Image feature selection method for schizophrenia medical record Spark model

Abstract

The invention relates to the technical field of medical information extraction, in particular to an image feature selection method for a schizophrenia medical record Spark model. The method comprises the following steps: acquiring brain magnetic resonance image data and resting state functional magnetic resonance imaging data of a plurality of subjects; performing multi-scale layered segmentation on the brain image of each subject by using a preset layered brain region template for the brain magnetic resonance image data to obtain brain region positioning data including a local level, a regional level and a global level; and performing adaptive hierarchical noise reduction processing on the resting state functional magnetic resonance imaging data in a Spark distributed environment to obtain brain function time sequence data after noise reduction. According to the method, the brain magnetic resonance image data is accurately divided into a local level, a regional level and a global level by using the preset layered brain region template, so that local structure details and global network characteristics of the brain region are considered, and effective integration of multi-scale information is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information extraction, and in particular to an image feature selection method for a schizophrenia medical record Spark model. Background Art

[0002] Schizophrenia is a complex, severe mental disorder with unknown etiology, which is mainly manifested by multiple disorders in perception, thinking, emotion, and behavior, as well as incoordination of mental activities. Spark MLlib provides a variety of feature selection methods suitable for image feature selection scenarios, such as the Filter method to select features based on the statistical relationship between features and target variables, and the Wrapper method to evaluate the importance of features by training models. In the Spark model for schizophrenia medical records, image feature selection is a key step that can help improve the accuracy and efficiency of the model. In the study of schizophrenia, medical images (such as CT, MRI, etc.) provide important brain morphology and functional information. By selecting features related to schizophrenia, we can better understand the pathological mechanism of the disease and improve the diagnosis and treatment effects.

[0003] However, traditional image feature selection methods for the Spark model of schizophrenia medical records often have the following problems: Existing Spark MLlib feature selection methods (such as chi-square test and information gain) are relatively limited when processing medical images, and it is difficult to accurately select the most biologically significant features. The inherent low signal-to-noise ratio and multi-level, complex brain region structure of fMRI data make it difficult for traditional denoising methods and single-scale analysis to simultaneously take into account the accurate extraction of local details and global networks, resulting in unstable analysis results and low reliability. Summary of the invention

[0004] Based on this, it is necessary for the present invention to provide an image feature selection method for the Spark model of schizophrenia medical records to solve at least one of the above technical problems.

[0005] To achieve the above purpose, an image feature selection method for a schizophrenia medical record Spark model includes the following steps: Step S1: acquiring brain magnetic resonance image data and resting-state functional magnetic resonance imaging data of multiple subjects; Step S2: using a preset hierarchical brain region template to perform multi-scale hierarchical segmentation on the brain magnetic resonance image data of each subject, and obtain brain region positioning data including three levels: local, regional and global. The multi-scale hierarchical segmentation specifically includes subcortical structure segmentation at the local level, functional partitioning segmentation based on anatomical connection by calculating the structural connection probability and spatial adjacency between brain regions at the regional level, and large-scale brain network segmentation based on hierarchical organization analysis of resting-state networks at the global level; Step S3: performing adaptive hierarchical denoising on the resting-state functional magnetic resonance imaging data in a Spark distributed environment to obtain denoised brain function time series data; Step S4: construct a hierarchical functional connection calculation framework based on the brain area positioning data and brain function time series data, calculate the functional connection strength within each level and between levels, and dynamically adjust the connection weights between levels based on the principle of synaptic plasticity to obtain brain function network feature data; Step S5: Performing a hierarchical feature importance evaluation based on Fisher score on the brain functional network feature data, and setting a hierarchical adaptive threshold for feature screening according to the pre-acquired schizophrenia prior knowledge to obtain a schizophrenia feature set.

[0006] Preferably, step S1 comprises the following steps: Step S11: performing T1-weighted magnetic resonance scanning on the brains of multiple subjects using a 3T magnetic resonance imaging device, thereby obtaining brain magnetic resonance image data, wherein the imaging contrast of the brain magnetic resonance image data is specifically determined by the T1 relaxation time; Step S12: using a 7T superconducting magnetic resonance imaging device to perform blood oxygen level-dependent contrast acquisition based on a gradient echo sequence on the brains of multiple subjects in a resting state, thereby obtaining blood oxygen level-dependent contrast data; Step S13: Analyze the local blood flow changes caused by neuronal activities on the blood oxygen level dependence contrast data, so as to obtain resting state functional magnetic resonance imaging data.

[0007] Preferably, step S2 comprises the following steps: Step S21: performing image resolution adaptive normalization processing on the brain magnetic resonance image data, and obtaining normalized image data with a spatial resolution of 1 mm isotropic voxel by resampling the voxel size based on B-spline interpolation; Step S22: performing subcortical structure segmentation on the standardized image data according to a preset hierarchical brain region template, and performing brain region positioning processing, thereby obtaining subcortical structure positioning data at a local level; Step S23: performing functional partitioning clustering based on anatomical connection by calculating the structural connection probability and spatial adjacency relationship between brain regions according to the local-level subcortical structure positioning data, thereby obtaining functional partitioning data at the regional level; Step S24: performing a hierarchical organization analysis of the cerebral cortex based on the resting-state network according to the functional partitioning data and the resting-state functional magnetic resonance imaging data, thereby obtaining large-scale brain network data at the global level, wherein the hierarchical organization analysis specifically calculates the time series correlation and spatial continuity between the functional partitions; Step S25: hierarchically integrate the subcortical structure positioning data, functional partitioning data, and large-scale brain network data to establish a correspondence matrix between the hierarchies, thereby obtaining brain region positioning data.

[0008] The present invention realizes the standardization of brain magnetic resonance image data and multi-level brain region positioning through a series of refined processing, providing high-quality basic data for subsequent analysis. First, the resolution of the image data is standardized and unified to 1 mm isocubic voxel format, eliminating the differences caused by equipment and scanning parameters, and providing a unified framework for subsequent analysis. Then, the hierarchical brain region template is used to segment and locate the subcortical structure, accurately identify important structures such as the thalamus and amygdala, and provide local hierarchical information for understanding disease changes. Based on this, functional partitioning clustering is performed by calculating the structural connection probability and spatial adjacency relationship to obtain regional hierarchical functional partitioning data, reflecting the structural and functional relationship of the brain. Further combined with resting state functional magnetic resonance imaging data, the time series correlation and spatial continuity between functional partitions are analyzed to obtain large-scale brain network data at the global level, revealing the overall architecture of the brain functional network. Finally, the data at the local, regional and global levels are integrated, and a matrix of corresponding relationships between levels is established to form a multi-level and multi-dimensional brain region positioning system, which provides comprehensive and high-quality input data for subsequent functional connection analysis and feature screening, ensuring the scientificity and effectiveness of the research method.

[0009] Preferably, step S22 includes the following steps: Step S221: performing edge enhancement processing by calculating the spatial gradient amplitude and direction of the three-dimensional image according to the standardized image data, thereby obtaining boundary enhancement data of the subcortical structure; Step S222: Analyze the grayscale distribution characteristics of the local area according to the boundary enhancement data, set a dynamic threshold, and perform segmentation, thereby obtaining preliminary segmentation mask data of the subcortical structure; Step S223: morphologically optimizing the preliminary segmentation mask data by using morphological operations, so as to obtain continuous and complete subcortical structure mask data, wherein the morphological optimization specifically processes holes and boundary discontinuities by designing a three-dimensional morphological operator sequence; Step S224: Perform connected domain analysis based on the subcortical structure mask data and mark different subcortical structures, so as to obtain fine subcortical structure positioning data at the local level.

[0010] Preferably, step S23 includes the following steps: Step S231: performing connection fiber bundle tracing on the local level subcortical structure positioning data, and performing white matter fiber connection analysis between subcortical structures, thereby obtaining a structural connection probability map; Step S232: Calculating the spatial adjacency relationship between brain regions based on the spatial distance metric according to the structural connection probability map, thereby obtaining brain region spatial relationship data; Step S233: performing weighted fusion on the structural connection probability map and the brain region spatial relationship data based on the structural connection relationship and the spatial adjacency relationship, thereby obtaining a comprehensive relationship matrix; Step S234: partition the comprehensive relationship matrix using a spectral clustering algorithm, thereby obtaining functional partition data at the regional level.

[0011] Preferably, step S25 comprises the following steps: Step S251: extracting the spatial position, volume and shape features of each structure from the subcortical structure positioning data, and constructing a feature vector to obtain a local hierarchical feature matrix; Step S252: Analyze the functional connection pattern and spatial distribution characteristics of each partition according to the functional partition data, so as to obtain a regional level feature matrix; Step S253: extracting network features including topological attributes and dynamic features from the large-scale brain network data, thereby obtaining a global hierarchical feature matrix; Step S254: local level feature matrix, regional level feature matrix and global level feature matrix are processed to obtain brain region localization data.

[0012] Preferably, step S3 comprises the following steps: Step S31: partitioning the resting state functional magnetic resonance imaging data in the Spark distributed environment, dividing the data into multiple time windows based on the time series, thereby obtaining time window series data; Step S32: performing multi-scale decomposition based on wavelet transform on the time window sequence data, decomposing the noise signal into different frequency bands, thereby obtaining wavelet frequency band coefficient data; Step S33: evaluating the signal-to-noise ratio of each frequency band according to the wavelet frequency band coefficient data, and performing adaptive noise threshold determination, thereby obtaining frequency band noise threshold data; Step S34: using the frequency band noise threshold data to perform soft threshold shrinkage processing on the wavelet frequency band coefficient data to suppress the noise component, thereby obtaining the denoised wavelet coefficient data; Step S35: reconstruct the wavelet coefficient data by inverse wavelet transform, and perform parallel calculation in the Spark environment to obtain preliminary denoised sequence data; Step S36: performing physiological noise removal processing including heartbeat and breathing on the preliminary noise reduction sequence data, thereby obtaining noise correction sequence data; Step S37: performing head motion parameter correction and spatial smoothing processing on the noise-corrected sequence data to obtain brain function time series data.

