A method for detecting schizophrenia neurovascular coupling abnormalities and assessing disease progression

Abstract

This invention discloses a method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia, belonging to the field of medical image processing and brain function analysis technology. It provides a method for estimating the hemodynamic response function (HRF) based on resting-state fMRI, used to quantitatively characterize neurovascular coupling (NVC) abnormalities in schizophrenia and to depict the dynamic changes in the disease as it progresses. By combining multi-center, large-sample resting-state fMRI data, a voxel-level estimation method for HRF using blind-source deconvolution, a disease progression sliding window grouping strategy, structural / functional imaging indicators, and clinical symptom scales, this invention constructs a framework for analyzing schizophrenia progression in cross-sectional, large-sample studies, replacing longitudinal studies. This framework can identify temporal patterns of brain tissue, neurovascular coupling, and functional impairment during disease progression, providing a new technical path for exploring the degenerative pathological mechanisms of schizophrenia.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing and brain function analysis technology, and relates to a method for quantifying neurovascular coupling (NVC) based on resting-state functional magnetic resonance imaging (fMRI). In particular, it relates to a comprehensive method for estimating the hemodynamic response function (HRF) through blind source deconvolution and using its parameter characteristics for schizophrenia progression assessment, brain structure-function coupling analysis and clinical prediction. Background Technology

[0002] Schizophrenia is a chronic mental disorder affecting approximately 1% of the global population. Although traditionally not classified as a neurodegenerative disease, numerous longitudinal imaging studies have revealed progressive brain structural damage in schizophrenia patients, particularly a progressive reduction in gray matter volume. Meta-analyses show that, compared to healthy controls, patients experience annual shrinkage of the entire brain, gray matter, and lateral ventricles, with these structural changes persisting for at least 20 years after onset. Simultaneously, functional imaging studies have further revealed the evolutionary trajectory of brain dysfunction in schizophrenia, with abnormal patterns gradually expanding from the subcortical and sensorimotor areas to the higher association cortex. This converging evidence supports a pathological framework for schizophrenia that resembles a neurodegenerative process. However, the lack of reliable biomarkers to characterize this progressive pathology or its underlying mechanisms limits further development in validating disease mechanisms, early intervention, and the construction of models of mental illness progression.

[0003] NVC (neural vascular perfusion) refers to the dynamic coordination between neuronal activity and local cerebral blood flow (CBF), a key physiological mechanism for maintaining brain function and metabolic homeostasis. NVC is regulated by a multi-component system including neurons, glial cells, and blood vessels, and can adaptively regulate perfusion according to metabolic needs. Existing studies have linked NVC disruption to various age-related and neurodegenerative diseases, and there is a growing consensus that NVC abnormalities may play an important role in the pathology of schizophrenia. For example, reduced cerebral blood flow can be observed in areas such as the superior temporal gyrus, cingulate gyrus, and middle frontal gyrus in patients with severe negative symptoms. Studies based on arterial spin labeling (ASL) and single-photon emission computed tomography (SPECT) have also repeatedly reported hypoperfusion in the frontal lobe and cingulate cortex of schizophrenia patients, while hyperperfusion is observed in some subcortical and cerebellar regions. These perfusion abnormalities suggest underlying vascular or glial pathology, thus proposing that NVC indicators may be important biological features reflecting disease-related brain changes.

