Structural network-genetic map biological network model for predicting ischemic stroke and construction method thereof

Abstract

The invention relates to a structural network-genetic map biological network model for predicting ischemic stroke and a construction method thereof, and the method comprises the steps: extracting and calculating seven multi-scale morphological features and pairwise Pearson correlation coefficients among the features from T1 weighted imaging data and diffusion tensor imaging data; constructing a 308 * 308 morphological similarity network matrix and a brain network module for identifying ischemic stroke neural dysfunction; 1782 sampling points are extracted from the Airy human brain map, and each sampling point comprises expression data of 10185 genes; the method comprises the following steps: mapping space coordinates of AHBA sampling points to a cortex package of a Desikan-Killiany map, carrying out normalization processing to output 308 * 10185 brain region gene-by-gene expression matrixes, and constructing a structural network-gene map biological network model for predicting ischemic stroke by adopting a partial least square regression method and a bootstrap method. Compared with the prior art, the model determines the specific molecular mechanism related to the phenotypic structure change of ischemic stroke injury, and the stroke occurrence probability is predicted according to the specific molecular mechanism.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, specifically to a structural network-gene map biological network model for predicting ischemic stroke and its construction method. Background Technology

[0002] Ischemic stroke is a leading cause of death and disability, and early prevention, diagnosis, treatment, and prognostic functional assessment of IS are crucial for reducing its mortality and disability rates. Currently, clinical diagnosis of IS primarily relies on clinical manifestations and neuroimaging techniques, supplemented by neurovascular imaging data or computed tomography (CT), magnetic resonance imaging (MRI), ultrasound, or angiography. However, this approach is insufficient for early prevention, diagnosis, treatment, and prognosis. Traditional neuroimaging studies often only rule out hemorrhagic stroke and assess ischemic stroke, but they cannot characterize changes in the overall brain network, failing to capture the full dynamic picture of ischemic stroke and thus cannot serve as early biomarkers for predicting IS severity and prognosis. Therefore, identifying these biomarkers is key to reducing IS mortality and disability rates.

[0003] Morphological similarity networks (MSNs) are mathematical models used to describe the degree of similarity in "structural features" among different regions of the brain. They quantify the inter-regional covariance of multimodal structural features and may provide a new way to identify biomarkers. However, their application in brain symmetry (IS) and their underlying molecular mechanisms remain largely unexplored. Summary of the Invention

[0004] The purpose of this invention is to provide a structural network-gene map biological network model for predicting ischemic stroke and a method for constructing the same. The aim is to use the structural network-gene map biological network model for predicting ischemic stroke to predict the risk of ischemic stroke, and to determine core biological pathways and potential targets based on the connectivity within gene modules and their relationship with phenotypic changes.

[0005] To achieve the above objectives, this invention provides a method for constructing a structural network-gene map biological network model for predicting ischemic stroke, comprising: S100: Extract seven multi-scale morphological features from T1-weighted imaging data and diffusion tensor imaging data, calculate the pairwise Pearson correlation coefficient between the morphological vectors, construct a 308×308 morphological similarity network matrix and a brain network module for identifying neurological deficits in ischemic stroke. S200 extracts and incorporates 1782 sampling points from the Allen brain map, with each sampling point containing expression data for 10185 genes; The spatial coordinates of the S300 and AHBA sampling points are mapped onto the cortical wrapper of the Desikan-Killiany map. After normalization, a gene expression matrix of 308×10185 brain regions is output. Partial least squares regression and bootstrap methods are used to construct a structural network-gene map biological network model for predicting ischemic stroke.

[0006] Preferably, in the above technical solution, in step S1, the steps of extracting seven multi-scale morphological features from T1-weighted imaging data and diffusion tensor imaging data, calculating the pairwise Pearson correlation coefficients between the morphological vectors, constructing a 308×308 morphological similarity network matrix, and creating a brain network module for recognizing neurological deficits in ischemic stroke are as follows: S110: Read T1-weighted imaging data, preprocess the T1-weighted imaging data into a standardized brain structure image, and segment the cerebral cortex of the brain structure image into 308 regions; S120: Read diffusion tensor imaging data, preprocess the diffusion tensor imaging data, calculate and generate the diffusion tensor matrix, anisotropy score, and average diffusion rate, compare, identify, and output the anisotropy score and average diffusion rate of white matter regions with significant group differences; S130: Using the 308 regions defined in step S110, extract 7 multi-scale morphological features of each region from the T1-weighted imaging data and diffusion tensor imaging data, calculate the pairwise Pearson correlation coefficient between the morphological vectors, construct a 308×308 morphological similarity network matrix, and identify the brain network module for ischemic stroke neurological deficits.

[0007] Preferably, in the above technical solution, step S110, the step of reading T1-weighted imaging data, preprocessing the T1-weighted imaging data into a standardized brain structure image, and segmenting the cerebral cortex of the brain structure image into 308 regions, specifically includes: S111, Read T1-weighted imaging data, and perform motion correction, skull dissection, spatial normalization, and gray-white matter segmentation on the T1-weighted imaging data to obtain a standardized brain structure image; S112, Segment the cerebral cortex of the brain structure image into 308 regions according to the Desikan-Killiany atlas, and segment each hemisphere into 34 gray matter regions.

[0008] Preferably, in the above technical solution, step S120, which involves reading diffusion tensor imaging data, preprocessing the diffusion tensor imaging data, calculating and generating the diffusion tensor matrix, anisotropy fraction, and average diffusion rate, and comparing, identifying, and outputting the anisotropy fraction and average diffusion rate of white matter regions with significant group differences, specifically includes: S121, Read the diffusion tensor imaging data, and perform format conversion, head motion correction, eddy current correction, and skull dissection processing on the diffusion tensor imaging data to obtain preprocessed diffusion tensor imaging data; S122, Calculate the diffusion tensor on the preprocessed diffusion tensor imaging data using the DTIFIT algorithm and generate the diffusion tensor matrix, anisotropy fraction, and average diffusion rate for each voxel; S123, Compare group-level whole-brain statistics using the Randomise function to rank groups with significant thresholds; S124, Based on the FSL white matter map, identify the white matter regions of the groups with significant thresholds and filter out the white matter regions; S125, Extract the FA and MD values ​​of the white matter regions.