[0013] Preferably, step S4 comprises the following steps: Step S41: partitioning the brain function time series data according to the brain region positioning data in the Spark distributed environment, and standardizing the time series in each level to obtain standardized time series data; Step S42: Calculate the functional connectivity strength between subcortical structures in the local level according to the standardized time series data, and quantify it through Pearson correlation coefficient analysis to obtain a local level functional connectivity matrix; Step S43: Calculate the functional connectivity of the functional partitions at the regional level according to the local level functional connectivity matrix, and perform weighted integration based on the time series correlation of the nodes inside the region, so as to obtain the regional level functional connectivity matrix; Step S44: evaluating node importance and connection significance through network theory methods according to the regional level functional connection matrix, and determining the functional connection pattern of the global level large-scale brain network, thereby obtaining the global level functional connection matrix; Step S45: Dynamically adjust the inter-level weights of the local hierarchical functional connection matrix, the regional hierarchical functional connection matrix, and the global hierarchical functional connection matrix through the principle of synaptic plasticity, establish a mapping relationship of hierarchical connections, and thus obtain brain functional network characteristic data.

[0014] Preferably, step S45 includes the following steps: Step S451: normalizing the local level functional connectivity matrix, the regional level functional connectivity matrix, and the global level functional connectivity matrix, and integrating the data in the Spark distributed environment to construct the level connection mapping matrix data; Step S452: constructing a weight update mathematical model based on the synaptic plasticity principle according to the hierarchical connection mapping matrix data, initializing each cross-level connection weight, thereby obtaining initial cross-level connection weight matrix data; Step S453: performing synchronization analysis on the time series activation of cross-level nodes according to the initial cross-level connection weight matrix data, thereby obtaining preliminary dynamic connection weight data, wherein the synchronization analysis includes calculating the update factors of positive reinforcement and reverse inhibition; Step S454: performing iterative dynamic weight adjustment processing on the preliminary dynamic connection weight data in the Spark distributed environment, and correcting the connection weight according to the preset global convergence condition, so as to obtain cross-level connection weight matrix data; Step S455: Establishing a hierarchical connection mapping relationship according to the cross-hierarchical connection weight matrix data, thereby obtaining brain function network feature data.

[0015] Preferably, step S5 comprises the following steps: Step S51: grouping the brain function network feature data into three levels: local, regional and global, and calculating the Fisher discriminant score for the functional connection features of each level, thereby obtaining hierarchical Fisher score data; Step S52: establishing a knowledge base of brain functional connection patterns related to schizophrenia based on the pre-collected clinical data of schizophrenia patients and healthy controls, thereby obtaining prior knowledge of schizophrenia; Step S53: setting differentiated importance weights for functional connectivity features at different levels based on prior knowledge of schizophrenia, and weighting the hierarchical Fisher score data to obtain weighted feature importance data; Step S54: determining feature selection thresholds at different levels based on the distribution characteristics and quantity of features according to the weighted feature importance data, thereby obtaining hierarchical threshold data; Step S55: using the hierarchical threshold data to screen the weighted feature importance data, thereby obtaining a candidate feature set; Step S56: performing redundancy analysis on the candidate feature set, calculating the correlation between features, and removing highly correlated redundant features according to a preset threshold, thereby obtaining an optimized feature set; Step S57: Matching and testing the optimized feature set with the prior knowledge of schizophrenia to generate the biological significance and clinical interpretability of the optimized features, thereby obtaining the schizophrenia feature set.

[0016] The present invention provides comprehensive and high-quality basic data for subsequent analysis by acquiring brain magnetic resonance image data and resting-state functional magnetic resonance imaging data of the subject. Brain magnetic resonance image data can clearly show the structural characteristics of the brain, while resting-state functional magnetic resonance imaging data reflects the functional activity pattern of the brain in a resting state. This data acquisition method combining structure and function provides multi-dimensional information support for in-depth research on the neural mechanism of schizophrenia, ensuring the comprehensiveness and accuracy of subsequent analysis. Multi-scale hierarchical segmentation of brain images using preset hierarchical brain area templates can accurately locate brain areas at three levels: local, regional and global. This hierarchical method not only covers multiple levels from micro to macro, but also can perform targeted analysis based on the characteristics of different levels. The subcortical structure at the local level, the functional partitioning at the regional level, and the division of large-scale brain networks at the global level enable researchers to comprehensively analyze the structural and functional characteristics of the brain from multiple angles, providing an accurate positioning basis for subsequent functional connection calculations and feature screening. This multi-scale hierarchical segmentation method can effectively avoid the limitations of single-level analysis, making the study of the pathological mechanism of schizophrenia more comprehensive and in-depth. Adaptive hierarchical denoising of resting-state functional magnetic resonance imaging data can effectively remove noise interference and improve data quality. By performing denoising in a Spark distributed environment, not only can the advantages of distributed computing be fully utilized to improve processing efficiency, but also the integrity and accuracy of the data can be ensured. This denoising method can accurately suppress noise components while retaining important physiological signals, thereby providing high-quality data input for subsequent functional connectivity calculations. In addition, steps such as physiological noise removal and head motion parameter correction further improve the reliability and availability of the data, ensuring the accuracy of subsequent analysis. A hierarchical progressive functional connectivity calculation framework is constructed based on brain region localization data and denoised brain function time series data, which can accurately calculate the functional connectivity strength within and between each level. By introducing the principle of synaptic plasticity to dynamically adjust the connection weights between levels, the dynamic changes of the brain functional network can be more realistically reflected. This dynamic adjustment mechanism enables the functional connectivity network to better adapt to the information transmission and interaction between different levels, thereby more accurately revealing the abnormal patterns of the brain functional network in patients with schizophrenia. The resulting brain functional network feature data provides high-quality input for subsequent feature evaluation and screening, ensuring the scientificity and accuracy of feature selection. By using a hierarchical feature importance evaluation method based on Fisher score and setting a hierarchical adaptive threshold based on pre-acquired schizophrenia prior knowledge, we can accurately screen out feature sets related to schizophrenia. This method can not only effectively remove redundant features and reduce feature dimensions, but also improve the biological significance and clinical interpretability of features.Through matching verification with prior knowledge, the selected feature set is further ensured to have high accuracy and reliability, providing strong support for the diagnosis, treatment and research of schizophrenia. This feature screening method based on a combination of prior knowledge and data-driven can effectively improve the pertinence and effectiveness of feature selection, and provide an important reference for subsequent clinical applications and research. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1 A schematic diagram of the steps of the image feature selection method for the Spark model of schizophrenia medical records of the present invention; Figure 2 for Figure 1 Detailed step flow diagram of step S1; Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0018] To achieve this, please refer to Figures 1 to 3 The present invention provides an image feature selection method for a schizophrenia medical record Spark model, the method comprising the following steps: Step S1: acquiring brain magnetic resonance image data and resting-state functional magnetic resonance imaging data of multiple subjects; In this embodiment, 50 subjects were first selected, and their T1-weighted structural image data and resting-state functional magnetic resonance imaging (rs-fMRI) data were obtained using a 3.0T magnetic resonance imaging (MRI) device. During the scan, the subjects were asked to close their eyes but stay awake. The acquisition time was 10 minutes and the time resolution was 2 seconds, that is, each subject obtained data at 300 time points. The spatial resolution of the T1-weighted image was set to 1mm×1mm×1mm, and the spatial resolution of the rs-fMRI data was 3mm×3mm×3mm. All data were stored in NIfTI format and subjected to basic preprocessing, including time slice correction, head motion correction, spatial registration to the MNI standard space, linear drift removal, and bandpass filtering (0.01-0.08Hz) to reduce noise interference and improve the accuracy of subsequent analysis.

[0019] Step S2: using a preset hierarchical brain region template to perform multi-scale hierarchical segmentation on the brain magnetic resonance image data of each subject, and obtain brain region positioning data including three levels: local, regional and global. The multi-scale hierarchical segmentation specifically includes subcortical structure segmentation at the local level, functional partitioning segmentation based on anatomical connection by calculating the structural connection probability and spatial adjacency between brain regions at the regional level, and large-scale brain network segmentation based on hierarchical organization analysis of resting-state networks at the global level; This embodiment uses the Harvard-Oxford brain region template as a hierarchical brain region template, in which the local level includes 68 subcortical structures, the regional level is divided into 116 functional partitions based on the AAL (Automated Anatomical Labeling) template, and the global level is divided into 7 functional networks (such as the default mode network, the central executive network, etc.) based on the large-scale brain network model. In the multi-scale segmentation process, the brain tissue is first segmented based on the T1-weighted structural image to remove non-brain tissues such as the skull. Then, the graph cutting algorithm (Graph Cut) based on the probability graph is used to finely segment the subcortical structure, and the spatial matching of each brain region is performed based on the standard brain map to achieve functional zoning at the regional level. Finally, the functional partitions are further aggregated into a global functional network through a network module analysis method (such as k-means clustering), and the multi-scale hierarchical segmentation is completed to obtain the three-level brain region positioning data for each subject.

[0020] Step S3: performing adaptive hierarchical denoising on the resting-state functional magnetic resonance imaging data in a Spark distributed environment to obtain denoised brain function time series data; In this embodiment, the acquired rs-fMRI data is subjected to adaptive hierarchical noise reduction processing under the Spark distributed computing framework. First, the signal loss data caused by head movement is filled based on the time series interpolation method, and physiological noise (such as breathing and heartbeat signals) is removed by independent component analysis (ICA). Secondly, at the local level, a denoising method based on wavelet transform is used to decompose the time series into multiple frequency components and remove high-frequency noise components; at the regional level, a method based on dynamic low-rank representation (DLR) is used to remove head movement artifacts and maintain the functional relationship between brain regions; at the global level, a sparse Bayesian regression model is used for global noise estimation, and the empirical mode decomposition (EMD) method is used to remove non-physiological artifacts. Finally, all denoised signal data are reintegrated to obtain optimized brain function time series data.

[0021] Step S4: construct a hierarchical functional connection calculation framework based on the brain area positioning data and brain function time series data, calculate the functional connection strength within each level and between levels, and dynamically adjust the connection weights between levels based on the principle of synaptic plasticity to obtain brain function network feature data; This embodiment constructs a hierarchical functional connection calculation framework based on brain region positioning data and denoised brain function time series data. First, at the local level, the dynamic time warping (DTW) method is used to calculate the functional connection strength between subcortical brain regions. Secondly, at the regional level, the sliding window Pearson correlation method is used to calculate the dynamic connection strength between functional partitions, and redundant connections are eliminated through sparse regularization constraints. Then, at the global level, the method based on Granger causal analysis (GCA) is used to calculate the causal influence weights between large-scale brain networks. At the same time, in order to enhance the dynamic adaptability between levels, based on the principle of synaptic plasticity, the cross-level connection weights are adaptively adjusted. Specifically, for the connection from the local to the regional level, the weight update strategy based on STDP (Spike-Timing-Dependent Plasticity) is adopted to strengthen the connection of synchronously activated brain regions; for the connection from the regional to the global level, the nonlinear steady-state distribution optimization method is adopted to keep the overall connectivity of the brain network in dynamic balance. Finally, the characteristic data of the brain functional network is output, including local, regional and global functional connection matrices.