[0004] Blood oxygen level dependent (BOLD) signals in fMRI are an indirect reflection of neural activity, and their generation process can be viewed as the convolution of potential neural events with the blood oxygen refractory signal (HRF). Therefore, morphological changes in the HRF can reflect the multi-level dynamics of neural, glial, and vascular components in the neurovascular coupling system, which is of great significance for understanding neural metabolic processes and cerebrovascular health. Traditionally, HRF extraction mainly relies on task-based fMRI, inducing activation of local brain regions through specific stimulus paradigms, and then estimating the corresponding HRF based on model-driven methods. However, these methods have two core limitations: first, they must rely on pre-designed cognitive or sensory tasks, making them difficult to universally apply to routinely acquired clinical data; second, the task paradigm can only activate specific brain regions, resulting in the estimation range of HRF being limited to the activated local areas, and failing to achieve a systematic characterization of neurovascular coupling features across the entire brain. In recent years, with the development of data-driven blind source deconvolution technology, task-free HRF estimation has become possible, providing a new path for calculating whole-brain voxel-level HRF in resting-state fMRI data. However, the related technical system is still under development and lacks a robust toolchain suitable for large-scale heterogeneous clinical studies. Furthermore, previous studies have also faced methodological bottlenecks in the imaging exploration of disease progression. Traditional approaches often involve meta-analysis or correlation analysis of multiple functional indicators to identify imaging features related to disease progression. Although such methods can integrate key results from different studies, they only yield single-dimensional markers linearly correlated with disease progression, making it difficult to reveal specific functional patterns corresponding to different stages of the disease. Moreover, while longitudinal designs are considered the gold standard for studying disease evolution trajectories, obtaining large-scale longitudinal imaging data covering several years or even decades is extremely difficult in reality, limited by factors such as cost, follow-up losses, and patient compliance. Therefore, in the absence of long-term longitudinal data, there is an urgent need for an alternative research framework that can reasonably infer the dynamic evolution of disease from large-sample cross-sectional data and identify neurobiological features with progression significance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies in the study of schizophrenia progression by providing a resting-state fMRI-based HRF estimation method for quantitatively characterizing NVC abnormalities in schizophrenia and depicting the dynamic changes in the disease as it progresses. This invention combines multi-center, large-sample resting-state fMRI data, a blind-source deconvolution HRF voxel-level estimation method, a disease progression sliding window grouping strategy, structural / functional imaging indicators, and clinical symptom scales to construct a framework for analyzing schizophrenia progression in cross-sectional, large-sample studies, replacing longitudinal studies. This framework can identify temporal patterns of brain tissue, neurovascular coupling, and functional impairment during disease progression, providing a new technical approach for exploring the degenerative pathological mechanisms of schizophrenia.

[0006] The objective of this invention is achieved through the following technical solution: a method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia, comprising the following steps:

[0007] Step 1: Collect multicenter MRI data from schizophrenia patients and healthy controls, including: resting-state fMRI, structural T1 images, and DTI data, DTI diffusion tensor imaging. A unified preprocessing workflow was used for all image features, and batch effect correction was performed on image features from different centers using the ComBat method to obtain comparable whole-brain structural and functional features.

[0008] Step 2: Spontaneous neural events are identified from resting-state fMRI data using a blind-source deconvolution model, and HRF estimation is performed on whole-brain voxels based on this. The peak amplitude parameter HRF Height is extracted from the estimated HRF. HRF Height represents the maximum response amplitude of the HRF curve in the time dimension, which is used to characterize the intensity of hemodynamic response caused by a unit neural event and serves as a quantitative indicator of NVC. HRF represents the hemodynamic response function, and NVC represents neurovascular coupling.

[0009] Step 3: Based on the HRF Height parameters obtained in Step 2, perform neurovascular coupling abnormality detection in the whole brain for both the patient group and the healthy control group;

[0010] At the voxel level, HRF Height values ​​were extracted for each voxel from the patient group and the healthy control group. Under the condition of controlling for age and gender, a two-sample t-test was used to calculate the difference in HRF Height between the two groups at that voxel, and the corresponding t-statistic atlas was obtained. To control the false positive risk of multiple comparisons, the statistical results at the whole-brain voxel level were corrected for multiple comparisons using the false discovery rate (FDR) correction method to obtain the corrected significance results. The voxel set that passed the correction and showed significant differences between groups was defined as the neurovascular coupling abnormal region, thereby constructing a spatial distribution atlas of whole-brain HRF Height abnormalities related to schizophrenia.

[0011] Step 4: Sort patient samples according to the length of the disease course, and construct overlapping disease course subgroups using a sliding window strategy with a fixed window width of 5 years and a step size of 1 year. In each disease course subgroup, perform a two-sample t-test on HRF height for patients and healthy controls at the voxel level, and use the multiple comparison correction method consistent with Step 3 to identify significantly abnormal voxels. For each voxel in the whole brain, record the disease course window number in which it first showed a significant HRF height abnormality relative to healthy controls in a certain disease course subgroup, and map this number back to the corresponding voxel position, thereby constructing a whole-brain "disease course initiation atlas" to describe the order in which neurovascular coupling abnormalities appear in different brain regions as the disease progresses.

