network by calculating the Pearson correlation coefficient of the gray matter volume of paired brain regions; A calculation unit, used for calculating the area under the curve of the node topology attribute of the IDSCN network within a sparsity threshold range of 0.05-0.4; A screening unit is used to screen the node attribute indicators with statistical differences between the patient group and the healthy control group through a two-sample t test; The classification unit is used to input the filtered node attribute indicators as feature vectors into the support vector machine classification model for disease classification.

[0036] Implementation method 9. A computer device described in this implementation method includes a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes a mental illness classification method based on individual difference structural covariation network and machine learning as described in any one of implementation methods 1 to 7.

[0037] Implementation method 10. A computer-readable storage medium described in this implementation method stores a computer program, and when the computer program is executed by a processor, the steps of a method for classifying mental illnesses based on individual difference structural covariation networks and machine learning are performed as described in any one of implementation methods 1 to 7.

[0038] Implementation method 11, see Figures 2 to 4 This embodiment provides a specific example of the mental illness classification method based on individual difference structural covariation network and machine learning described in embodiment 1, and is also used to explain embodiments 2 to 7. Specifically: Step 1: Imaging data acquisition: T1WI images were acquired using a GE 3.0 T scanner (750 Signa-HDX). During the scan, patients were provided with cotton pads to stabilize their heads and cotton ball earplugs to isolate noise. All subjects were instructed to keep their heads still and avoid sleeping and thinking as much as possible. T1WI image acquisition sequence and parameters: Gradient echo sequence was used to acquire images with the following parameters: TR = 2530 ms; TE = 3.39 ms; FOV = 256 × 256; slice thickness = 1.3 mm; interslice spacing = 0.5 mm; flip angle = 7°; matrix = 256 × 192; voxel = 3 mm × 3 mm × 3 mm; a total of 176 sagittal sections were acquired.

[0039] Step 2: Data preprocessing process includes data format conversion, bias field correction, and skull stripping: (1) Data format conversion: convert the data format from DICOM to NIFTI; T1 images in NIFTI format were manually inspected for quality and anatomical abnormalities. For normal T1 images, the "Segment" module in the SPM toolkit cat12 was used on the MATLAB platform to preprocess the T1 images, perform bias field correction, skull stripping, tissue segmentation, and extract gray matter volume. (2) The purpose of bias field correction is to eliminate the unevenness of low- and medium-frequency intensity in the image and improve the accuracy of image segmentation. In the "Segment" module of SPM, check the "Bias Correction" option to perform bias field correction based on the Gaussian mixture model and Markov random field algorithm; (3) Skull stripping: removing non-brain tissue (such as skull, scalp, dura mater, etc.) from the head image, leaving only the brain parenchyma (gray matter, white matter and cerebrospinal fluid).

[0040] Step 3: Extracting brain gray matter volume. The process includes tissue segmentation, image normalization, extraction of brain gray matter, and smoothing of the extracted brain gray matter. (1) Tissue segmentation: In the "Segment" module of SPM, input the AAL brain atlas into the "Atlas" to generate brain gray matter, brain gray matter, and cerebrospinal fluid tissue probability maps. Check the image segmentation quality based on the report file generated in this step. For images with an image segmentation quality grade of A or B, perform subsequent analysis. (2) Image normalization: The gray matter images were normalized to the Montreal Neurological Institute space using the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) algorithm with a voxel size of 2 mm × 2 mm × 2 mm to eliminate individual structural differences. (3) Extracting gray matter volume: In the Segment module of SPM, check “ModulatedNomalized” in Grey Matter to output the normalized gray matter volume; (4) Smoothing: A 6 mm × 6 mm × 6 mm full-width at half-maximum Gaussian kernel was used to smooth the gray matter image to eliminate the differences in anatomical structures among individuals.