[0022] Step S5: Performing a hierarchical feature importance evaluation based on Fisher score on the brain functional network feature data, and setting a hierarchical adaptive threshold for feature screening according to the pre-acquired schizophrenia prior knowledge to obtain a schizophrenia feature set.

[0023] This embodiment evaluates the hierarchical feature importance of the acquired brain functional network feature data based on Fisher scores, and uses prior knowledge related to schizophrenia for feature screening. Specifically, the Fisher scores of the functional connection features at each level are first calculated to measure the ability of the features to distinguish between the schizophrenia group and the healthy control group. At the local level, the top 10% features with the highest Fisher scores are screened for the connection features of the subcortical structure; at the regional level, the recursive feature elimination (RFE) method is used to continuously iterate and optimize during the feature selection process to ensure that the screened features remain stable in multiple cross-validations; at the global level, an adaptive threshold is set based on the abnormal brain network pattern of schizophrenia (such as overactivation of the default mode network) to remove low-weight connections that are not related to schizophrenia. Finally, the most diagnostically valuable feature set is selected and used for subsequent disease classification model training and prediction.

[0024] Preferably, step S1 comprises the following steps: Step S11: performing T1-weighted magnetic resonance scanning on the brains of multiple subjects using a 3T magnetic resonance imaging device, thereby obtaining brain magnetic resonance image data, wherein the imaging contrast of the brain magnetic resonance image data is specifically determined by the T1 relaxation time; In this embodiment, 40 healthy subjects were selected and T1-weighted magnetic resonance scans were performed on their brains using a 3T magnetic resonance imaging (MRI) device (such as Siemens Prisma 3T). During the imaging process, the subjects were required to remain still and wear noise-reducing earplugs to reduce motion artifacts and external noise interference. T1-weighted imaging uses a three-dimensional MPRAGE (MagnetizationPrepared Rapid Acquisition Gradient Echo) sequence, and the scanning parameters are set to echo time (TE) 2.98ms, repetition time (TR) 2300ms, flip angle 9°, field of view (FOV) 256mm×256mm, matrix resolution 256×256, layer thickness 1mm, and finally obtain T1-weighted brain magnetic resonance image data with an isovoxel resolution of 1mm³. The imaging contrast is determined by the T1 relaxation time, that is, different tissues (such as gray matter, white matter, and cerebrospinal fluid) show different signal intensities under the MPRAGE sequence due to differences in T1 relaxation time to ensure that the structure is clearly distinguishable. After acquisition, the image data were stored in NIfTI format and spatially registered to the MNI standard brain template using SPM software for subsequent analysis.

[0025] Step S12: using a 7T superconducting magnetic resonance imaging device to perform blood oxygen level-dependent contrast acquisition based on a gradient echo sequence on the brains of multiple subjects in a resting state, thereby obtaining blood oxygen level-dependent contrast data; In this embodiment, the same subject as step S11 is selected, and a 7T superconducting magnetic resonance imaging device (such as Siemens Magnetom Terra 7T) is used to perform blood oxygen level dependent (BOLD) contrast acquisition based on a gradient echo (GRE) sequence on the brain in a resting state. The subject needs to close his eyes but stay awake during the scan and avoid any thinking or movement. The BOLD signal acquisition uses a single-shot gradient echo EPI (Echo Planar Imaging) sequence, and the parameters are set to TE25ms, TR1000ms, flip angle 60°, FOV220mm×220mm, matrix resolution 128×128, layer thickness 2mm, no gap, acquisition time 10 minutes, and each subject finally obtains BOLD image data at 600 time points. The BOLD signal reflects the level of neuronal activity based on the local magnetic susceptibility differences caused by changes in the oxygenation state of hemoglobin. After imaging is completed, the data is stored in NIfTI format and corrected for head motion, time slices, and space normalization to improve signal stability and comparability.

[0026] Step S13: Analyze the local blood flow changes caused by neuronal activities on the blood oxygen level dependence contrast data, so as to obtain resting state functional magnetic resonance imaging data.

[0027] This embodiment analyzes the local blood flow changes caused by neuronal activity based on the acquired BOLD signal data to extract resting-state functional magnetic resonance imaging (rs-fMRI) data. First, the BOLD signal is preprocessed based on low-pass filtering (0.01-0.08Hz) to remove high-frequency physiological noise (such as heartbeat and breathing artifacts). Then, a method based on the steady-state oxygenation blood flow coupling model is used to infer the spontaneous neuronal activity of the brain region by calculating the amplitude changes of the local blood oxygen level-dependent signal. Next, the ALFF (Amplitude of Low-Frequency Fluctuations) method is used to calculate the low-frequency amplitude of each brain region to quantify the intensity of neural activity, and the ReHo (Regional Homogeneity) method is used to evaluate the time series synchronization between local brain regions and identify the local characteristics of the resting-state brain functional network. Finally, the functional connection strength between the brain regions of the whole brain is calculated by Pearson correlation analysis, and the functional connection matrix is ​​constructed to finally obtain the resting-state functional magnetic resonance imaging data for subsequent analysis.

[0028] Preferably, step S2 comprises the following steps: Step S21: performing image resolution adaptive normalization processing on the brain magnetic resonance image data, and obtaining normalized image data with a spatial resolution of 1 mm isotropic voxel by resampling the voxel size based on B-spline interpolation; This embodiment uses the B-spline interpolation method to resample the voxel size based on the acquired brain magnetic resonance image data to standardize the spatial resolution. First, read the brain magnetic resonance image in NIfTI format and extract its voxel size information to ensure that the anisotropic resolution of the input image can be uniformly converted. Then, the image is resampled in three dimensions using the third-order B-spline interpolation method to calculate the grayscale value of the new voxel point so that it matches the original image signal as smoothly as possible. B-spline interpolation has better spatial smoothness than bilinear interpolation, while avoiding edge artifacts that may be caused by cubic Lagrange interpolation. During the resampling process, ANTs (Advanced Normalization Tools) are used for spatial transformation to maintain the morphological integrity of the brain structure. Finally, the spatial resolution of all image data is standardized to 1 mm isotropic voxels and re-stored in NIfTI format to ensure the accuracy consistency of subsequent segmentation and analysis.

[0029] Step S22: performing subcortical structure segmentation on the standardized image data according to a preset hierarchical brain region template, and performing brain region positioning processing, thereby obtaining subcortical structure positioning data at a local level; This embodiment uses a method based on a hierarchical brain region template to perform subcortical structure segmentation and brain region localization on the standardized brain magnetic resonance image data. First, the FIRST tool in FSL (FMRIB Software Library) is used to automatically segment the subcortical structure of the T1-weighted image. The target structures include subcortical regions such as the thalamus, caudate nucleus, putamen, amygdala, and hippocampus. During the segmentation process, the brain regions of different subjects are morphologically corrected by the Bayesian probability map matching method, and the segmentation boundaries are optimized based on the maximum a posteriori estimation. Then, the segmentation results are spatially aligned with the preset hierarchical brain region template, and FLIRT (FMRIB's Linear Image Registration Tool) is used for affine transformation to ensure that the individual brain region coordinates of the subjects are consistent with the standard MNI coordinates. Finally, each subcortical structure is labeled and stored to generate subcortical structure localization data at the local level, providing accurate anatomical reference for subsequent functional zoning clustering.

[0030] Step S23: performing functional partitioning clustering based on anatomical connection by calculating the structural connection probability and spatial adjacency relationship between brain regions according to the local-level subcortical structure positioning data, thereby obtaining functional partitioning data at the regional level; This embodiment calculates the structural connection probability and spatial adjacency between brain regions based on the local level subcortical structure positioning data, and performs functional partition clustering based on anatomical connection. First, DTI (Diffusion TensorImaging) data is used for fiber tracking, and the iFOD2 method of MRtrix3 is used to estimate the distribution probability of white matter fiber bundles based on the spherical deconvolution model, and the anatomical connection probability between subcortical structures is calculated. Then, the spatial adjacency matrix of the subcortical brain region is constructed to define the pairs of brain regions that are directly connected anatomically. Next, the method based on fuzzy C-means (FCM) is used to perform functional partition clustering, and the number of partitions is optimized to meet the functional distribution characteristics of brain regions, so that brain regions with higher functional similarity are classified as functional partitions at the same regional level. Finally, the clustering results are mapped to the standard brain space and stored as functional partition data to provide a regional-level brain functional organizational structure.

[0031] Step S24: performing a hierarchical organization analysis of the cerebral cortex based on the resting-state network according to the functional partitioning data and the resting-state functional magnetic resonance imaging data, thereby obtaining large-scale brain network data at the global level, wherein the hierarchical organization analysis specifically calculates the time series correlation and spatial continuity between the functional partitions; This embodiment is based on functional partition data and resting state functional magnetic resonance imaging data, and performs hierarchical organization analysis based on resting state networks to construct a large-scale brain network at the global level. First, the BOLD time series data of each functional partition is extracted, and bandpass filtering (0.01-0.08Hz) is performed to remove high-frequency noise. Then, the Pearson correlation coefficient is used to calculate the time series correlation between functional partitions to obtain the functional connection matrix, and a robust network connection structure is extracted based on the sparse regularization method (L1 regularization). Next, a modularity optimization method (such as the Louvain algorithm) is used to cluster functional partitions to identify brain function modules with strong functional consistency. Furthermore, the betweenness centrality and global efficiency of each functional partition in the network are calculated to measure its role in the large-scale brain network. Finally, the time series correlation between functional partitions is combined with spatial continuity to optimize the hierarchical organizational structure of the network and generate large-scale brain network data at the global level.