[0012] Step 5: Calculate the brain structure and function indicators for each subject and analyze their correlation with HRF Height; the brain structure and function indicators include: gray matter volume, white matter FA, average low-frequency amplitude, and regional consistency ReHo;

[0013] At the voxel or brain region level, Pearson correlation coefficients were calculated between HRF Height and white matter FA, the average power spectrum amplitude in the low-frequency band, and regional consistency to obtain the corresponding correlation distribution maps. Steiger's Z test was used to compare the differences in correlation coefficients between the patient group and the healthy control group. All brain structures and brain function indicators with correlation coefficients greater than a set threshold and inter-group differences greater than a set threshold were included.

[0014] Step 6: Using the brain structure and function indicators obtained in Step 5 as input variables, use the XGBoost regression model based on gradient boosting decision trees to predict the patient's disease course or symptom score; find the HRF Height corresponding to the brain structure and function indicators calculated in Step 5, and then calculate the Spearman rank correlation coefficient between the corresponding HRF Height and the PANSS positive symptom score, negative symptom score, and psychopathology score. The absolute value of the correlation coefficient is used to quantify the association strength between HRF Height and clinical symptoms. If the strength is greater than the set threshold, the score is accepted; otherwise, the score is not accepted.

[0015] Furthermore, the formula for calculating white matter FA in step 5 is as follows:

[0016] ;

[0017] ;

[0018] in, , , These are the three principal eigenvalues ​​of the diffusion tensor; It is the average diffusion rate, and the value of FA ranges from [0,1]. The larger the value, the stronger the diffusion directionality and the more complete the protein fiber structure.

[0019] Furthermore, in step 5, ALFF is defined as the square root of the power spectrum of the voxel BOLD time series in the 0.01–0.1 Hz frequency band. Let the BOLD time series of a certain voxel be... Its Fourier transform for:

[0020] ;

[0021] ALFF is defined as the average value of the power spectrum amplitude in the low-frequency range:

[0022] ;

[0023] The low-frequency band F ranges from [0.01Hz to 0.1Hz]. A higher ALFF value indicates a higher intensity of local spontaneous neural activity.

[0024] Furthermore, in step 5, ReHo obtains the Kendall coordination coefficient by calculating the time series of a voxel and its neighboring voxels.

[0025] ;

[0026] ;

[0027] in For the first The time point, the first The rank of individual elements; For the first The sum of the ranks of all voxels at time points; K is the number of neighboring voxels; n represents the number of time points; the larger the ReHo value, the higher the temporal synchronicity of neural activity in the local area.

[0028] Furthermore, based on the results of step 6, the correlation coefficient between the predicted values ​​and the true values ​​is calculated through cross-validation to evaluate the clinical predictive ability of HRF features.

[0029] Furthermore, based on the results of step 6, a mediation model was constructed between disease course, structure, neurovascular coupling, and function, with disease course as the independent variable, brain structural indicators and HRF Height as mediating variables, and brain functional indicators as the dependent variable. Regression analysis was used to estimate the coefficients of each path, and the confidence interval of the mediation effect was calculated through 5000 nonparametric bootstrap resampling. When the 95% confidence interval of the indirect effect did not include zero, HRF Height was determined to have a significant statistical mediating effect between disease course-related structural damage and functional abnormalities.

[0030] The beneficial effects of this invention are as follows: This invention is the first to realize a whole-brain voxel-level HRF estimation method based on resting-state fMRI, breaking through the limitation of traditional task-based HRF which can only cover focal activation regions, and can comprehensively and quantitatively obtain NVC features. Simultaneously, it proposes a disease progression modeling framework based on a disease progression sliding window, using cross-sectional data instead of longitudinal design. This framework can accurately characterize the dynamic evolutionary path of schizophrenia in large-sample cross-sectional data, revealing the temporal patterns of gray matter, white matter, and primary to higher brain networks. Furthermore, this invention clarifies the key mediating role of NVC abnormalities between structural atrophy and functional decline, and confirms that HRF features have the potential to predict disease course and the severity of clinical symptoms, thus providing new technical means for the mechanistic elucidation, progression monitoring, and biomarker development of schizophrenia. Overall, the multimodal fusion analysis system of imaging and symptoms constructed in this invention provides an innovative strategy with broad application value for the mechanistic research and clinical application of schizophrenia and other neuropsychiatric diseases. Attached Figure Description

[0031] Figure 1 This is a flowchart of the method for detecting neurovascular coupling abnormalities and assessing disease progression according to the present invention.

[0032] Figure 2 The results show the intergroup differences in HRF between patients with schizophrenia and healthy controls.