[0009] Preferably, in the above technical solution, in step S130, the steps of extracting 7 multi-scale morphological features of each of the 308 regions defined in step S110 from T1-weighted imaging data and diffusion tensor imaging data, calculating the pairwise Pearson correlation coefficients between the morphological vectors, constructing a 308×308 morphological similarity network matrix, and identifying the brain network module for ischemic stroke neurological deficits are as follows: S131, Map the diffusion tensor imaging data to the T1-weighted imaging data space, and extract multi-scale morphological features for each of the 308 regions defined by the Desikan-Killiany atlas from the diffusion tensor imaging data and the T1-weighted imaging data; S132, Normalize each of the multi-scale morphological features with zero mean, and output standardized multi-scale morphological features; S133, Read the multi-scale morphological features from step S132, calculate the pairwise Pearson correlation coefficients between the multi-scale morphological features, construct a 308×308 morphological similarity network matrix and a brain network module for identifying ischemic stroke neurological deficits, wherein the brain network module for identifying ischemic stroke neurological deficits is MSN = intercept + β1 × (gender) + β2 × (age) + β3 × (gender × age).

[0010] Preferably, in the above technical solution, in step S200, the step of extracting and incorporating 1782 sampling points from the Allen brain map, with each sampling point containing expression data of 10185 genes, is as follows: S210, Gene annotation processing is performed on the Allen Human Brain Atlas whole brain transcriptome dataset; S220, Probe selection is performed on the data output from step S210; S230, Normalization processing is performed on the data output from step S220; S240, 1782 sampling points are extracted and incorporated from the data output from step S230, and each sampling point contains expression data of 10185 genes.

[0011] Preferably, in the above technical solution, in step S300, the spatial coordinates of the AHBA sampling points are mapped onto the cortical wrapper of the Desikan-Killiany map, and the normalized processing outputs a 308×10¹⁸⁵ brain region gene expression matrix. The steps of constructing a structural network-gene map biological network model for predicting ischemic stroke using partial least squares regression, bootstrap method, and other methods are as follows: S310: The spatial coordinates of AHBA sampling points were mapped onto the cortical wrapper of the Desikan-Killiany atlas. The data were normalized, and a gene-by-gene expression matrix of brain regions (number of subjects × 308 × 10185) was output. S320: Partial least squares regression was used to study the relationship between morphological features of ischemic stroke regions and the transcriptional activity of 10,185 genes, resulting in 10 PLS1 components. S330: The bootstrap method was used to estimate the standardized weight Z of each gene in PLS1, setting Z>8 and Z<-8, obtaining Z>8 for 515 genes and Z<-8 for 762 genes. Since Z < -8, 1277 significantly differentially expressed genes are output; S340, a protein-protein interaction network is constructed using the 1277 significantly differentially expressed genes. The protein-protein interaction network consists of 1189 nodes and 389 edges, with an average node degree of 0.654 and an average local clustering coefficient of 0.216; S350, gene ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis is performed on the protein-protein interaction network to screen out specific molecular mechanisms within gene modules related to phenotypic changes, and construct a structural network-gene map biological network model for predicting ischemic stroke.

[0012] Preferably, in the above technical solution, the 10 PLS1 components are URM1, SYCP2, NCOA3, EEPD1, PECR, CHCHD6, PID1, GMFB, PKIA and TMEM263.

[0013] Preferably, in the above technical solution, specific molecular mechanisms within gene modules related to phenotypic changes are screened out, and the specific molecular mechanisms are cellular responses to metal ions, calcium signaling, and cAMP signaling.

[0014] A computer device includes: a processor and a memory storing program instructions thereon, the program instructions, when executed by the processor, causing the processor to perform the method for constructing a structural network-gene map biological network model for predicting ischemic stroke as described in any one of claims 1 to 8.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: By combining MSNs with genes, a gene-by-gene expression matrix of 308 × 10¹⁸5 brain regions was constructed. By setting the threshold of gene Z-scores to ±8, 1277 significantly differentially expressed genes were screened from 10¹⁸5 genes to construct a protein-protein interaction network. GO and KEGG pathway enrichment analysis was used to explore the relationship between phenotypic structural changes in ischemic stroke injury and specific molecular mechanisms. Specific molecular mechanisms were identified, including cellular responses to metal ions, calcium signaling, and cAMP signaling. Finally, a structural network-gene map biological network model for predicting ischemic stroke was established.

[0016] This structural network-gene map biological network model for predicting ischemic stroke injury can accurately predict the probability of ischemic stroke by detecting changes in core modules such as cellular response to metal ions, calcium signaling, and cAMP signaling. Attached Figure Description

[0017] Figure 1 The bar chart (A) shows the normal distribution of morphological similarity between the case group and the control group, and the box plot (B) shows the overall morphological similarity distribution between the case group and the control group in this invention. Figure 2 The brain structure diagrams (A) and (B) are for the analysis of the average morphological similarity of the control group and the t-statistic. Figure 3 This is a diagram showing the regions where there are significant differences in morphological similarity between the case group and the control group in this invention; Figure 4 This is a graph showing the correlation and t-statistic analysis between the average regional morphological similarity between the control group and the case control group in this invention; Figure 5 This is a graph showing the correlation between morphological similarity and NIHSS scores in the case groups of this invention. Figure 6 This is a graph showing the correlation between morphological similarity and 90-day mRS scores in the case groups of this invention. Figure 7 This is a graph showing the Yeo network difference analysis between the case and control groups of this invention; Figure 8 This is a graph showing the von network difference analysis of the case and control groups in this invention; Figure 9 This is a graph showing the ROC analysis of the diagnostic value of morphological similarity to AIS in this invention; Figure 10 The graph shows the explanatory power of PLS ​​components for the differences in morphological similarity between case and control groups (A) and the graph shows the cumulative explanatory power of PLS ​​genomic components for the morphological similarity network (B). Figure 11 This is a graph showing the stability permutation test of the PLS results of this invention; Figure 12 Figure (D) shows the average z-value of whole-brain gene expression in this invention. Figure (E) shows the difference in morphological similarity between the left hemisphere of the case group and the control group. Figure 13 This is a graph showing the correlation between the average score of 308 brain regions and the T-value in this invention; Figure 14 This is a graph showing the bootstrap distribution frequency of the z-score of the PLS1 gene; Figure 15 This is a graph showing a significant correlation between gene expression related to morphological similarity in this invention; Figure 16 This is a diagram showing the regional morphological differences between genes exhibiting significant positive expression in PLS1 gene expression and the case and control groups in this invention; Figure 17 This is a diagram showing the genes and regions with significant negative expression in PLS1 gene expression in this invention, along with morphological differences. Figure 18 This is a diagram showing the PPI network construction of the PLS+ gene (Z value > 8) and the PLS- gene (Z value < -8) of this invention; Figure 19 This is a diagram of gene ontology enrichment analysis according to the present invention; Figure 20 This is a diagram of pathway enrichment analysis from the Kyoto Encyclopedia of Genes and Genomes of this invention; Figure 21 This is a schematic diagram of the structure of an electronic device that implements a preferred embodiment of the present invention for constructing a structural network-gene map biological network model for predicting ischemic stroke. Detailed Implementation

[0018] The technical solutions in the embodiments of this invention are clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0019] This study complies with the Declaration of Helsinki, and the research protocol was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi University of Chinese Medicine (Ethics No.: 2022-048-02). The case group consisted of patients with acute ischemic stroke hospitalized in the Department of Neurology of the First Affiliated Hospital of Guangxi University of Chinese Medicine from June 2023 to November 2024, while the control group consisted of residents who participated in health checkups in the surrounding communities during the same period. Written informed consent was obtained from each participant before registration.