[0032] Step S25: hierarchically integrate the subcortical structure positioning data, functional partitioning data, and large-scale brain network data to establish a correspondence matrix between the hierarchies, thereby obtaining brain region positioning data.

[0033] This embodiment is based on subcortical structure positioning data, functional partitioning data and large-scale brain network data, and hierarchical integration is performed to establish a correspondence matrix between the hierarchies, thereby obtaining the final brain region positioning data. First, the hierarchical mapping relationship between the subcortical structure, functional partitioning and large-scale network is defined, the contribution ratio of each subcortical brain region to the regional hierarchical functional partitioning is calculated, and the distribution similarity is measured by Kullback-Leibler divergence to optimize the corresponding relationship. Then, a multi-level brain network is constructed based on the graph theory method, and a weighted adjacency matrix is ​​used to represent the connection strength between local, regional and global levels. Next, a dynamic adjustment mechanism of the connection weight based on the principle of synaptic plasticity is adopted to optimize the information transmission method between the levels, so that the brain region positioning data at different levels maintain consistency in anatomical structure and functional characteristics. Finally, a complete brain region positioning data is generated to provide standardized input for subsequent brain function network analysis.

[0034] Preferably, step S22 includes the following steps: Step S221: performing edge enhancement processing by calculating the spatial gradient amplitude and direction of the three-dimensional image according to the standardized image data, thereby obtaining boundary enhancement data of the subcortical structure; In this embodiment, standardized image data is first used to calculate the grayscale changes in the neighborhood of each voxel in the three-dimensional image, and the spatial gradient amplitude and direction of the image are calculated using the gradient operation method, wherein the gradient amplitude reflects the speed of grayscale changes, and the gradient direction indicates the main direction of grayscale changes. In specific operations, the central difference method is used to obtain the first-order derivatives of the image in the X, Y, and Z directions respectively, and the gradient amplitude is obtained by square root and square root, and then the inverse tangent function is used to calculate the angle distribution of the gradient direction, and finally an edge-enhanced image is obtained by combining, which has a high response value at the boundary of the subcortical structure, thereby effectively distinguishing the boundaries between tissues. In terms of operating parameters, the window size is set to 3×3×3 voxels to balance noise suppression and edge detection accuracy, and the calculation results are normalized to ensure that the dynamic range of the enhanced data is suitable for subsequent processing. The subcortical structure boundary enhancement data finally generated can significantly highlight the boundary information between various brain structures.

[0035] Step S222: Analyze the grayscale distribution characteristics of the local area according to the boundary enhancement data, set a dynamic threshold, and perform segmentation, thereby obtaining preliminary segmentation mask data of the subcortical structure; Based on the obtained boundary enhancement data, this embodiment performs a detailed analysis of the grayscale distribution characteristics in the local area, and uses a dynamic threshold segmentation method to achieve preliminary segmentation by counting the mean, variance and histogram distribution of the pixel grayscale values ​​in each area. Specifically, first, in the predefined region of interest, the average grayscale level in the region is determined by the local mean filtering method, and then the dynamic threshold is determined based on the local standard deviation, which means that pixels with grayscale values ​​greater than the dynamic threshold are considered to belong to the edge of the subcortical structure, and the threshold is automatically adjusted according to the peak and valley relationship of the local grayscale histogram to ensure that it adapts to the differences in grayscale distribution of different brain regions, and then the boundary enhancement data is binarized, and finally preliminary subcortical structure segmentation mask data is generated, in which high grayscale areas are marked as candidate areas for subcortical structures, while low grayscale areas are regarded as background or noise.

[0036] Step S223: morphologically optimizing the preliminary segmentation mask data by using morphological operations, so as to obtain continuous and complete subcortical structure mask data, wherein the morphological optimization specifically processes holes and boundary discontinuities by designing a three-dimensional morphological operator sequence; This embodiment uses three-dimensional morphological operations to optimize the morphology of the preliminary segmentation mask data, with the aim of eliminating holes and boundary discontinuities caused by noise or segmentation errors. In terms of specific operations, the preliminary mask is first processed using a three-dimensional dilation operation, using a structural element designed as a 3×3×3 cube to fill small holes; then a three-dimensional erosion operation is performed to eliminate redundant noise introduced by dilation while maintaining the continuity of the original structure; then a combination of an open operation (erosion followed by dilation) and a closed operation (dilation followed by erosion) is used to further correct areas with uneven or broken edges to ensure the continuity and smoothness of the mask edge; the entire morphological operator sequence is iteratively optimized for multiple times, and the number of iterations of each operation in the parameter setting is usually between 1 and 3, which is adjusted according to the actual image quality, and finally a continuous and complete subcortical structure mask data is obtained.

[0037] Step S224: Perform connected domain analysis based on the subcortical structure mask data and mark different subcortical structures, so as to obtain fine subcortical structure positioning data at the local level.

[0038] In this embodiment, based on the morphologically optimized subcortical structure mask data, the connected domain analysis method is used to mark the independent connected regions in the image to achieve fine segmentation of different subcortical structures. In the specific operation, the entire mask data is scanned using a three-dimensional connected domain labeling algorithm (for example, the 26-neighborhood method), and all connected pixel blocks are grouped. Each connected domain is regarded as an independent structural candidate region; then, by comparing the volume, shape and position characteristics of each connected domain, it is matched with the preset subcortical structure database, so as to accurately assign a label to each connected domain; in addition, for areas with weak edges or segmentation errors, the regional growing algorithm is combined for fine-tuning to ensure that each marked subcortical structure is a continuous, complete and anatomically consistent area, and finally outputs the fine subcortical structure positioning data at the local level, which contains the spatial position, boundary shape and label information of each subcortical structure, providing a high-precision anatomical basis for subsequent multi-level brain area analysis.

[0039] Preferably, step S23 includes the following steps: Step S231: performing connection fiber bundle tracing on the local level subcortical structure positioning data, and performing white matter fiber connection analysis between subcortical structures, thereby obtaining a structural connection probability map; This embodiment uses an extended fiber tracking algorithm to track white matter fibers between subcortical structures based on local hierarchical subcortical structure positioning data. First, the preprocessed diffusion-weighted imaging data is used to reconstruct the local fiber direction distribution through a spherical model. Then, a probabilistic fiber tracking algorithm (for example, a random walk model based on the Monte Carlo method) is used to generate a large number of fiber bundles for the seed points of each subcortical structure. Each seed point is set to generate 5,000 fibers, and the maximum step length is set to 0.5 mm and the curvature threshold is set to no more than 45 degrees during the fiber tracking process to ensure that the tracked fibers meet anatomical constraints. Subsequently, the starting point and end point of each fiber are matched in a predefined subcortical structure segmentation map, and the structural connection probability between the two regions is calculated by counting the number of all fibers between the two regions, and the connection strength is represented by a probability value. Finally, a structural connection probability map is generated, which expresses the relative possibility and strength of the connection between the subcortical structures in a matrix form, providing a quantitative basis for subsequent analysis.

[0040] Step S232: Calculating the spatial adjacency relationship between brain regions based on the spatial distance metric according to the structural connection probability map, thereby obtaining brain region spatial relationship data; This embodiment is based on the structural connection probability map generated above, and constructs spatial adjacency relationship data by calculating the spatial distance between brain regions. First, the three-dimensional coordinates of each subcortical structure in the standardized image are extracted, and the centroid coordinates of each structure are used as representatives; then, a distance matrix is ​​obtained by calculating the Euclidean distance between any two centroids, and a preset spatial distance threshold (for example, set to 20 mm) is used to determine whether there is a direct spatial adjacency relationship between the two regions. If the distance is less than the threshold, it is considered that the two brain regions are spatially adjacent. The closer the distance, the higher the weight. By normalizing the distance by the inverse, the closer brain regions correspond to higher spatial correlation weights, and finally the spatial relationship data of the brain regions are constructed. The data presents the spatial connection strength between the regions in the form of a weighted adjacency matrix.

[0041] Step S233: performing weighted fusion on the structural connection probability map and the brain region spatial relationship data based on the structural connection relationship and the spatial adjacency relationship, thereby obtaining a comprehensive relationship matrix; In this embodiment, the structural connection probability map and the spatial relationship data of brain regions are weightedly fused to obtain a comprehensive relationship matrix that comprehensively reflects the anatomical connection and spatial adjacency between brain regions. First, the two types of data are normalized separately, and the structural connection probability value and the spatial adjacency weight are mapped to the range of 0 to 1. Then, a linear weighted fusion strategy is adopted to weighted average the normalized structural connection probability and the normalized spatial relationship data according to a preset weight ratio (for example, the structural connection weight is set to 0.7 and the spatial adjacency weight is set to 0.3). The weighted result of the corresponding element of each matrix during the fusion process is the comprehensive connection value of the pair of brain regions. This value reflects the possibility of fiber connection between the two regions and also takes into account their proximity in physical space. The final generated comprehensive relationship matrix is ​​a symmetric matrix that can fully display the anatomical and spatial connection information between brain regions, providing sufficient basis for spectral clustering analysis.

[0042] Step S234: partition the comprehensive relationship matrix using a spectral clustering algorithm, thereby obtaining functional partition data at the regional level.

[0043] In this embodiment, a spectral clustering algorithm is used to partition the aforementioned comprehensive relationship matrix to obtain functional partitioning data at the regional level. First, a normalized Laplace matrix is ​​constructed for the comprehensive relationship matrix, that is, a normalized graph Laplace operator is calculated after the degree matrix is ​​constructed using the comprehensive relationship matrix, and the eigenvectors corresponding to the first k smallest non-zero eigenvalues ​​are extracted by eigenvalue decomposition, wherein the k value is determined automatically according to a predetermined number of partitions (for example, set to 10 regions) or by the Eigengap method; then, the representation of each brain region in these eigenvector spaces is used as a clustering feature, and all brain regions are divided into several clusters using the K-means clustering algorithm, each cluster representing a functional partition; in the clustering process, the number of iterations is set to 100, and the Euclidean distance is used as the distance metric to ensure the stability and accuracy of the clustering results; finally, each cluster is mapped back to the original brain region coordinate space to generate functional partitioning data at the regional level, which not only reflects the anatomical connection advantages between brain regions, but also integrates the spatial adjacency information, providing a more refined and physiologically relevant partitioning basis for subsequent brain functional network analysis.