[0033] Figure 3 A disease progression map of NVC abnormalities in schizophrenia patients. Detailed Implementation

[0034] This invention proposes a method for HRF estimation and dynamic assessment of NVC in schizophrenia based on resting-state fMRI. Through multi-center image data integration, blind-source deconvolution HRF estimation, disease progression sliding window analysis, and multimodal correlation modeling of images and clinical data, it achieves a systematic quantification of whole-brain NVC abnormalities in schizophrenia and their evolutionary patterns throughout the disease course. The method includes the following steps:

[0035] Step 1: This invention first acquires large-sample multimodal imaging data of schizophrenia patients and healthy controls from five independent multicenter datasets, including structural T1 MRI, resting-state fMRI, and DTI data; the five datasets contain a total of 1191 participants, covering multicenter samples from different regions, using different scanners, and with different acquisition protocols. All patients were clinically diagnosed according to the DSM-IV criteria, and clinical symptoms were assessed using the PANSS scale. All participants had no neurological diseases, substance dependence, family history of mental illness, or contraindications to MRI, and underwent imaging scans after obtaining informed consent. This step ensures that this invention is based on large-sample multicenter data in a real clinical environment for NVC assessment, providing a foundation for subsequent cross-site model construction;

[0036] Step 2: To ensure comparability across datasets, this invention employs a consistent preprocessing procedure for all participants' data. Resting-state fMRI data were processed using DPARSF, SPM12, and a publicly available white matter functional analysis platform. The process sequentially included deletion of the first five time points, temporal correction, head movement correction, structural image segmentation and gray matter analysis, white matter and cerebrospinal fluid mask construction, noise covariate regression, spatial smoothing in gray matter and white matter, standard spatial normalization, and 0.01–0.1 Hz bandpass filtering. Participants with head movements exceeding 3 mm or 3° were excluded. To avoid contamination between gray matter and white matter signals, gray matter and white matter data were processed independently within their respective masks. DTI data underwent eddy current correction, field map correction, and tensor fitting using the FSL toolbox, and was aligned to the MNI space through linear and nonlinear registration. This step ensures consistent spatial and temporal standardization of all functional and structural data across large-scale samples from multiple scanners and projects.

[0037] Step 3: After preprocessing, this invention calculates structural and functional imaging parameters across the entire brain. Structural features include gray matter volume calculated using the CAT12 toolbox and white matter FA based on DTI, reflecting tissue morphology and white matter fiber microstructure. Functional features include ALFF and ReHo, reflecting local neural activity intensity and time-series synchronicity, respectively. All features are calculated in the MNI space and satisfy cross-dataset consistency. This step provides foundational data support for subsequent HRF-structure-function coupling analysis.

[0038] Step 4: To quantify NVC, this invention employs blind-source deconvolution to estimate HRF at the voxel level from resting-state fMRI. Specifically, spontaneous event points with amplitudes exceeding the mean plus 1.5 standard deviations of the voxel's time series are detected in the BOLD time series of each voxel, constructing an event sequence vector. This event sequence is then convolved with a function set containing standard HRF basis functions and their time and dispersion derivatives to model the optimal HRF for that voxel using least-squares fitting. The peak amplitude parameter HRF Height is extracted from the HRF, defined as the maximum response amplitude of the HRF curve in the time dimension, used to characterize the intensity of the hemodynamic response caused by a unit neural event, serving as a quantitative indicator of NVC. To eliminate multi-center scan differences, this invention uses the ComBat algorithm to perform batch effect correction on all HRF parameters, ensuring cross-site comparability. This step achieves quantification of HRF across the entire brain, including both gray and white matter, and is the core technical aspect of this invention.

[0039] Step 5: After obtaining the HRF height, this invention performs HRF difference analysis between patients and healthy controls at the tissue, regional, and voxel levels. At the tissue level, the average HRF height of gray and white matter is compared; at the regional level, the average HRF of each region is calculated according to predetermined partitions (10 white matter networks, 7 cortical networks, subcortical structures, and cerebellum); at the voxel level, the differences are compared voxel-by-voxel across the entire brain. Two-sample t-tests are used at all levels, and FDR correction is applied. The results show that schizophrenic patients exhibit significant HRF decreases in the visual network, somatosensory-motor network, thalamus, cerebellum, and major white matter fiber tracts, while some limbic system regions show HRF increases. This step reveals a widespread and specific NVC abnormality pattern across the entire brain in schizophrenia.