[0020] The inclusion criteria for the case group were: (1) first stroke with hemiplegia lasting 72 h, meeting the diagnostic criteria for cerebral infarction; (2) right-handed; (3) age 18-80 years; (4) infarct lesion located in the unilateral basal ganglia and / or coronary artery region; (5) NIHSS score 1-25; (6) no impairment of consciousness and relatively stable condition. The exclusion criteria were: (1) planned or received intravenous thrombolysis or endovascular treatment for the stroke (including mechanical thrombectomy, arterial thrombolysis or angioplasty); (2) stroke secondary to brain tumor, traumatic brain injury, hematologic disease or other non-vascular causes; (3) comorbidities such as claudication, osteoarthritis, rheumatoid arthritis, gouty arthritis, etc., causing limb motor dysfunction and affecting neurological assessment; (4) severe liver and kidney dysfunction; (5) comorbidities with other life-threatening diseases and expected survival time of less than 3 months; (6) pregnancy or breastfeeding.

[0021] The inclusion criteria for the control group were: (1) healthy, with no history of brain injury, stroke, dementia, schizophrenia or other neurological or mental illness; and (2) matched with the case group based on age, sex, years of education and right-handedness.

[0022] The following data were collected from each participant: name, gender, age, education level; medical history, including stroke, diabetes, and hypertension; and smoking and alcohol consumption. The results are shown in Table 1 below.

[0023] Table 1. Comparison of characteristics between the case group and the control group

[0024] As shown in Table 1, this study included 36 patients with acute ischemic stroke and 36 healthy controls. There were no statistically significant differences between the two groups in terms of gender, age, years of education, smoking history, alcohol consumption history, hypertension, and diabetes. P >0.05).

[0025] On the day of enrollment, participants were assessed for the severity of neurological deficits in each ischemic stroke patient using the National Institutes of Health Stroke Scale (NIHSS). Short-term neurological outcomes were assessed via telephone follow-up 90 days after onset using the modified Rankin Scale (mRS).

[0026] LMAGING Data Acquisition: Within 24 hours of enrollment, participants underwent scanning on a Siemens Novus 3.0T superconducting MRI system equipped with an 8-channel phased array head coil. Acquired sequences included T1-weighted and diffusion tensor imaging (DTI). Imaging parameters were as follows: (1) T1W: TR = 1900 ms, TE = 2.13 ms, matrix = 128×128, FOV = 256×256 mm², voxels = 1.0×1.0×1.0 mm³, 176 slices, slice thickness = 1 mm; (2) DTI: TR = 11000 ms, TE = 94 ms, b-value = 0 and 1000 s / mm², matrix = 128×128, FOV = 256×256 mm², slice thickness = 2 mm; diffusion gradients were applied in 30 directions. During the scan, participants remained supine, awake, and relaxed. The head was stabilized with soft foam padding. Participants were instructed to minimize head movement and wear earplugs to reduce scanner noise.

[0027] S100: Extract seven multi-scale morphological features from T1-weighted imaging data and diffusion tensor imaging data, calculate the pairwise Pearson correlation coefficients between the morphological vectors, construct a 308×308 morphological similarity network matrix and a brain network module for identifying neurological deficits in ischemic stroke.

[0028] S110, Read T1-weighted imaging data, preprocess the T1-weighted imaging data into a standardized brain structure image, and segment the cerebral cortex of the brain structure image into 308 regions. The specific steps are as follows: S111, Read T1-weighted imaging data, and perform motion correction, skull dissection, spatial normalization and gray-white matter segmentation on the T1-weighted imaging data to obtain a standardized brain structure image; The participants' MRI images were imported into ITK-SNAP (version 3.8.0). ITK-SNAP's semi-automatic segmentation method was used to segment the cerebral infarction lesions in the participants' MRI images, generating a binary segmentation mask (lesion areas as 1, other areas as 0). The number of voxels with values ​​of 1 in the mask was counted. The volume of a single voxel is 1.0 × 1.0 × 1.0 mm³. The volume of the cerebral infarction lesions in the MRI images was quantified according to the formula: volume = number of voxels * volume of a single voxel. The spatial location of all lesions in the right hemisphere was unified to the same hemisphere.

[0029] T1-weighted data were preprocessed using FreeSurfer v6.0 to generate standardized brain structure images. The preprocessing workflow is as follows: (1) Motion correction is performed on the T1-weighted image to eliminate motion artifacts and the influence of motion on subsequent processing results; (2) The T1-weighted image in step (1) is subjected to skull stripping, that is, the skull, scalp and other non-brain tissue structures in the image are separated, and only the brain tissue is retained; (3) Spatial normalize the T1-weighted images in step (2), that is, map the images to a standard brain template space so that the T1-weighted images of different participants have the same spatial coordinates and size; (4) Perform gray-white matter segmentation on the T1 weighted image in step (3), that is, segment the brain tissue in the image into different tissue types such as gray matter, white matter and cerebrospinal fluid according to the gray value, and obtain a standardized brain structure image.

[0030] S112, Based on the Desikan-Killiany atlas, the cerebral cortex of the brain structure image is divided into 308 regions, and each hemisphere is divided into 34 gray matter regions. The specific method is as follows: Based on the Desikan-Killiany atlas, FreeSurfer segmented the cerebral cortex of the standardized brain structure image into 308 regions and each hemisphere into 34 gray matter regions.

[0031] S120: Read the diffusion tensor imaging data, preprocess the diffusion tensor imaging data, calculate and generate the diffusion tensor matrix, anisotropy fraction, and average diffusion rate, compare, identify, and output the anisotropy fraction and average diffusion rate of white matter regions with significant group differences. The specific steps are as follows: S121, Read the diffusion tensor imaging data, and perform format conversion, head motion correction, eddy current correction, and skull dissection processing on the diffusion tensor imaging data to obtain preprocessed diffusion tensor imaging data. The specific method is as follows: The DTI data were analyzed using the Region-Based Spatial Statistics (TBSS) module from the Oxford University FSL software package. The preprocessing workflow included format conversion, head motion correction, eddy current correction, and skull dissection to obtain preprocessed DT1 data.