[0044] Preferably, step S25 comprises the following steps: Step S251: extracting the spatial position, volume and shape features of each structure from the subcortical structure positioning data, and constructing a feature vector to obtain a local hierarchical feature matrix; Embodiments of the present invention This embodiment aims at the subcortical structure positioning data, and constructs a local hierarchical feature matrix by extracting the three-dimensional spatial position, volume and shape features of each subcortical structure. The specific operation first uses the preprocessed standardized T1-weighted MRI image and the corresponding subcortical structure mask data, and uses the three-dimensional region growing algorithm to calculate the center coordinates of each structure as its spatial position, and then estimates the volume by counting the number of voxels of each structure in the image, and extracts shape features including the compactness of the structure boundary, the surface area to volume ratio and the irregularity parameter. These shape indicators are obtained by boundary extraction and contour analysis. On this basis, after each feature is standardized, the principal component analysis (PCA) method is used to reduce the feature dimension, and a feature vector composed of each eigenvalue of each structure is constructed. Finally, the feature vectors of all subcortical structures are arranged into a matrix to form a local hierarchical feature matrix. The noise in the region is filtered throughout the process, and the minimum threshold of the region size is set to 50 voxels to ensure the reliability and consistency of the extracted data.

[0045] Step S252: Analyze the functional connection pattern and spatial distribution characteristics of each partition according to the functional partition data, so as to obtain a regional level feature matrix; Embodiments of the present invention This embodiment uses functional partition data to analyze the functional connection pattern and spatial distribution characteristics of each region, thereby constructing a regional hierarchical feature matrix. The specific operation is to first extract the preprocessed resting fMRI signal from each functional partition, and use the time series correlation analysis method to calculate the functional connection strength between the brain regions in the partition to obtain the connection pattern within the partition; at the same time, based on the spatial coordinates of the partition in the standardized brain map, calculate the central position, spatial range and shape distribution characteristics of the partition, such as the eccentricity and spatial distribution discreteness of the partition, which are obtained by counting the position coordinates of each pixel or voxel in the region. Then, the functional connection pattern characteristics and spatial distribution characteristics of each functional partition are normalized, and the functional consistency of the partition is verified by cluster analysis. Finally, the comprehensive characteristics of each partition (including functional connection strength index and spatial morphology index) are combined into a column of feature vectors, and arranged in a matrix form to form a regional hierarchical feature matrix. In actual operation, the time window is set to 60 seconds and the step length is set to 30 seconds to ensure the stability and representativeness of the time series data.

[0046] Step S253: extracting network features including topological attributes and dynamic features from the large-scale brain network data, thereby obtaining a global hierarchical feature matrix; Embodiments of the present invention This embodiment is based on large-scale brain network data, and uses network analysis methods to extract its topological properties and dynamic features, and construct a global hierarchical feature matrix. Specific steps First, the pre-processed resting state fMRI data and large-scale network partitioning results are used to calculate the functional connection matrix between each network node, and then the graph theory indicators such as node degree, clustering coefficient, betweenness centrality, global efficiency, etc. are used to measure the network topology structure. Each indicator is obtained by statistically analyzing the connection of nodes in the network; at the same time, the sliding window method is applied to the time series data for dynamic network analysis to extract the time-varying characteristics of network connections, such as the dynamic connectivity change amplitude and stability coefficient, and the window length is specifically set to 30 seconds and the overlap rate is 50%. After obtaining the topological indicators and dynamic features, the indicator data of all nodes are standardized, and the multivariate statistical method is used for feature fusion to construct the feature vector of each large-scale network. Finally, the feature vectors of each large-scale network are integrated and arranged into a matrix to form a global hierarchical feature matrix. The matrix not only reflects the static network structure, but also captures the dynamic change characteristics, providing a quantitative basis for the overall brain network function evaluation.

[0047] Step S254: local level feature matrix, regional level feature matrix and global level feature matrix are processed to obtain brain region localization data.

[0048] Embodiments of the present invention This embodiment integrates the local level, regional level and global level feature matrices to construct the final brain region positioning data. The specific operation is to first align the feature matrices of the three levels, determine the correspondence between the matrices by matching the shared samples, and use multi-level feature fusion methods (such as feature cascade and weighted average) to achieve feature fusion. When fusion, different weights are set according to the signal-to-noise ratio and statistical stability of the data at each level (for example, local level weight 0.4, regional level weight 0.35, global level weight 0.25) to ensure that the comprehensive data reflects the advantages of the features at each level. Then, the fused feature data is subjected to dimensionality reduction processing, and the most discriminative feature vectors are extracted using the linear discriminant analysis (LDA) method, and the dimensionality-reduced data is mapped back to the original brain region space, and a brain region positioning map with anatomical and functional significance is reconstructed through spatial interpolation and registration technology, in which the position, volume and morphological information of each brain region correspond to the multi-level features. The final brain region localization data not only contains the fine positioning information of the anatomical structure, but also reflects the functional connectivity and network dynamic characteristics, providing reliable multi-dimensional information support for subsequent disease diagnosis and analysis.

[0049] Preferably, step S3 comprises the following steps: Step S31: partitioning the resting state functional magnetic resonance imaging data in the Spark distributed environment, dividing the data into multiple time windows based on the time series, thereby obtaining time window series data; In this embodiment, the resting-state functional magnetic resonance imaging data is first loaded into the Spark distributed computing platform in NIfTI format, and the data is segmented using Spark's built-in data partitioning function. The entire data set is divided into multiple continuous time windows in the order of the time series, and the length of each window is set to 30 seconds according to the actual data acquisition parameters (for example, if TR is 2 seconds, each window contains 15 time points). During the partitioning process, a custom function is used to ensure that the time sequence of the data in each partition is complete and correct. The final time window series data is stored in the form of RDD or DataFrame, providing efficient data access and computing basis for subsequent parallel processing.

[0050] Step S32: performing multi-scale decomposition based on wavelet transform on the time window sequence data, decomposing the noise signal into different frequency bands, thereby obtaining wavelet frequency band coefficient data; This embodiment adopts a multi-scale decomposition method based on wavelet transform for each time window series data, decomposes the signal at different scales to distinguish between noise and real signals. In specific operations, the Daubechies wavelet basis (such as db4) is first selected to perform discrete wavelet transform on the time series data in each window. The number of decomposition layers is set to 5, and the original signal is decomposed into low-frequency approximation coefficients and high-frequency detail coefficients through a recursive filter group, wherein the high-frequency part mainly contains noise components and the low-frequency part contains useful neural activity signals. The decomposition results are stored in the form of wavelet frequency band coefficient data matrix, and the coefficient data of each layer corresponds to information of different frequency bands, providing a basis for subsequent noise processing.

[0051] Step S33: evaluating the signal-to-noise ratio of each frequency band according to the wavelet frequency band coefficient data, and performing adaptive noise threshold determination, thereby obtaining frequency band noise threshold data; This embodiment evaluates the signal-to-noise ratio (SNR) of each frequency band based on the obtained wavelet frequency band coefficient data. The specific method is to calculate the mean and standard deviation of the statistical coefficients in each decomposition layer, determine the SNR of each frequency band by comparing the amplitude difference between the signal and the noise in different frequency bands, and then use an adaptive threshold algorithm (for example, a method based on local maximum and median absolute deviation) to determine the noise threshold. The process of determining the noise threshold data of each frequency band takes into account the discrete degree of the coefficient distribution in the frequency band. The threshold is set to be the statistic multiplied by a preset factor (for example, 1.5 times). The final generated frequency band noise threshold data is recorded in the form of an array, and each element corresponds to the threshold of a decomposition layer, which is used to guide the subsequent soft threshold processing.

[0052] Step S34: using the frequency band noise threshold data to perform soft threshold shrinkage processing on the wavelet frequency band coefficient data to suppress the noise component, thereby obtaining the denoised wavelet coefficient data; This embodiment uses the frequency band noise threshold data determined in step S33 to perform soft threshold shrinkage processing on the wavelet frequency band coefficient data. In specific operations, for each wavelet coefficient, if its absolute value is less than the noise threshold of the corresponding frequency band, the coefficient is set to zero; if it is greater than the threshold, the coefficient is shrunk according to the threshold, that is, the original sign is retained after subtracting the threshold from its amplitude, thereby achieving a smooth transition and reducing artifacts and discontinuities that may be introduced by hard threshold processing. The coefficient data after soft threshold processing retains the main features of the signal while effectively suppressing the noise component. The denoised wavelet coefficient data finally generated provides a clean basis for the reconstructed signal of each time window after decomposition at each layer.

[0053] Step S35: reconstruct the wavelet coefficient data by inverse wavelet transform, and perform parallel calculation in the Spark environment to obtain preliminary denoised sequence data; This embodiment performs inverse wavelet transform reconstruction on the wavelet coefficient data that has been processed by soft thresholding to restore the denoised time series data. In the specific operation, the same wavelet basis (such as db4) and decomposition level as the forward wavelet transform are selected, and the coefficients of each frequency band in each time window are merged and reconstructed into a complete time series signal through inverse transform. To improve computing efficiency, this inverse transform process is parallelized in the Spark distributed environment through the MapReduce paradigm to ensure that the data in each partition is processed independently and the results are seamlessly integrated. The preliminary denoised sequence data finally output has a higher signal-to-noise ratio and less high-frequency noise interference in the time domain.

[0054] Step S36: performing physiological noise removal processing including heartbeat and breathing on the preliminary noise reduction sequence data, thereby obtaining noise correction sequence data; Based on the preliminary denoised sequence data, this embodiment further removes the influence of physiological noise such as heartbeat and breathing on the signal. First, a spectrum analysis-based method is used to detect typical heart rate (usually in the range of 0.8 to 1.2 Hz) and breathing rate (usually in the range of 0.2 to 0.4 Hz) in the signal. Then, a band-stop filter is used to design a corresponding filter to suppress the noise in these frequency bands. The bandwidth and attenuation parameters of the filter are adaptively adjusted according to the specific physiological parameters of the subject (such as pre-measured heart rate and breathing rate). In addition, independent component analysis (ICA) can be combined to further separate and remove physiological noise components. The noise-corrected sequence data obtained after these processes significantly reduces the energy of heartbeat and breathing noise in the frequency domain, making the neural signal component more prominent.

[0055] Step S37: performing head motion parameter correction and spatial smoothing processing on the noise-corrected sequence data to obtain brain function time series data.