[0040] Step 6: To characterize the evolutionary trajectory of NVC in schizophrenia throughout its course, this invention divides participants into 26 overlapping disease course subgroups based on a 5-year window width and a 1-year step size. HRF comparisons between patients and healthy controls are performed in each disease course group. Subsequently, the disease course window where the first significant abnormality appears is recorded in each voxel of the whole brain, thereby creating a "disease course initiation atlas." Analysis reveals that the early disease course primarily affects the thalamus, lingual gyrus, and precentral and anterior central gyruses, subsequently spreading to higher cortical networks, starting in the superficial pivotal regions of the white matter, and ultimately affecting the deep white matter and transhemispheric connective tissue. This step is the first to construct a dynamic disease course model of schizophrenia NVC using large-sample cross-sectional data.

[0041] Step 7: This invention further evaluates the coupling relationship between HRF height and brain structure and function, including Pearson correlation analysis between HRF height and gray matter volume, white matter femoral head (FA), white matter upper lip fat (ALFF), and Reho (ReHo). Analysis was performed in both the patient group and the healthy control group, and Steiger's Z-test was used to differentiate between groups. Results showed that HRF height exhibited a stable positive correlation with gray matter volume, ALFF, and Reho, while HRF height in white matter did not show a significant association with FA. This step reveals the close link between decreased NVC and structural atrophy and functional decline.

[0042] Step 8: To further explore the relationship between HRF and disease severity, this invention used Spearman analysis to assess the correlation between HRF height and PANSS positive / negative symptoms, general symptoms, and disease course. The results showed a significant negative correlation between HRF height and negative symptoms, most pronounced in the frontal lobe, cingulate gyrus, temporal lobe, and precuneus. Furthermore, an XGBoost-based regression model demonstrated that HRF features can effectively predict patient disease course, with the combined gray and white matter HRF features exhibiting the highest predictive accuracy, indicating that HRF can serve as an imaging marker related to the course of schizophrenia. This step explored the potential application of HRF in clinical assessment.

[0043] Step 9: Finally, this invention constructs a mediation model of disease course (independent variable), structure and HRF (mediation), and function (dependent variable), using 5000 bootstrap iterations to obtain confidence intervals for the mediation effect. The results show that in gray matter, HRF is a significant mediator of the impact of disease course on function, and structure and HRF also have a joint mediating effect; in white matter, only HRF plays a mediating role. This analysis reveals that disease progression may occur through a chain reaction mechanism of "structural atrophy leading to NVC decline and then to functional weakening." This step provides mechanistic evidence for understanding the multi-scale neural mechanisms involved in the progression of schizophrenia.

[0044] In this embodiment, a total of 1108 subjects were included after quality control, including 498 patients with schizophrenia and 610 healthy controls. Experimental results showed that the method proposed in this invention can stably and reliably estimate HRF parameters across the entire brain, identify widespread NVC decline areas in schizophrenia patients, and reveal a progressive pattern of disease involvement from the primary cortex to higher networks through dynamic disease progression analysis. Furthermore, HRF height reflects various structural and functional impairments and is highly correlated with negative symptoms and disease progression; mediation analysis further illustrates the potentially crucial role of NVC abnormalities in disease progression. This invention provides a systematic methodological foundation for the assessment of schizophrenia progression, understanding of neural mechanisms, and development of potential biomarkers.