[0032] S122, The DTIFIT algorithm is used to calculate the diffusion tensor of the preprocessed diffusion tensor imaging data and generate the diffusion tensor matrix, anisotropy fraction, and average diffusion rate for each voxel. The specific method is as follows: The DTIFIT algorithm in FDT is used to calculate the diffusion tensor of the preprocessed DT1 data. The diffusion tensor matrix of each voxel is obtained by the algorithm, and the anisotropy fraction (FA) and mean diffusion rate (MD) are generated.

[0033] S123, use the Randomise function to compare group-level whole-brain statistics and rank the groups with significance thresholds. The specific method is as follows: Randomise function was used for group-level whole-brain statistical comparisons, with 5000 rows arranged. After multiple comparison correction, the significance threshold was P<0.05.

[0034] S124, Based on the FSL white matter map, identify the group-level white matter regions with significant thresholds, screen out the white matter regions, and extract the FA and MD values ​​of the white matter regions. The specific method is as follows: Based on the FSL white matter map in FSLView, the group-level white matter regions with significant thresholds are identified, white matter regions with significant group differences are screened out, and the corresponding FA and MD values ​​of the white matter regions are extracted.

[0035] S130: Using the 308 regions defined in step S110, extract 7 multi-scale morphological features for each region from T1-weighted imaging data and diffusion tensor imaging data, calculate the pairwise Pearson correlation coefficients between the morphological vectors, construct a 308×308 morphological similarity network matrix and a brain network module for identifying ischemic stroke neurological deficits.

[0036] S131, the diffusion tensor imaging data is mapped to the T1-weighted imaging data space. Using the 308 regions defined by the Desikan-Killiany map, multi-scale morphological features of each region are extracted from the diffusion tensor imaging data and the T1-weighted imaging data. The specific steps are as follows: Read T1-weighted imaging data and diffusion tensor imaging data, convert them into a unified format, take any voxel coordinate of the T1 image as the target coordinate y, identify the corresponding voxel coordinate of the DTI data as the coordinate to be transferred x, and adopt the rigid transformation model y=Rx+t. The evolution of the transformation parameters from x to y follows the following equation:

[0037] Where Rk+1 is the final rotation value of the coordinate to be transferred x to the target coordinate y, tk+1 is the final translation value of the coordinate to be transferred x to the target coordinate y, Rk is the rotation matrix parameter of the current DTI data, tk is the translation parameter of the current DTI data, α is the learning rate, and the partial derivative of the metric function MI with respect to the rotation parameters (Euler angles θ, ϕ, ψ) is:

[0038] The metric function MI is related to the translation parameter (t) x ,t y ,t z Partial derivatives of )

[0039] The metric function MI is defined as follows:

[0040] Where pT1(g) is the gray-level probability distribution of the T1 image, pDTI(h) is the gray-level probability distribution of the DTI image, and Pjoint(g,h) is the joint probability distribution.

[0041] Based on the above steps, DTI data is mapped to T1 data space to form a three-dimensional space of T1WI and DTI.

[0042] Multiscale morphological features were calculated and extracted from T1WI and DTI for 308 regions defined by the Desikan-Killiany map. These multiscale morphological features included surface area (SA), cortical thickness (CT), gray matter volume (GM), Gaussian curvature (GC), mean curvature (MC), fractional anisotropy (FA), and mean diffusion (MD). A total of 278 features were extracted for each participant.

[0043] S133: Read the multi-scale morphological features from step S132, calculate the pairwise Pearson correlation coefficients between the multi-scale morphological features, construct a 308×308 morphological similarity network matrix with 308 regions as nodes and the pairwise Pearson correlation coefficients as edge weights, and output the brain network module for identifying ischemic stroke neurological deficits. The brain network module for identifying ischemic stroke neurological deficits is MSN = intercept + β1 × (gender) + β2 × (age) + β3 × (gender × age). The specific method is as follows: The multi-scale morphological features of each region are read, and each multi-scale morphological feature is standardized using zero-mean normalization (z-score transformation) to minimize the variability between features and improve the robustness of multi-feature fusion analysis.

[0044] Pairwise Pearson correlation coefficients were calculated between the morphological feature vectors of all standardized cortical regions. A 308×308 morphological similarity network (MSN) matrix was generated for each subject, with 308 regions as nodes and the pairwise Pearson correlation coefficients as edge weights. Using MSN as the independent variable and gender and age as covariates, multivariate logistic regression was applied to identify brain network modules significantly associated with neurological deficits in ischemic stroke, based on the following model: MSN = intercept + β1×(gender) + β2×(age) + β3×(gender×age), where the intercept is 0.85, β1 is -0.04, β2 is 0.35, and β3 is -0.02.

[0045] The frequency distribution of global morphological similarity follows a normal distribution. Figure 1(A) It can be seen that the morphological similarity of the case group (Stroke) is lower than that of the control group (Controls), showing a higher frequency in the pattern and a lower frequency in extreme cases. The Kolmogorov-Smirnov test shows that... Figure 1 (B), from Figure 1 (B) It can be seen that there is a statistically significant difference in the overall morphological similarity distribution between the acute ischemic stroke group (Patients) and the control group (Controls) (P=0.029).

[0046] Analysis of the average morphological similarity of the healthy control group, such as Figure 2 As shown in (A), the positive values ​​of the prefrontal and temporal cortex are higher, while the negative values ​​of the occipital lobe, postcentral gyrus, and frontal pole are higher, indicating the reproducibility of the regional morphological similarity patterns in healthy individuals, with more uniform morphological features within the same region.

[0047] Figure 2 (B) is the t-statistic of regional morphological similarity between the case group and the control group. Figure 3 Regions showing significant differences in morphological similarity between the case and control groups (FD < 0.05 correction). From Figure 2 (B) and Figure 3 We identified 30 cortical regions, among which significant differences in morphological similarity were observed between cases and healthy controls. In the dominant hemisphere, the morphological similarity of the lingual gyrus, postcentral gyrus, and middle frontal gyrus was increased in the stroke group, while the fusiform gyrus, inferior temporal gyrus, precingulate gyrus, and superior frontal gyrus showed decreased morphological similarity. In the right hemisphere, the morphological similarity of the lateral occipital gyrus, lingual gyrus, pericalcarine gyrus, and postcentral gyrus was increased, while the morphological similarity of the fusiform gyrus, lateral orbitofrontal gyrus, paracentral gyrus, pericalcarine gyrus, precuneus, middle frontal gyrus, and superior temporal gyrus was decreased (FDR-corrected P < 0.05; Table 2).

[0048] Table 2

[0049] Figure 4 This involves correlation and t-statistic analysis between the average regional morphological similarity of the control group and the case control group. Figure 4 We found that the average local morphological similarity of the control group was negatively correlated with the case-control t-value in space. That is, the cortical areas with increased morphological similarity in the control group often showed a decreased value in the case-control t-plot, while the areas with decreased similarity in the control group often showed an increased t-value.