[0056] In this embodiment, the noise-corrected sequence data is further subjected to head motion parameter correction and spatial smoothing processing to obtain the final brain function time series data. Specifically, a head motion correction algorithm (such as an algorithm based on six-parameter rigid registration) is first used to correct the tiny movements of the subject during the scanning process, and the displacement and rotation errors are estimated and corrected by matching the image at each time point with the reference image. The correction process is accelerated by parallel computing in the Spark environment. Next, a Gaussian smoothing kernel is used for spatial smoothing processing, and the standard deviation of the smoothing kernel is set to 6 mm to adapt to the spatial resolution of the fMRI data. The smoothing processing helps to improve the signal-to-noise ratio and the consistency of the signal within the region. After these preprocessing steps, the brain function time series data finally output has high spatiotemporal consistency and low noise interference, which is suitable for subsequent functional connection and network analysis research.

[0057] Preferably, step S4 comprises the following steps: Step S41: partitioning the brain function time series data according to the brain region positioning data in the Spark distributed environment, and standardizing the time series in each level to obtain standardized time series data; In this embodiment, the preprocessed brain function time series data is partitioned in the Spark distributed environment. First, the data is labeled according to the previously generated brain region positioning data, and a mapping relationship is established between each time series and its corresponding brain region or layer (such as subcortical structure, functional partition or large-scale network). Then, the entire data set is divided according to these mapping labels using Spark's partitioning function to ensure that each partition only contains time series data in the same brain region or the same layer. Next, a standardized processing method is used for the time series data in each partition, that is, the mean and standard deviation of each time series are first calculated, and then the value at each time point is subtracted from the mean and divided by the standard deviation. The difference transformation (often called Z-score normalization) is performed to eliminate the influence of signal amplitude differences in different brain regions. The standardized data is saved in Spark's DataFrame. Each record contains the standardized time series, the corresponding brain region label and the time index. The partitioning process uses a custom partitioner to ensure balanced data distribution. The parameters usually set the length of each time window to 30 seconds (for example, when TR is 2 seconds, it contains 15 data points) to obtain sufficient statistical stability. At the same time, linear interpolation is used to compensate for missing data. The final standardized time series data provides a unified, smooth and parallel-processable basic data for subsequent functional connection calculations.

[0058] Step S42: Calculate the functional connectivity strength between subcortical structures in the local level according to the standardized time series data, and quantify it through Pearson correlation coefficient analysis to obtain a local level functional connectivity matrix; This embodiment uses the standardized time series data obtained in step S41 to quantify the functional connection strength between local hierarchical subcortical structures. First, for each pair of subcortical structures, a signal sequence is extracted from its corresponding standardized time series, and then the Pearson correlation coefficient analysis method is used for calculation, that is, the covariance between each pair of sequences is calculated and divided by the product of their respective standard deviations, so as to obtain a correlation coefficient between -1 and 1. The closer the value is to 1, the higher the synchronization of the signals in the two regions is, and the closer it is to -1, the reverse synchronization is indicated. During the calculation process, the data of each subject are analyzed independently, and the window length is usually consistent with the time window used in S41 to ensure timing matching; all calculated correlation coefficients are arranged into a matrix according to the labels of the subcortical structures, and each row and each column represents a subcortical structure. The matrix is ​​the local hierarchical functional connection matrix, and the matrix is ​​stored in a distributed file system for further processing. At the same time, in order to improve the calculation accuracy, the extreme values ​​are robustly corrected to ensure that the functional connection strength reflected in the matrix has high biological significance and statistical reliability.

[0059] Step S43: Calculate the functional connectivity of the functional partitions at the regional level according to the local level functional connectivity matrix, and perform weighted integration based on the time series correlation of the nodes inside the region, so as to obtain the regional level functional connectivity matrix; This embodiment integrates and weights the functional partitions at the regional level based on the local-level functional connectivity matrix. First, the time series data of the subcortical structures belonging to the same functional partition in the local level are aggregated according to the preset functional partition data, and the inter-regional connectivity index is constructed by calculating the weighted average correlation coefficient between all nodes within the partition. The weight coefficient can be allocated according to the signal-to-noise ratio or signal stability of each node to ensure that the nodes with stronger or more stable signals in the functional partition play a greater role in the overall weighting; then, the inter-regional correlation calculation is performed on the aggregated data, similar to the Pearson correlation coefficient method used in S42, so as to construct a new matrix, in which each element represents the comprehensive functional connection strength between two functional partitions. The sliding window analysis method is used throughout the process to capture time-varying features. The window length is usually set to 60 seconds and overlapped by 50% to balance the time domain resolution and statistical stability. The regional-level functional connectivity matrix reflects both the functional integration within the region and the interaction between different regions. The matrix data is normalized and stored in a distributed environment for subsequent network analysis.

[0060] Step S44: evaluating node importance and connection significance through network theory methods according to the regional level functional connection matrix, and determining the functional connection pattern of the global level large-scale brain network, thereby obtaining the global level functional connection matrix; After obtaining the regional-level functional connectivity matrix, this embodiment uses network theory methods to evaluate the importance of each functional partition (node) and detect the connection significance, so as to determine the large-scale brain network functional connectivity pattern at the global level. First, graph theory indicators such as node degree, betweenness centrality and clustering coefficient are calculated for each node in the matrix, where the node degree reflects the number of connections between the region and other regions, betweenness centrality quantifies the bridge role of the node in the entire network, and the clustering coefficient is used to evaluate the tightness of the connection within the neighborhood of the node; these indicators are obtained by statistical analysis methods, and the Z-score standardization is used to unify the scales of different indicators, and then non-parametric statistical tests are used. Methods (such as permutation tests) are used to determine the significance level of each connection, and connections with significance below a preset p-value threshold (such as 0.05) are removed. In addition, the topological structure of the remaining connections is optimized, and modular algorithms (such as the Louvain algorithm) are used to identify functional modules in large-scale brain networks. The final global hierarchical functional connectivity matrix not only reflects the connection strength between regions, but also includes the relative importance of each node in the network and module affiliation information, providing a detailed structural basis for the analysis of the overall network functional pattern. All processing parameters can be adjusted according to experimental data and prior knowledge to ensure the physiological relevance and statistical robustness of the network model.

[0061] Step S45: Dynamically adjust the inter-level weights of the local hierarchical functional connection matrix, the regional hierarchical functional connection matrix, and the global hierarchical functional connection matrix through the principle of synaptic plasticity, establish a mapping relationship of hierarchical connections, and thus obtain brain functional network characteristic data.

[0062] In this embodiment, the weights of the local, regional and global functional connection matrices are adjusted dynamically based on the principle of synaptic plasticity to establish a connection mapping relationship between the levels. First, the three-level connection matrices are normalized and then initially fused. Then, based on the synaptic plasticity model (such as Spike-Timing-Dependent The principle of STDP (Stimulus Plasticity) is used to assign dynamic update rules to each cross-level connection. The specific operation is to detect the time series synchronization between cross-level nodes. If synchronous activation is detected, the connection weight is enhanced accordingly, otherwise it is suppressed. The process is carried out in parallel in the Spark environment through an iterative algorithm. Each update in the iterative process considers the mutual influence between the local connection matrix, the regional connection matrix and the global connection matrix. The text description of the weight update formula is: the current connection weight is equal to the weight at the previous moment plus an update factor proportional to the degree of synchronization, and the penalty term related to the asynchronous component is subtracted. Iterate until the global convergence condition is met (such as the update factor is less than the preset threshold of 0.001). The final generated brain functional network feature data is stored in the form of a comprehensive matrix. The matrix not only reflects the original strength of the functional connections at each level, but also reflects the hierarchical mapping relationship after dynamic adjustment, providing high dynamic range and biologically meaningful multi-level functional connection characteristics for subsequent pathological feature analysis and clinical decision-making.

[0063] Preferably, step S45 includes the following steps: Step S451: normalizing the local level functional connectivity matrix, the regional level functional connectivity matrix, and the global level functional connectivity matrix, and integrating the data in the Spark distributed environment to construct the level connection mapping matrix data; In this embodiment, the local, regional and global functional connection matrices are first normalized respectively, using a minimum-maximum normalization method, that is, the value of each element in each matrix is ​​subtracted from the minimum value in the matrix and then divided by the difference between the maximum value and the minimum value, so that the value range of all matrices is standardized to between 0 and 1, thereby eliminating the deviation caused by the different scales of data at each level; the normalized matrices are then integrated in the Spark distributed environment by joining data sets (Join), and the normalized local, regional and global matrices are aligned and merged according to the corresponding relationship between brain regions using Spark's DataFrame API, and finally a hierarchical connection mapping matrix data is constructed, which stores the comprehensive value of cross-level connections of each brain region in the form of a multidimensional array, providing a unified data basis for subsequent dynamic weight updates, and at the same time, during the data integration process, the data partition size is set so that each partition contains no more than 5000 data units, so as to ensure the efficiency and stability of distributed computing.

[0064] Step S452: constructing a weight update mathematical model based on the synaptic plasticity principle according to the hierarchical connection mapping matrix data, initializing each cross-level connection weight, thereby obtaining initial cross-level connection weight matrix data; This embodiment is based on the hierarchical connection mapping matrix data constructed in step S451, constructs a weight update mathematical model according to the principle of synaptic plasticity, and initializes the cross-level connection weights. In the specific operation, a dynamic weight update framework is first designed. The core idea is to regard each cross-level connection weight as the product of the initial connection strength and the subsequent activity regulation factor, and then apply a preset initialization function to the elements in each mapping matrix. The function is calculated based on the normalized value in the mapping matrix and a constant initialization factor (for example, the initial factor is set to a random value or a fixed value between 0.5 and 1.0), so as to set the initial value of the weight of each connection within a reasonable range; the initialization process is implemented in the Spark environment using the Map operation, and the weights in each data partition are independently initialized and then globally merged. The final initial cross-level connection weight matrix data provides the starting conditions for subsequent iterative updates, and the initialization parameters can be fine-tuned according to the actual data distribution to ensure that the model has good convergence and biological rationality during subsequent weight adjustments.