[0041] Step 4: IDSCN network construction: ; ; Because the patient's age and total brain volume are the most important factors affecting the gray matter volume, the Pearson correlation coefficient between the gray matter volumes of paired brain regions of n healthy controls (HCs) was calculated using the patient's age and total brain volume as covariates to construct a 90×90 reference network, which is expressed as ; The patient Added to the HCs group, using the same method, based on A new 90 × 90 structured covariance network was constructed based on the gray matter volume data of n subjects (n HCs and 1 patient), denoted as the perturbation network ; Calculate the difference between the perturbation network and the reference network, expressed as ,calculate The Z value of is defined as the connecting edge.

[0042] Step 5. Using Gretna software, calculate the small-world coefficient within a sparsity threshold range of 0.05 to 0.5. When the small-world coefficient is greater than 1.1, the IDSCN network is considered to have the small-world property. Further calculations are performed on the node properties of the IDSCN network within this sparsity threshold range. Using Gretna software, with a sparsity threshold range of 0.05 to 0.4 and a step size of 0.01, the area under the curve (AUC) of the IDSCN node properties is calculated at each sparsity level. Node properties include betweenness centrality, degree centrality, clustering coefficient, node efficiency, local efficiency, and characteristic path length.

[0043] Step 6: Statistical analysis: To investigate the differences in node attributes specific to SZ, OCD, and BD patients, the node attributes of the SZ, OCD, and BD patient groups were compared with those of HCs using a two-sample t-test, with a statistical standard of P < 0.05 (Bonferroni correction).

[0044] Step 7: Statistically significant node attribute values ​​were extracted from the patient group. These node attribute values ​​were used as features for support vector machine (SVM) analysis using LIBSVM software. The "leave one out" method was used to validate SVM performance. Specifically, one sample served as the test set, and the remaining samples served as the training set. The training set was used to build the SVM model, and the test set was used to test the model's classification performance. The number of repeated classification and validation runs matched the number of samples classified. To verify the effectiveness of the classification strategy, 5000 permutation tests were repeated to assess overall classification accuracy. Model performance was evaluated using accuracy, sensitivity, specificity, and the area under the receiver operating characteristic (ROC) curve.

[0045] By statistically analyzing the local node topology properties of the IDSCN network, we revealed brain network changes unique to SZ, OCD, and BD patients. Using these features as features, we achieved classification accuracy rates exceeding 70% between patients and healthy controls, and between patients, demonstrating application value in the classification of psychiatric disorders. Table 1 illustrates the classification results based on the specific node attribute changes in the IDSCN of SZ patients, and the corresponding ROC curves for the classification of SZ patient-specific node attributes are shown in Figure 1. Figure 2 shown.

[0046] Table 1 Classification results based on changes in IDSCN-specific node attributes in SZ patients

[0047] where SZ: schizophrenia, OCD: obsessive-compulsive disorder, BD: bipolar disorder, HCs: healthy controls, IDSCN: individual difference structural covariation network, aBC: area under the curve of betweenness centrality, and aEfficiency: area under the curve of node efficiency.

[0048] The classification results based on the changes in IDSCN-specific node attributes of OCD patients are shown in Table 2 , and the ROC curve for the classification of OCD patient-specific node attributes is shown in Table 2 . Figure 3 shown.

[0049] Table 2 Classification results based on changes in IDSCN-specific node attributes in OCD patients

[0050] Where, aDC: area under the degree centrality curve.

[0051] The classification results based on the changes in IDSCN-specific node attributes of BD patients are shown in Table 3 , and the ROC curve for the classification of BD patients’ specific node attributes is shown in Table 3 . Figure 4 shown.

[0052] Table 3 Classification results based on changes in IDSCN-specific node attributes in BD patients

[0053] Those skilled in the art will appreciate that embodiments of the present disclosure may be provided as methods, systems, or computer program products. Thus, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code. The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes. These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure and are not intended to limit its scope of protection. Although the present disclosure has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that after reading the present disclosure, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the disclosed claims.