[0050] Using the Yeo_2011 cortical segmentation template, 308 regions were divided into seven subnetworks: visual, sensorimotor, ventral attention, dorsal attention, limbic, frontotemporal, and default mode network. Significant differences were found between the stroke group and the control group in the visual, limbic, and default mode networks (P<0.05), while the frontotemporal network showed a decreasing trend (P=0.058). Figure 7 ).

[0051] Based on the von Economo cortical cell structure classification template, 308 regions were further divided into 7 subnetworks: primary sensory cortex, secondary sensory cortex, primary motor cortex, insular cortex, limbic system, and supplementary cortex 1 and 2. Statistically significant differences were found between the stroke group and the control group in the PSS and limbic network (P<0.05). Figure 8 ).

[0052] All statistical analyses were performed using R (version 4.2.2). Descriptive analyses were performed on demographic data. Normally distributed quantitative data are expressed as mean and standard deviation, while non-normally distributed quantitative data are expressed as median and percentile. Categorical data are expressed as frequency and percentile. Receiver operating characteristic (ROC) curves were used to characterize the diagnostic performance of MSNs as biomarkers for acute ischemic stroke in different regions. A higher area under the curve (AUC) indicates better diagnostic performance. Since MSN, NIHSS score, and mRS score do not conform to a normal distribution, Spearman's rank correlation coefficient was used to test their correlation. A two-sided p-value of 0.05 was considered statistically significant. Receiver operating characteristic (ROC) curves were used to evaluate the early diagnostic value of 30 differential areas for ischemic stroke. Ten regions had significant diagnostic value (AUC>0.7), including the fusiform gyrus, cingulate gyrus, and superior frontal gyrus in the left hemisphere, and the fusiform gyrus, lateral frontal lobe, paracentral gyrus, precuneus gyrus, superior frontal gyrus, and superior temporal gyrus in the right hemisphere, as shown in Table 2 below. Figure 9 As shown.

[0053] Table 2. ROC analysis of the diagnostic value of morphological similarity for AIS.

[0054] Note: lh represents the left hemisphere, and rh represents the right hemisphere.

[0055] Spearman correlation analysis was performed on the morphological similarity values ​​of 30 brain regions at 90 days after onset, showing significant differences between NIHSS scores and mRS scores. M-cell similarity values ​​of the left postcentral gyrus, left anterior cingulate gyrus, right postcentral gyrus, and right superior frontal gyrus were significantly correlated with the severity of neurological deficits (NIHSS; P < 0.05). Figure 5Compared with the control group, AIS was significantly different (AUC>0.7). The MSN values ​​of the bilateral postcentral gyrus were significantly correlated with NIHSS scores and 90-day mRS (P<0.05). Furthermore, the morphological similarity values ​​of the left postcentral gyrus, right paratibial gyrus, and right postcentral gyrus were significantly correlated with neurological outcomes (mRS; P<0.05). Figure 6 ).

[0056] The spatial coordinates of the S300 and AHBA sampling points are mapped onto the cortical wrapper of the Desikan-Killiany map. After normalization, a gene expression matrix of 308×10185 brain regions is output. Partial least squares regression and bootstrap methods are used to construct a structural network-gene map biological network model for predicting ischemic stroke.

[0057] S310 maps the spatial coordinates of AHBA sampling points onto the cortical wrapper of the Desikan-Killiany atlas, normalizes the data, and outputs a gene-by-gene expression matrix of brain regions representing the number of subjects × 308 × 10185. The specific method is as follows: Whole-brain gene expression prediction: MSN was correlated with gene expression patterns using the Allen Human Brain Atlas (AHBA) whole-brain transcriptome dataset. Data were sourced from the AHBA repository and contained post-mortem brain samples from six donors (mean age = 42.50 ± 13.38 years; male / female = 5 / 1). The AHBA whole-brain transcriptome dataset underwent data preprocessing following a standardized AHBA pipeline, including the following steps: (1) Gene annotation: Map the original probe IDs of AHBA to the latest gene symbols; (2) Probe selection: (3) Normalization: Normalize the gene expression data to make the expression levels of different genes comparable.

[0058] A total of 1,782 sampling points were extracted, and each sampling point contained expression data of 10,185 genes.

[0059] Constructing a structural network-gene map biological network: The spatial coordinates of AHBA sampling points were recorded within the cortical wrapper of the Desikan-Killiany atlas. To minimize inter-donor variability due to age, ethnicity, and sex differences, robust normalization (SCnorm) was applied to eliminate systematic variations and outliers. The normalized data were merged and aggregated to obtain a dataset, which was then used to construct a 308×10¹⁸⁵ gene expression matrix for brain regions.

[0060] S320, partial least squares regression was used to study the relationship between regional morphological features of ischemic stroke and the transcriptional activity of 10,185 genes, resulting in 10 PLS1 components. The specific method is as follows: Partial least squares regression (PLS) was used to investigate the relationship between regional morphological features in ischemic stroke and the transcriptional activity of 10,185 genes, thereby constructing a biological structural network-gene map. In this model, gene expression profiles serve as predictors of variation in MSN measurement regions. Cross-validation was used to identify the PLS components most closely associated with MSN alterations. Subsequent regression analysis quantified the correlation coefficients and the proportion of variance related to MSN changes explained by each component. This method ultimately identified the gene modules most closely associated with cortical structural damage.

[0061] After preprocessing using the ALLEN map normalization pipeline, we obtained a matrix of 72 subjects × 308 brain regions × 10185 genes. Ten components were derived, and the genes in the PLS1 component explained 28.25% of the MSN variance (permutation test, P < 0.001).

[0062] S330 uses the bootstrap method to estimate the standardized weight Z of each gene in PLS1, setting Z>8 and Z<-8, obtaining Z>8 for 515 genes and Z<-8 for 762 genes, outputting 1277 genes with significant differential expression, as follows: The bootstrap method was used to estimate the standardized weights (z-scores) of each gene in PLS1. The mean scores of 308 brain regions were significantly positively correlated with the T scores (r=0.521, P<0.001). Figure 13 ).

[0063] To further identify differentially expressed genes, we set a threshold for the gene-normalized weighted Z-score (Z = gene weight / bootstrap standard error). The bootstrap distribution of gene Z-scores is shown below. Figure 14 As shown, P=0.05 is marked with a blue line, and P<0.05 when Z±8 is marked with a red line. Based on the normalized weights (z-score) of each gene in PLS1, 515 genes had Z-scores >8, and 762 genes had Z-scores <-8. The five genes most positively correlated with PLS1 (PLS+) were URM1, SYCP2, NCOA3, EEPD1, and PECR. The five genes negatively correlated with PLS1 (PLS−) were CHCHD6, PID1, GMFB, PKIA, and TMEM263. PLS+ genes were positively correlated with T-scores, and PLS− genes were negatively correlated with T-scores. Figure 15 ). Figure 16The diagram shows the regional morphological differences between genes showing significant positive expression in PLS1 (such as URM1) and the case and control groups. Figure 17 This is a diagram showing the morphological differences between genes exhibiting significant negative expression in PLS1 (such as CHCHD6) and their regions. Figure 16 It can be seen that the gene showing significant positive expression in PLS1 is positively correlated with the regional morphological differences between the case group and the control group (r=0.599, p<0.001), while from... Figure 17 It can be seen that the gene with significant negative expression in PLS1 is negatively correlated with the regional morphological differences between the case group and the control group (r=-0.6232, p<0.001).