[0065] Step S453: performing synchronization analysis on the time series activation of cross-level nodes according to the initial cross-level connection weight matrix data, thereby obtaining preliminary dynamic connection weight data, wherein the synchronization analysis includes calculating the update factors of positive reinforcement and reverse inhibition; This embodiment performs synchronization analysis on the time series activation of cross-level nodes based on the initial cross-level connection weight matrix data to obtain preliminary dynamic connection weight data. In the specific operation, the standardized time series data of the relevant brain areas of each cross-level connection are first extracted, and then the synchronization of activation between different brain areas is evaluated by using the cross-correlation analysis method of the time series. By comparing the phase and amplitude of the activation signal, an update factor reflecting the positive reinforcement and reverse inhibition effects is calculated. The determination process of the update factor includes the count of peak synchronization events in the signal window and the inhibition evaluation of the low-correlation period. The positive update factor reflects that the connection weight should be appropriately increased under high synchronization, while the reverse update factor promotes the connection weight to be reduced under low synchronization or anti-phase activity. The synchronization analysis uses a rolling window technology (for example, the window length is set to 30 seconds and the step length is 15 seconds) for local statistics, and the update factor of each connection is calculated in parallel on the Spark distributed platform. Finally, these factors are combined with the initial weight to form preliminary dynamic connection weight data, which provides feedback basis and preliminary dynamic response indicators for subsequent iterative adjustments.

[0066] Step S454: performing iterative dynamic weight adjustment processing on the preliminary dynamic connection weight data in the Spark distributed environment, and correcting the connection weight according to the preset global convergence condition, so as to obtain cross-level connection weight matrix data; In this embodiment, the preliminary dynamic connection weight data is iteratively adjusted in a Spark distributed environment to achieve global optimization of the connection weights. An iterative algorithm is used in the specific operation. In each iteration, the weights are updated according to the difference between the current cross-level connection weight data and the result of the previous iteration. The update process follows the preset global convergence condition (for example, when the change amplitude of all connection weights is less than 0.001, it is considered to have converged). In each iteration, each cross-level connection is weighted and adjusted using the aforementioned synchronization update factor. At the same time, an adaptive learning rate parameter is introduced (for example, the initial learning rate is set to 0.05 and gradually decays according to the number of iterations) to ensure smooth convergence of the update process. On the Spark platform, the iterative process triggers Map and Reduce operations through a loop to achieve local updates of the weights in each data partition and then perform global integration until the convergence condition is met. The cross-level connection weight matrix data finally output not only has higher stability, but also can more truly reflect the biological characteristics of dynamic interactions between brain regions.

[0067] Step S455: Establishing a hierarchical connection mapping relationship according to the cross-hierarchical connection weight matrix data, thereby obtaining brain function network feature data.

[0068] This embodiment is based on the cross-level connection weight matrix data finally obtained, and uses these data to establish a mapping relationship of hierarchical connections. In the specific operation, the updated weight matrix is ​​first rearranged according to brain regions and layers, and the connection weights between each brain region are statistically summarized to form a mapping table that comprehensively describes the connection status between each layer. This mapping table is represented in the form of a multidimensional matrix or a graph structure, in which each node represents a brain region, and the edge weight between nodes is the final weight value of the cross-level connection; then, the community detection algorithm (such as Louvain algorithm) and visualization tools in graph theory are used to analyze the mapping relationship to verify whether the connection pattern between the layers conforms to the expected synaptic plasticity principle and physiological law; finally, the mapping relationship data is saved in a standardized format (such as CSV or Parquet format) and stored in a distributed file system, and used as brain functional network feature data for subsequent clinical decision-making and machine learning model training. In specific application scenarios, this data can help researchers discover abnormal connection patterns in brain networks and provide quantitative support for the diagnosis of diseases such as schizophrenia.

[0069] Preferably, step S5 comprises the following steps: Step S51: grouping the brain function network feature data into three levels: local, regional and global, and calculating the Fisher discriminant score for the functional connection features of each level, thereby obtaining hierarchical Fisher score data; In this embodiment, after the obtained brain function network feature data are grouped according to three levels: local, regional and global, each functional connection feature in each level is evaluated using the Fisher discriminant analysis method, that is, the mean difference and the intra-group variance ratio of the schizophrenia patient group and the healthy control group on the feature are calculated respectively. In the specific operation, the mean and standard deviation of each feature in the two groups of data are first calculated respectively, and then the discriminant score is calculated. The textual expression of the score is the mean difference between the groups divided by the sum of the intra-group standard deviations. The higher the value, the stronger the ability of the feature to distinguish the two groups. The Fisher score data of all features are stored in a distributed database according to the level. The parameters generally adopt the unbiased estimation method when calculating the standard deviation, and are calculated in parallel on the Spark platform to improve efficiency, thereby obtaining the Fisher discriminant score data of the local, regional and global levels.

[0070] Step S52: establishing a knowledge base of brain functional connection patterns related to schizophrenia based on the pre-collected clinical data of schizophrenia patients and healthy controls, thereby obtaining prior knowledge of schizophrenia; This embodiment is based on the detailed clinical data of schizophrenia patients and healthy controls collected in advance, including symptom assessment, cognitive test and neuroimaging data, to build a knowledge base of brain functional connection patterns related to schizophrenia. Specifically, by integrating multimodal data and literature reports from multiple centers, data mining and machine learning methods (such as clustering and pattern recognition algorithms) are used to identify abnormal functional connection patterns that recur in schizophrenia, and combined with expert annotations to form a priori knowledge base. The knowledge base is stored in the form of a relational database and is accompanied by descriptive statistics and clinical relevance indicators. In the application scenario, a data set with a sample size of not less than 100 cases is often used to ensure statistical robustness, and finally a set of schizophrenia prior knowledge for guiding feature screening is obtained.

[0071] Step S53: setting differentiated importance weights for functional connectivity features at different levels based on prior knowledge of schizophrenia, and weighting the hierarchical Fisher score data to obtain weighted feature importance data; This embodiment assigns differentiated importance weights to functional connectivity features at different levels based on the prior knowledge of schizophrenia obtained in step S52. Specifically, the initial weights are first set for features at each level based on the key brain networks and connection patterns recorded in the prior knowledge. For example, higher weights are assigned to connections in the default mode network and the central executive network, while lower weights are assigned to weaker related networks. Then, the Fisher score data at each level is multiplied by the corresponding weights and weighted averaged, so that features with clinical indicative significance are more prominent in the weighted feature importance data. This process is parallelized through Map operations on the Spark platform, and the weight factors used are usually dynamically adjusted between 0.5 and 2.0. The weighted feature importance data finally outputted can more accurately reflect the actual contribution of each feature in the diagnosis of schizophrenia.

[0072] Step S54: determining feature selection thresholds at different levels based on the distribution characteristics and quantity of features according to the weighted feature importance data, thereby obtaining hierarchical threshold data; This embodiment is based on the weighted feature importance data obtained in step S53, and performs a detailed analysis on the statistical characteristics of feature distribution at different levels (such as mean, standard deviation and quantile), and adopts an adaptive threshold determination method. In the specific operation, a histogram and a box plot of the feature importance data are first drawn to identify outliers and central trends in the data, and then the feature selection threshold of each level is determined according to the preset distribution characteristics (such as selecting the 75th percentile or using the mean plus a standard deviation as the threshold). In terms of parameter setting, in the application scenario, the threshold can be set at an upper-middle level of the weighted importance data, such as 80% of the cumulative distribution position, to ensure that only features with higher discrimination are screened out, and finally the threshold data corresponding to each level is saved as an independent threshold set.

[0073] Step S55: using the hierarchical threshold data to screen the weighted feature importance data, thereby obtaining a candidate feature set; This embodiment uses the hierarchical threshold data obtained in step S54 to screen the weighted feature importance data. The specific operation adopts a layer-by-layer comparison method to compare the weighted score of each functional connection feature with the threshold of the corresponding layer. If the feature score is higher than the threshold, it is retained, otherwise it is removed. At the same time, the filter operation (Filter) is used in the Spark distributed environment to quickly screen the data set to ensure balanced data distribution and feature representativeness. The screened candidate feature set is saved in a matrix form, where each column represents a screened functional connection feature. The candidate feature set provides preliminary input for further optimization and redundancy analysis. The parameters usually require that the number of features contained in the candidate feature set is between 10% and 30% of the original feature set to balance information richness and computational burden.

[0074] Step S56: performing redundancy analysis on the candidate feature set, calculating the correlation between features, and removing highly correlated redundant features according to a preset threshold, thereby obtaining an optimized feature set; This embodiment performs redundancy analysis on the candidate feature set screened in step S55. In the specific operation, the Pearson correlation coefficient matrix between the features is calculated using the correlation analysis method, and all feature pairs in the matrix are traversed to identify those feature pairs with correlation coefficients higher than a preset threshold (such as 0.9). These features are considered to be highly redundant. Then, a stepwise elimination method is used to preferentially retain features that perform more prominently in Fisher scores and weighted importance. At the same time, principal component analysis (PCA) is used to assist in verifying the contribution rate and independence of the features. The optimized feature set formed after eliminating redundant features has lower multicollinearity and higher information uniqueness. The entire process is performed in a Spark environment using distributed matrix operations to ensure efficiency and accuracy when processing large-scale data.

[0075] Step S57: Matching and testing the optimized feature set with the prior knowledge of schizophrenia to generate the biological significance and clinical interpretability of the optimized features, thereby obtaining the schizophrenia feature set.

[0076] In this embodiment, the optimized feature set obtained in step S56 is matched and verified with the prior knowledge of schizophrenia established in step S52. Specifically, by comparing the physiological and clinical descriptions corresponding to each optimized feature in the prior knowledge base, the biological significance of the feature is verified using text matching and rule-based methods. For example, it is checked whether the feature involves abnormal brain networks known in schizophrenia (such as the default mode network or the central executive network) and whether its connection pattern is consistent with literature reports. At the same time, neuroscience experts are invited to conduct qualitative evaluation to ensure that each optimized feature has clinical interpretability and biological significance. Finally, the features that have passed the matching verification are integrated to form a schizophrenia feature set, which is output in the form of a multidimensional array and annotated table for subsequent machine learning model training and clinical decision support system use. The parameters and matching rules of the entire process are dynamically adjusted according to specific clinical data and expert feedback to achieve the best matching effect.