[0045] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia, comprising the following steps: Step 1: Collect multicenter MRI data from schizophrenia patients and healthy controls, including resting-state fMRI, structural T1 images, and diffusion tensor imaging (DTI) data. Use a uniform preprocessing workflow for all image features and use the ComBat method to perform batch effect correction on image features from different centers to obtain comparable whole-brain structural and functional features. Step 2: Spontaneous neural events are identified from resting-state fMRI data using a blind-source deconvolution model, and HRF estimation is performed on whole-brain voxels based on this. The peak amplitude parameter HRF Height is extracted from the estimated HRF. HRF Height represents the maximum response amplitude of the HRF curve in the time dimension, which is used to characterize the intensity of hemodynamic response caused by a unit neural event and serves as a quantitative indicator of NVC. HRF represents the hemodynamic response function, and NVC represents neurovascular coupling. Step 3: Based on the HRF Height parameters obtained in Step 2, perform neurovascular coupling abnormality detection in the whole brain for both the patient group and the healthy control group; At the voxel level, HRF Height values ​​were extracted for each voxel from the patient group and the healthy control group. Under the condition of controlling for age and gender, a two-sample t-test was used to calculate the difference in HRF Height between the two groups at that voxel, and the corresponding t-statistic atlas was obtained. To control the false positive risk of multiple comparisons, the statistical results at the whole-brain voxel level were corrected for multiple comparisons using the false discovery rate (FDR) correction method to obtain the corrected significance results. The voxel set that passed the correction and showed significant differences between groups was defined as the neurovascular coupling abnormal region, thereby constructing a spatial distribution atlas of whole-brain HRF Height abnormalities related to schizophrenia. Step 4: Sort patient samples according to the length of the disease course, and construct overlapping disease course subgroups using a sliding window strategy with a fixed window width of 5 years and a step size of 1 year. In each disease course subgroup, perform a two-sample t-test on HRF height for patients and healthy controls at the voxel level, and use the multiple comparison correction method consistent with Step 3 to identify significantly abnormal voxels. For each voxel in the whole brain, record the disease course window number in which it first showed a significant HRF height abnormality relative to healthy controls in a certain disease course subgroup, and map this number back to the corresponding voxel position, thereby constructing a whole-brain "disease course initiation atlas" to describe the order in which neurovascular coupling abnormalities appear in different brain regions as the disease progresses. Step 5: Calculate the brain structure and function indices for each subject and analyze their correlation with HRF Height; the brain structure and function indices include: gray matter volume, white matter FA, low-frequency amplitude ALFF, and regional consistency ReHo; At the voxel or brain region level, Pearson correlation coefficients were calculated between HRF Height and white matter FA, ALFF, and ReHo to obtain the corresponding correlation distribution maps. Steiger's Z test was used to compare the differences in correlation coefficients between the patient group and the healthy control group. All brain structural and functional indicators with correlation coefficients greater than a set threshold and inter-group differences greater than a set threshold were included. Step 6: Using the brain structure and function indicators obtained in Step 5 as input variables, use the XGBoost regression model based on gradient boosting decision trees to predict the patient's disease course or symptom score; find the HRF Height corresponding to the brain structure and function indicators calculated in Step 5, and then calculate the Spearman rank correlation coefficient between the corresponding HRF Height and the PANSS positive symptom score, negative symptom score, and psychopathology score. The absolute value of the correlation coefficient is used to quantify the association strength between HRF Height and clinical symptoms. If the strength is greater than the set threshold, the score is accepted; otherwise, the score is not accepted.

2. The method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia as described in claim 1, characterized in that, The formula for calculating white matter FA in step 5 is as follows: ； ； in, , , These are the three principal eigenvalues ​​of the diffusion tensor; It is the average diffusion rate, and the value of FA ranges from [0,1]. The larger the value, the stronger the diffusion directionality and the more complete the protein fiber structure.

3. The method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia as described in claim 1, characterized in that, In step 5, ALFF is defined as the square root of the power spectrum of the voxel BOLD time series in the 0.01–0.1 Hz frequency band. Let the BOLD time series of a certain voxel be... Its Fourier transform for: ； ALFF is defined as the average value of the power spectrum amplitude in the low-frequency range: ； The low-frequency band F ranges from [0.01Hz to 0.1Hz]. A higher ALFF value indicates a higher intensity of local spontaneous neural activity.

4. The method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia as described in claim 1, characterized in that, In step 5, ReHo obtains the Kendall coordination coefficient by calculating the time series of a voxel and its neighboring voxels; ； ； in For the first The time point, the first The rank of individual elements; For the first The sum of the ranks of all voxels at time points; K is the number of neighboring voxels; n represents the number of time points; the larger the ReHo value, the higher the temporal synchronicity of neural activity in the local area.

5. The method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia as described in claim 1, characterized in that, Based on the results of step 6, the correlation coefficient between the predicted and actual values ​​is calculated through cross-validation to evaluate the clinical predictive ability of the HRF parameters.

6. The method for detecting neurovascular coupling abnormalities and assessing disease progression in schizophrenia as described in claim 1, characterized in that, Based on the results of step 6, a mediation model was constructed between disease course, structure, neurovascular coupling, and function. Disease course was used as the independent variable, brain structural indicators and HRF Height were used as mediating variables, and brain functional indicators were used as the dependent variable. Regression analysis was used to estimate the coefficients of each path, and the confidence interval of the mediation effect was calculated by 5000 nonparametric bootstrap resampling. When the 95% confidence interval of the indirect effect did not include zero, HRF Height was determined to have a significant statistical mediating effect between disease course-related structural damage and functional abnormalities.

Images