Claims

1. A method for classifying mental illness based on individual difference structural covariation networks and machine learning, characterized in that: The method comprises: Step 1: Acquire structural magnetic resonance T1-weighted images and perform image preprocessing, including data format conversion, bias field correction, and skull stripping; Step 2: Segment the preprocessed T1 image into brain regions based on the AAL brain atlas and extract the gray matter volume of each brain region; Step 3: Construct the IDSCN network by calculating the Pearson correlation coefficient of the gray matter volume of paired brain regions; Step 4: Calculate the area under the node topology attribute curve of the IDSCN network within the sparsity threshold range of 0.05-0.4; Step 5: Use the two-sample t-test to screen the node attribute indicators with statistical differences between the patient group and the healthy control group; Step 6: Use the selected node attribute indicators as feature vectors and input them into the support vector machine classification model for disease classification.

2. The method for classifying mental illnesses based on individual difference structural covariation networks and machine learning according to claim 1, characterized in that: The pre-processing in step 1 further comprises: The DARTEL algorithm was used for brain region registration and gray matter volume normalization; Gray matter images were spatially smoothed using a 6-mm full-width-at-half-maximum Gaussian kernel; Gray matter volume was corrected using total brain volume and age as covariates.

3. The method for classifying mental illnesses based on individual difference structural covariation networks and machine learning according to claim 1, characterized in that: The step 2 includes: The Pearson correlation coefficient matrix PCCn of the gray matter volumes of paired brain regions of n healthy subjects was calculated as the reference network; The target subject is added to the reference group, and the Pearson correlation coefficient matrix PCCn+1 of n+1 subjects is calculated as the perturbation network; The difference between the perturbation network and the reference network was calculated and Z-score normalized to obtain the individual difference network.

4. The method for classifying mental illnesses based on individual difference structural covariation networks and machine learning according to claim 3, characterized in that: The Z-score normalization includes: 。 5. The method for classifying mental illnesses based on individual difference structural covariation networks and machine learning according to claim 1, characterized in that: The method for determining the sparsity threshold in step 4 includes: Calculate the small world coefficient in the range of 0.05-0.5; The maximum sparsity range corresponding to the small-world coefficient > 1.1 is selected as the calculation threshold interval.

6. The method for classifying mental illnesses based on individual difference structural covariation networks and machine learning according to claim 1, characterized in that: The SVM classification model adopts an optimized combination of a linear kernel function and a regularization parameter C, and the results are verified by a leave-one-out cross-validation strategy.

7. The method for classifying mental illnesses based on individual difference structural covariation networks and machine learning according to claim 1, characterized in that: The node attributes include: The area under the curve of betweenness centrality in the bilateral middle frontal gyrus; The area under the node efficiency curve of the left calcarine groove; At least one combination of the area under the degree centrality curve of the right putamen.

8. A mental illness classification system based on individual difference structural covariation network and machine learning, characterized by: The system comprises: The preprocessing unit is used to acquire structural magnetic resonance T1-weighted images and perform image preprocessing, including data format conversion, bias field correction, and skull stripping; The segmentation unit is used to segment the preprocessed T1 image into brain regions based on the AAL brain atlas and extract the gray matter volume of each brain region; IDSCN network construction unit, used to construct the IDSCN network by calculating the Pearson correlation coefficient of the gray matter volume of paired brain regions; A calculation unit, used for calculating the area under the curve of the node topology attribute of the IDSCN network within a sparsity threshold range of 0.05-0.4; A screening unit is used to screen the node attribute indicators with statistical differences between the patient group and the healthy control group through a two-sample t test; The classification unit is used to input the filtered node attribute indicators as feature vectors into the support vector machine classification model for disease classification.

9. A computer device, characterized in that: The invention comprises a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes a method for classifying mental illnesses based on individual difference structural covariation network and machine learning according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of a method for classifying mental illnesses based on individual difference structural covariation networks and machine learning as described in any one of claims 1 to 7.

Images