Claims

1. An image feature selection method for the Spark model of schizophrenia medical records, characterized in that: The following steps are involved: Step S1: acquiring brain magnetic resonance image data and resting-state functional magnetic resonance imaging data of multiple subjects; Step S2: using a preset hierarchical brain region template to perform multi-scale hierarchical segmentation on the brain magnetic resonance image data of each subject, and obtain brain region positioning data including three levels: local, regional and global. The multi-scale hierarchical segmentation specifically includes subcortical structure segmentation at the local level, functional partitioning segmentation based on anatomical connection by calculating the structural connection probability and spatial adjacency between brain regions at the regional level, and large-scale brain network segmentation based on hierarchical organization analysis of resting-state networks at the global level; Step S3: performing adaptive hierarchical denoising on the resting-state functional magnetic resonance imaging data in a Spark distributed environment to obtain denoised brain function time series data; Step S4: construct a hierarchical functional connection calculation framework based on the brain area positioning data and brain function time series data, calculate the functional connection strength within each level and between levels, and dynamically adjust the connection weights between levels based on the principle of synaptic plasticity to obtain brain function network feature data; Step S5: Performing a hierarchical feature importance evaluation based on Fisher score on the brain functional network feature data, and setting a hierarchical adaptive threshold for feature screening according to the pre-acquired schizophrenia prior knowledge to obtain a schizophrenia feature set.

2. The image feature selection method for the schizophrenia medical record Spark model according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: performing T1-weighted magnetic resonance scanning on the brains of multiple subjects using a 3T magnetic resonance imaging device, thereby obtaining brain magnetic resonance image data, wherein the imaging contrast of the brain magnetic resonance image data is specifically determined by the T1 relaxation time; Step S12: using a 7T superconducting magnetic resonance imaging device to perform blood oxygen level-dependent contrast acquisition based on a gradient echo sequence on the brains of multiple subjects in a resting state, thereby obtaining blood oxygen level-dependent contrast data; Step S13: Analyze the local blood flow changes caused by neuronal activities on the blood oxygen level dependence contrast data, so as to obtain resting state functional magnetic resonance imaging data.

3. The image feature selection method for the schizophrenia medical record Spark model according to claim 2, characterized in that: Step S2 includes the following steps: Step S21: performing image resolution adaptive normalization processing on the brain magnetic resonance image data, and obtaining normalized image data with a spatial resolution of 1 mm isotropic voxel by resampling the voxel size based on B-spline interpolation; Step S22: performing subcortical structure segmentation on the standardized image data according to a preset hierarchical brain region template, and performing brain region positioning processing, thereby obtaining subcortical structure positioning data at a local level; Step S23: performing functional partitioning clustering based on anatomical connection by calculating the structural connection probability and spatial adjacency relationship between brain regions according to the local-level subcortical structure positioning data, thereby obtaining functional partitioning data at the regional level; Step S24: performing a hierarchical organization analysis of the cerebral cortex based on the resting-state network according to the functional partitioning data and the resting-state functional magnetic resonance imaging data, thereby obtaining large-scale brain network data at the global level, wherein the hierarchical organization analysis specifically calculates the time series correlation and spatial continuity between the functional partitions; Step S25: hierarchically integrate the subcortical structure positioning data, functional partitioning data, and large-scale brain network data to establish a correspondence matrix between the hierarchies, thereby obtaining brain region positioning data.

4. The image feature selection method for the schizophrenia medical record Spark model according to claim 3, characterized in that: Step S22 includes the following steps: Step S221: performing edge enhancement processing by calculating the spatial gradient amplitude and direction of the three-dimensional image according to the standardized image data, thereby obtaining boundary enhancement data of the subcortical structure; Step S222: Analyze the grayscale distribution characteristics of the local area according to the boundary enhancement data, set a dynamic threshold, and perform segmentation, thereby obtaining preliminary segmentation mask data of the subcortical structure; Step S223: morphologically optimizing the preliminary segmentation mask data by using morphological operations, so as to obtain continuous and complete subcortical structure mask data, wherein the morphological optimization specifically processes holes and boundary discontinuities by designing a three-dimensional morphological operator sequence; Step S224: Perform connected domain analysis based on the subcortical structure mask data and mark different subcortical structures, so as to obtain fine subcortical structure positioning data at the local level.

5. The image feature selection method for the schizophrenia medical record Spark model according to claim 4, characterized in that: Step S23 includes the following steps: Step S231: performing connection fiber bundle tracing on the local level subcortical structure positioning data, and performing white matter fiber connection analysis between subcortical structures, thereby obtaining a structural connection probability map; Step S232: Calculating the spatial adjacency relationship between brain regions based on the spatial distance metric according to the structural connection probability map, thereby obtaining brain region spatial relationship data; Step S233: performing weighted fusion on the structural connection probability map and the brain region spatial relationship data based on the structural connection relationship and the spatial adjacency relationship, thereby obtaining a comprehensive relationship matrix; Step S234: partition the comprehensive relationship matrix using a spectral clustering algorithm, thereby obtaining functional partition data at the regional level.

6. The image feature selection method for the schizophrenia medical record Spark model according to claim 5, characterized in that: Step S25 includes the following steps: Step S251: extracting the spatial position, volume and shape features of each structure from the subcortical structure positioning data, and constructing a feature vector to obtain a local hierarchical feature matrix; Step S252: Analyze the functional connection pattern and spatial distribution characteristics of each partition according to the functional partition data, so as to obtain a regional level feature matrix; Step S253: extracting network features including topological attributes and dynamic features from the large-scale brain network data, thereby obtaining a global hierarchical feature matrix; Step S254: local level feature matrix, regional level feature matrix and global level feature matrix are processed to obtain brain region localization data.

7. The image feature selection method for the schizophrenia medical record Spark model according to claim 6, characterized in that: Step S3 includes the following steps: Step S31: partitioning the resting state functional magnetic resonance imaging data in the Spark distributed environment, dividing the data into multiple time windows based on the time series, thereby obtaining time window series data; Step S32: performing multi-scale decomposition based on wavelet transform on the time window sequence data, decomposing the noise signal into different frequency bands, thereby obtaining wavelet frequency band coefficient data; Step S33: evaluating the signal-to-noise ratio of each frequency band according to the wavelet frequency band coefficient data, and performing adaptive noise threshold determination, thereby obtaining frequency band noise threshold data; Step S34: using the frequency band noise threshold data to perform soft threshold shrinkage processing on the wavelet frequency band coefficient data to suppress the noise component, thereby obtaining the denoised wavelet coefficient data; Step S35: reconstruct the wavelet coefficient data by inverse wavelet transform, and perform parallel calculation in the Spark environment to obtain preliminary denoised sequence data; Step S36: performing physiological noise removal processing including heartbeat and breathing on the preliminary noise reduction sequence data, thereby obtaining noise correction sequence data; Step S37: performing head motion parameter correction and spatial smoothing processing on the noise-corrected sequence data to obtain brain function time series data.

8. The image feature selection method for the schizophrenia medical record Spark model according to claim 7, characterized in that: Step S4 includes the following steps: Step S41: partitioning the brain function time series data according to the brain region positioning data in the Spark distributed environment, and standardizing the time series in each level to obtain standardized time series data; Step S42: Calculate the functional connectivity strength between subcortical structures in the local level according to the standardized time series data, and quantify it through Pearson correlation coefficient analysis to obtain a local level functional connectivity matrix; Step S43: Calculate the functional connectivity of the functional partitions at the regional level according to the local level functional connectivity matrix, and perform weighted integration based on the time series correlation of the nodes inside the region, so as to obtain the regional level functional connectivity matrix; Step S44: evaluating node importance and connection significance through network theory methods according to the regional level functional connection matrix, and determining the functional connection pattern of the global level large-scale brain network, thereby obtaining the global level functional connection matrix; Step S45: Dynamically adjust the inter-level weights of the local hierarchical functional connection matrix, the regional hierarchical functional connection matrix, and the global hierarchical functional connection matrix through the principle of synaptic plasticity, establish a mapping relationship of hierarchical connections, and thus obtain brain functional network characteristic data.

9. The image feature selection method for the schizophrenia medical record Spark model according to claim 8, characterized in that: Step S45 includes the following steps: Step S451: normalizing the local level functional connectivity matrix, the regional level functional connectivity matrix, and the global level functional connectivity matrix, and integrating the data in the Spark distributed environment to construct the level connection mapping matrix data; Step S452: constructing a weight update mathematical model based on the synaptic plasticity principle according to the hierarchical connection mapping matrix data, initializing each cross-level connection weight, thereby obtaining initial cross-level connection weight matrix data; Step S453: performing synchronization analysis on the time series activation of cross-level nodes according to the initial cross-level connection weight matrix data, thereby obtaining preliminary dynamic connection weight data, wherein the synchronization analysis includes calculating the update factors of positive reinforcement and reverse inhibition; Step S454: performing iterative dynamic weight adjustment processing on the preliminary dynamic connection weight data in the Spark distributed environment, and correcting the connection weight according to the preset global convergence condition, so as to obtain cross-level connection weight matrix data; Step S455: Establishing a hierarchical connection mapping relationship according to the cross-hierarchical connection weight matrix data, thereby obtaining brain function network feature data.

10. The image feature selection method for the schizophrenia medical record Spark model according to claim 9, characterized in that: Step S5 includes the following steps: Step S51: grouping the brain function network feature data into three levels: local, regional and global, and calculating the Fisher discriminant score for the functional connection features of each level, thereby obtaining hierarchical Fisher score data; Step S52: establishing a knowledge base of brain functional connection patterns related to schizophrenia based on the pre-collected clinical data of schizophrenia patients and healthy controls, thereby obtaining prior knowledge of schizophrenia; Step S53: setting differentiated importance weights for functional connectivity features at different levels based on prior knowledge of schizophrenia, and weighting the hierarchical Fisher score data to obtain weighted feature importance data; Step S54: determining feature selection thresholds at different levels based on the distribution characteristics and quantity of features according to the weighted feature importance data, thereby obtaining hierarchical threshold data; Step S55: using the hierarchical threshold data to screen the weighted feature importance data, thereby obtaining a candidate feature set; Step S56: performing redundancy analysis on the candidate feature set, calculating the correlation between features, and removing highly correlated redundant features according to a preset threshold, thereby obtaining an optimized feature set; Step S57: Matching and testing the optimized feature set with the prior knowledge of schizophrenia to generate the biological significance and clinical interpretability of the optimized features, thereby obtaining the schizophrenia feature set.

Images