[0064] S340, a protein-protein interaction network was constructed using the 1277 significantly differentially expressed genes. This network consists of 1189 nodes and 389 edges, with an average node degree of 0.654 and an average local clustering coefficient of 0.216, as detailed below: Principal Component Enrichment Analysis (PCI): The associations between thousands of genes and phenotypes identified in the PLS analysis were aggregated into a gene set, allowing for the extraction of modules associated with cortical injury in ischemic stroke. A PPI network was constructed using STRING 10.5 for genes with PLS1 weights Z%3E-8 and Z%3C-8 (both FDR < 0.05). Specifically, STRING 10.5 was used to convert 1277 significantly differentially expressed genes into STRING-recognized protein IDs, resulting in 1189 mapped and interacting proteins. 398 protein-protein interactions were identified. These 1189 were used as the number of PPI nodes, and the 389 as the number of edges. The average node degree was calculated as (2 × number of edges) / number of nodes, yielding an average node degree of 0.654. ​​The local clustering coefficient was defined as... , where e i It is a node i The number of actual edges between neighboring nodes, k i It is a node i The average local clustering coefficient of this PPI network is 0.216. Therefore, a protein-protein interaction (PPI) network consisting of 1189 nodes and 389 edges was constructed using STRING 10.5. The average node degree of this network is 0.654, and the average local clustering coefficient is 0.216. The PPI enrichment p-value is 0.004. Figure 18 This is a graph showing the PPI network construction for PLS+ genes (Z-value > 8) and PLS- genes (Z-value < -8). Figure 18It can be seen that significant enrichment was mainly observed in genes such as UBA52, PIK3CA, CALML3, PIK3CB, GNG2, and MAPK3. Figure 18 ).

[0065] S350, Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis are performed on the protein-protein interaction network to screen out specific molecular mechanisms within gene modules related to phenotypic changes, and a structural network-gene map biological network model for predicting ischemic stroke is constructed, as follows: The protein-protein interaction (PPI) network was divided into multiple detection modules. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on the gene set of each module. The GO and KEGG enrichment analysis results for each module were obtained. Figure 19 GO enrichment analysis of PLS+ genes (Z-value > 8) and PLS- genes (Z-value < -8) shows that... Figure 19 The GO analysis showed that the enriched biological functions were mainly related to cellular responses to metal ions and inorganic substances, stress responses to metal and copper ions, detoxification of copper ions, regulation of membrane potential, and cellular responses to copper and zinc ions. Figure 20 KEGG enrichment analysis of PLS+ genes (Z-value > 8) and PLS- genes (Z-value < -8) shows that... Figure 20 KEGG pathway enrichment analysis revealed significant enrichment in pathways such as calcium signaling, cAMP signaling, apelin signaling, cardiomyocyte adrenergic signaling, GABAergic synapses, EGFR tyrosine kinase inhibitor resistance, glutamatergic synapses, and Rap1 signaling. Based on GO and KEGG enrichment analyses of each module, modules associated with ischemic stroke phenotypes were identified as metal ion responses, calcium signaling, cAMP signaling, and apelin signaling. These identified modules were then used as core components to construct a structural network-gene map biological network model for predicting ischemic stroke, which was visualized using Cytoscape.

[0066] The structural network-gene map biological network model for predicting ischemic stroke has the following applications: First, it can predict the probability of ischemic stroke by observing changes in the core modules. Second, when new gene expression data is available, the positions of these genes in the network can be calculated to observe which modules they are closer to, or the network diffusion algorithm (such as RandomWalk) can be used to assess the perturbation of the entire network, thereby predicting the risk of ischemic stroke. Third, it can determine core biological pathways and potential targets based on the connectivity within gene modules and their relationship with phenotypic changes. Fourth, it can use core biological pathways and potential targets as indicators for drug use in the treatment of ischemic stroke, making treatment more targeted and thus improving the treatment effect of ischemic stroke.

[0067] Transcriptome analysis revealed that the spatial pattern of MSN abnormalities was significantly correlated with the expression of 1277 genes (PLS1 component explained 28.25% of the variance, P<0.001). These genes were enriched in biological pathways related to stroke pathology, including cellular responses to metal ions, calcium signaling, and cAMP signaling.

[0068] MSNs are highly sensitive biomarkers for quantifying AIS network alterations and have significant diagnostic and prognostic value. Our findings link these structural changes to specific molecular mechanisms—cellular responses to metal ions, calcium signaling, and cAMP signaling—establishing MSNs as a tool for integrating macroscopic network damage and microscopic transcriptomic mechanisms in stroke.

[0069] from Figure 21 An electronic device can be seen that carries the aforementioned structural network-gene map biological network model for predicting ischemic stroke.

[0070] A system for carrying the structural network-gene map biological network model for predicting ischemic stroke, the system comprising: Reading and building module: used to extract 7 multi-scale morphological features from T1-weighted imaging data and diffusion tensor imaging data, calculate the pairwise Pearson correlation coefficient between the morphological vectors, construct a 308×308 morphological similarity network matrix and a brain network module for identifying neurological deficits in ischemic stroke. Extraction module: Extracted and incorporated 1782 sampling points from Allen's brain map, with each sampling point containing expression data of 10185 genes; Analysis and construction module: The spatial coordinates of AHBA sampling points are mapped onto the cortical wrapper of the Desikan-Killiany map. After normalization, a gene expression matrix of 308×10185 brain regions is output. Partial least squares regression and bootstrap methods are used to construct a structural network-gene map biological network model for predicting ischemic stroke.

[0071] The read and build module includes: T1-weighted imaging data processing unit: used to read T1-weighted imaging data, preprocess the T1-weighted imaging data into standardized brain structure images, and divide the cerebral cortex of the brain structure images into 308 regions; Diffusion tensor imaging data processing unit: used to read diffusion tensor imaging data, preprocess the diffusion tensor imaging data, calculate and generate diffusion tensor matrix, anisotropy fraction and average diffusion rate, compare, identify and output the anisotropy fraction and average diffusion rate of white matter regions with significant group differences. Extraction and Construction Unit: This unit extracts seven multi-scale morphological features from the 308 regions defined by the T1-weighted imaging data processing unit and diffusion tensor imaging data for each region. It calculates the pairwise Pearson correlation coefficients between the morphological vectors, constructs a 308×308 morphological similarity network matrix with the 308 regions as nodes and the pairwise Pearson correlation coefficients as edge weights, and outputs a brain network module that identifies neurological deficits in ischemic stroke.

[0072] The T1-weighted imaging data processing unit includes: First processing unit: used to read T1-weighted imaging data, perform motion correction, skull dissection, spatial normalization and gray-white matter segmentation on the T1-weighted imaging data, and output standardized brain structure images; Segmentation unit: used to read the Desikan-Killiany map, and to segment the cerebral cortex of the brain structure image into 308 regions based on the read Desikan-Killiany map, and to segment each hemisphere into 34 gray matter regions.

[0073] The diffusion tensor imaging data processing unit includes: The second processing unit is used to read diffusion tensor imaging data, perform format conversion, head motion correction, eddy current correction and skull dissection on the diffusion tensor imaging data, and output the preprocessed diffusion tensor imaging data. First calculation unit: used to calculate the diffusion tensor of the preprocessed diffusion tensor imaging data using the DTIFIT algorithm and generate the diffusion tensor matrix, anisotropy fraction and average diffusion rate for each voxel; Filtering unit: Used to compare group-level whole-brain statistics using the Randomise function and rank groups with significance thresholds; First identification and extraction unit: used to identify the group-level white matter regions with significant thresholds based on the FSL white matter map, screen out the white matter regions, and extract the FA and MD values ​​of the white matter regions.

[0074] The extraction and construction unit includes: The second identification and extraction unit is used to map diffusion tensor imaging data to T1 weighted imaging data space, and extract multi-scale morphological features of each region from the diffusion tensor imaging data and T1 weighted imaging data, based on 308 regions defined by the Desikan-Killiany map. The third processing unit is used to normalize each of the multi-scale morphological features with zero mean and output standardized multi-scale morphological features. The first construction unit is used to read the multi-scale morphological features in step S132, calculate the pairwise Pearson correlation coefficients between the multi-scale morphological features, construct a 308×308 morphological similarity network matrix and a brain network module for identifying neurological deficits in ischemic stroke. The brain network module for identifying neurological deficits in ischemic stroke is MSN = intercept + β1 × (gender) + β2 × (age) + β3 × (gender × age).

[0075] The analysis and construction module includes: The fourth processing unit is used to map the spatial coordinates of AHBA sampling points onto the cortical wrapper of the Desikan-Killiany map, normalize the data, and output a gene expression matrix of brain regions for the number of subjects × 308 × 10185. The first analysis unit was used to study the relationship between the morphological features of the ischemic stroke region and the transcriptional activity of 10,185 genes using partial least squares regression, resulting in 10 PLS1 components, namely URM1, SYCP2, NCOA3, EEPD1, PECR, CHCHD6, PID1, GMFB, PKIA and TMEM263. The first judgment unit is used to estimate the standardized weight Z of each gene in PLS1 using the bootstrap method. Z>8 and Z<-8 are set, resulting in Z>8 for 515 genes and Z<-8 for 762 genes, outputting 1277 genes with significant differential expression. The second building block is used to construct a protein-protein interaction network using the 1277 significantly differentially expressed genes. The protein-protein interaction network consists of 1189 nodes and 389 edges, with an average node degree of 0.654 and an average local clustering coefficient of 0.216. The third building block is used to perform gene ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis on the protein-protein interaction network, screen out specific molecular mechanisms within gene modules related to phenotypic changes, and construct a structural network-gene map biological network model for predicting ischemic stroke. The specific molecular mechanisms are cellular responses to metal ions, calcium signaling, and cAMP signaling.

[0076] A computer device applies the method for constructing a structural network-gene map biological network model for predicting ischemic stroke to one or more electronic devices. The electronic device is a device that performs numerical calculations and / or information processing and / or model building according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc. The electronic device can be any electronic product capable of human-computer interaction, such as personal computers, tablets, smartphones, personal digital assistants (PDAs), game consoles, interactive network television (IPTV), smart wearable devices, robots, etc. The network in which the electronic device operates includes, but is not limited to, the Internet, wide area networks (WANs), metropolitan area networks (MANs), local area networks (LANs), virtual private networks (VPNs), etc.

[0077] In one embodiment of the present invention, the electronic device includes, but is not limited to, a memory and a processor. The memory stores program instructions that, when executed by the processor, cause the processor to perform the method for constructing a structural network-gene map biological network model for predicting ischemic stroke, and a computer program, such as a data determination program, stored in the memory and executable on the processor. The electronic device may include related components required for program execution, or combinations of certain components, or different components; for example, the electronic device may include input / output devices, network access devices, buses, etc.

[0078] The electronic device may also include network devices and / or user devices. The network devices include, but are not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0079] The processor executes the operating system of the electronic device and various installed applications. The processor executes the applications to implement the steps in the above-described data determination method embodiments, such as steps S100, S200, and S300.

[0080] Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments. For example, it extracts seven multi-scale morphological features from T1-weighted imaging data and diffusion tensor imaging data, calculates the pairwise Pearson correlation coefficients between the morphological vectors, constructs a 308×308 morphological similarity network matrix and a brain network module for identifying neurological deficits in ischemic stroke; it extracts and incorporates 1782 sampling points from the Allen brain atlas, each sampling point containing the expression data of 10185 genes; it maps the spatial coordinates of the AHBA sampling points onto the cortical wrapping of the Desikan-Killiany atlas, normalizes and outputs a 308×10185 gene-by-gene expression matrix of brain regions, and constructs a structural network-gene atlas biological network model for predicting ischemic stroke using partial least squares regression, bootstrap method, etc.

[0081] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the electronic device. For example, the computer program may be divided into a reading and building module, an extraction module, and an analysis and building module.

[0082] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0083] The memory can be external memory and / or internal memory of an electronic device. Further, the memory can be a circuit with storage function that does not have a physical form in an integrated circuit, such as RAM (Random-Access Memory), FIFO (First In First Out), etc. Alternatively, the memory can also be a memory with a physical form, such as a memory module, a TF card (Trans-flash Card), etc.

[0084] If the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above.

[0085] The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0086] This invention can be implemented in various ways and is not limited to the embodiments described. Those skilled in the art will understand that the invention can be implemented in other specific ways without changing the technical concept or essential features. Therefore, it should be understood that the embodiments described above are exemplary and not intended to limit the invention.

Claims

1. A method for constructing a structural network-gene profile biological network model for predicting ischemic stroke, characterized by, The method comprises the following steps: S100, 7 multi-scale morphological features are extracted from T1 weighted imaging data and diffusion tensor imaging data, the pair-wise Pearson correlation coefficients between the morphological vectors are calculated, a 308*308 morphological similarity network matrix is constructed, and a brain network module for identifying neurological impairment of ischemic stroke is identified; S200, 1782 sampling points are extracted from the Allen human brain atlas, and each sampling point contains expression data of 10185 genes; S300, the spatial coordinates of the AHBA sampling points are mapped onto the cortical parcellation of the Desikan-Killiany atlas, normalized to output a 308*10185 brain region gene expression matrix, and a structural network-gene atlas biological network model for predicting ischemic stroke is constructed by using partial least squares regression and bootstrap methods.

2. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke according to claim 1, wherein, In step S1, the 7 multi-scale morphological features are extracted from the T1 weighted imaging data and the diffusion tensor imaging data, the pair-wise Pearson correlation coefficients between the morphological vectors are calculated, a 308*308 morphological similarity network matrix is constructed, and a brain network module for identifying neurological impairment of ischemic stroke is identified. S110, reading T1 weighted imaging data, preprocessing the T1 weighted imaging data into a standardized brain structure image, and segmenting the cerebral cortex of the brain structure image into 308 regions; S120, reading diffusion tensor imaging data, preprocessing the diffusion tensor imaging data, calculating and generating a diffusion tensor matrix, anisotropy fraction and mean diffusivity, comparing, identifying and outputting the anisotropy fraction and mean diffusivity of the white matter region with significant group difference; S130, 7 multi-scale morphological features of each region are extracted from the T1 weighted imaging data and the diffusion tensor imaging data in the 308 regions defined in step S110, the pair-wise Pearson correlation coefficients between the morphological vectors are calculated, a 308*308 morphological similarity network matrix is constructed, and a brain network module for identifying neurological impairment of ischemic stroke is identified.

3. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke according to claim 2, wherein, In step S110, the reading of T1 weighted imaging data, preprocessing the T1 weighted imaging data into a standardized brain structure image, and segmenting the cerebral cortex of the brain structure image into 308 regions are specifically as follows: S111, reading T1 weighted imaging data, performing motion correction, skull stripping, spatial normalization and gray-white matter segmentation on the T1 weighted imaging data to obtain a standardized brain structure image; S112, segmenting the cerebral cortex of the brain structure image into 308 regions according to the Desikan-Killiany atlas, and segmenting each hemisphere into 34 gray matter regions.

4. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke according to claim 2, wherein, In step S120, the reading of diffusion tensor imaging data, preprocessing the diffusion tensor imaging data, calculating and generating a diffusion tensor matrix, anisotropy fraction and mean diffusivity, comparing, identifying and outputting the anisotropy fraction and mean diffusivity of the white matter region with significant group difference are specifically as follows: S121, reading the diffusion tensor imaging data, performing format conversion, head motion correction, eddy current correction and skull stripping on the diffusion tensor imaging data to obtain preprocessed diffusion tensor imaging data; S122, calculating the diffusion tensor of the preprocessed diffusion tensor imaging data using the DTIFIT algorithm and generating the diffusion tensor matrix, anisotropy fraction and average diffusion rate of each voxel; S123, comparing the group-level whole brain statistics using the Randomise function to arrange the group-level with significant threshold; S124, identifying the white matter regions of the group-level with significant threshold based on the FSL white matter atlas, screening out the white matter regions and extracting the FA and MD values of the white matter regions.

5. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke according to claim 2, wherein, In step S130, the 308 regions defined in step S110 extract 7 multi-scale morphological features of each region from the T1 weighted imaging data and the diffusion tensor imaging data, calculate the pair-wise Pearson correlation coefficient between the morphological vectors, construct a 308x308 morphological similarity network matrix and identify the brain network module of ischemic stroke neurological deficit as follows: S131, mapping the diffusion tensor imaging data to the T1 weighted imaging data space, extracting multi-scale morphological features of each region from the diffusion tensor imaging data and the T1 weighted imaging data in the 308 regions defined by the Desikan-Killiany atlas; S132, normalizing each of the multi-scale morphological features with zero mean, and outputting the normalized multi-scale morphological features; S133, reading the multi-scale morphological features in step S132, calculating the pair-wise Pearson correlation coefficient between the multi-scale morphological features, constructing a 308x308 morphological similarity network matrix and identifying the brain network module of ischemic stroke neurological deficit, the brain network module of ischemic stroke neurological deficit is MSN = intercept + β1x (gender) + β2x (age) + β3x (gender x age).

6. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke according to claim 1, wherein, In step S300, the spatial coordinates of the AHBA sampling points are mapped onto the cortical wrapping of the Desikan-Killiany atlas, normalized to output a 308x10185 brain region gene expression matrix, and the steps of partial least squares regression, bootstrap method, and construction of the structural network-gene atlas biological network model for predicting ischemic stroke are as follows: S310, mapping the spatial coordinates of the AHBA sampling points to the cortical wrapping of the Desikan-Killiany atlas, normalizing the data to output a subject number x 308 x 10185 brain region gene expression matrix; S320, using partial least squares regression to study the relationship between the ischemic stroke region morphological features and the transcriptional activity of 10,185 genes, and obtaining 10 PLS1 components; S330, bootstrap method is used to estimate the standardized weight Z of each gene in PLS1, Z>8 and Z<-8 are set, Z>8 of 515 genes and Z<-8 of 762 genes are obtained, and 1277 significant differential expression genes are output; S340, a protein-protein interaction network is constructed by using the 1277 significant differential expression genes, the protein-protein interaction network is composed of 1189 nodes and 389 edges, the average node degree is 0.654, and the average local clustering coefficient is 0.216; S350, gene ontology and Kyoto gene and genome encyclopedia pathway enrichment analysis are performed on the protein-protein interaction network, specific molecular mechanisms in the gene module related to phenotype change are screened out, and a structural network-gene atlas biological network model for predicting ischemic stroke is constructed.

7. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke according to claim 6, wherein, The 10 PLS1 components are URM1, SYCP2, NCOA3, EEPD1, PECR, CHCHD6, PID1, GMFB, PKIA and TMEM263.

8. The method of constructing a structural network-gene profile biological network model for predicting ischemic stroke as claimed in claim 6 wherein, Specific molecular mechanisms in the gene module related to phenotype change are screened out, and the specific molecular mechanisms are cell response to metal ions, calcium signaling and cAMP signaling.

9. The method according to any one of claims 1-8, wherein the method is used to construct a structural network-gene atlas biological network model for predicting ischemic stroke.

10. A computer device, comprising: Comprise: a processor, and a memory having program instructions stored thereon, the program instructions, when executed by the processor, causing the processor to perform the method of any one of claims 1-8.

Images