Wilson disease brain injury prediction method based on multi-modal MRI (Magnetic Resonance Imaging) and gene analysis

Abstract

The invention discloses a Wilson disease brain injury prediction method based on multi-modal MRI (magnetic resonance imaging) and gene analysis, which combines functional magnetic resonance imaging, diffusion tensor imaging and three-dimensional T1 weighted imaging, carries out confluence analysis with ATP7B gene mutation data of a patient, and establishes a multi-layer network model to predict the brain injury of the Wilson disease. According to the method, the relationship between a gene expression mode and brain area network function characteristics is analyzed, biomarkers related to Wilson disease brain injury are identified, then a diagnosis model and treatment effect evaluation are optimized by utilizing a machine learning algorithm, subtle changes of brain structures and functions can be predicted before obvious lesion, and the timeliness and effect of treatment intervention can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis. Background Art

[0002] Wilson disease (WD), also known as hepatolenticular degeneration, is a rare autosomal recessive genetic disease caused mainly by mutations in the ATP7B gene, which leads to dysfunction of the copper transport protein. This dysfunction hinders the normal metabolism and excretion of copper ions, causing copper to accumulate in key organs such as the liver and brain, ultimately causing serious liver and neurological diseases. Due to copper accumulation, patients may experience behavioral and cognitive dysfunction, psychotic symptoms, and motor coordination disorders.

[0003] Although the course of Wilson disease can be effectively controlled by medication and dietary adjustments, the diagnostic challenge is the need for early identification and intervention before clinical symptoms become apparent. Currently, the diagnosis of Wilson disease mainly relies on clinical symptoms, examination of Kayser-Fleischer rings, and evaluation of copper levels in blood and urine and their derived biochemical indices. However, these diagnostic methods are often insensitive in the early stages when there are no symptoms or mild nonspecific symptoms, resulting in delayed diagnosis.

[0004] In addition, although traditional MRI technology can be used to evaluate brain structural changes, its sensitivity and specificity for early microlesions are insufficient, and significant brain damage can often be detected only after clinical symptoms appear. Based on these limitations, it cannot meet the needs of effective monitoring and intervention for early-stage and asymptomatic WD patients.

[0005] Based on this, a prediction method for brain damage in Wilson disease based on multimodal MRI and genetic analysis is needed. Summary of the invention

[0006] To achieve the above object, the present invention provides the following scheme: A method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis, comprising the following steps: S1, using functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI) and three-dimensional T1-weighted imaging (3D-T1WI) to perform a comprehensive scan of brain structure and function, and using an imaging algorithm to automatically detect and correct image distortion caused by patient head movement, the specific method is as follows:

[0007] Using a set of head motion parameters derived from historical imaging data, which involve common head movement patterns and amplitudes, to set detection thresholds and correction strategies;

[0008] Monitor the position and movement of the patient's head in real time during the MRI scan, using rapid imaging sequences to capture the head state with extremely short intervals between each sequence to track rapid movements and identify any head movement outside the normal range by comparing it with preset parameters;

[0009] When abnormal movement is detected, the necessary correction parameters are immediately calculated, including adjusting the angle, focus and other scanning parameters of the MRI scan. The correction operation is based on the deviation between the current head position and the expected scanning position, and the quality of subsequent images is continuously monitored.

[0010] S2. Collect ATP7B gene mutation data of WD patients, including gene sequence and expression level, and use bioinformatics tools to predict the specific effect of gene mutation on protein function. The specific method is as follows:

[0011] Collect mutation data on the ATP7B gene, integrate the location, type and known clinical manifestations of the mutation with relevant protein structure data and functional annotations in public databases, build a comprehensive data set, use the data on known gene mutations and their corresponding protein function effects to train a recurrent neural network model, and optimize its predictive ability by learning the complex relationship between the biochemical properties of mutations and functional changes;

[0012] Predict the functional impact of understudied ATP7B gene mutations and output the protein function changes caused by each mutation;

[0013] S3. Collect and integrate gene expression data and brain MRI imaging data of WD patients, associate gene expression patterns with functional characteristics of brain area networks, and construct a multi-layer network, one layer of which represents gene expression data and the other layer represents brain MRI data; by analyzing the interaction and connection between the two layers of data, associate gene expression patterns with functional characteristics of brain area networks, and identify key biomarkers associated with the pathological process of WD;

[0014] S4. Integrate the previously obtained data to construct a prediction model to predict brain structural and functional abnormalities in patients with Wilson disease and conduct in-depth analysis of the potential pathophysiological mechanisms. The specific steps are as follows:

[0015] The multimodal MRI imaging data, gene mutation data, and clinical data of the Leipzig scoring system collected in S1, S2, and S2 were integrated to construct a dataset, and the gene expression and brain imaging phenotypes across sampling points were analyzed by spatial correlation analysis to extract key imaging omics and genomics features;

[0016] Recurrent neural networks are used to process and analyze the integrated data. The recurrent neural networks capture the complex nonlinear relationships in the data and identify biomarkers that predict brain damage in Wilson disease.

[0017] Furthermore, the multimodal MRI data obtained in the S1 include functional magnetic resonance imaging (fMRI) data, on the basis of which time series analysis is performed to identify the activity patterns of brain functional networks in the resting state, including specific brain regions including the basal ganglia, thalamus and cerebellum, which are closely related to the neuropsychiatric symptoms of WD patients, and the functional connectivity matrix is ​​calculated to analyze the functional coupling strength between different brain regions, and to identify pairs of brain regions with abnormal functional connectivity, specifically including abnormal connections between the basal ganglia and the prefrontal cortex; then, graph theory analysis methods are applied to evaluate the small-world properties of the brain functional network, including network efficiency, modularity and node centrality, to quantify the degree of abnormality of the brain functional network of WD patients, compare the characteristics of the brain functional network of WD patients and healthy controls, and reveal specific changes in the brain functional network of WD patients.

[0018] Furthermore, the diffusion tensor imaging (DTI) data used in the S1 was used to perform fiber tracking on the DTI data to construct a brain white matter structural network. Specific brain regions included the corpus callosum, internal capsule, and thalamus. By analyzing the topological properties of the structural network, including path length, clustering coefficient, and global efficiency, abnormally connected brain regions in the structural network were identified, specifically including regions associated with cognitive and motor dysfunction in WD patients. By comparing the brain structural network properties of WD patients and healthy controls, specific changes in the brain structural network of WD patients were revealed.

[0019] Furthermore, the following analysis and prediction process is performed using the multimodal MRI data and gene data obtained in S1 and S3:

[0020] Combining the functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) data obtained from S1 and the ATP7B gene mutation data collected in S3, statistical and machine learning techniques are used to identify brain imaging phenotypic characteristics associated with specific gene mutations. A predictive model is constructed using a machine learning algorithm, and the genotype data is used as input to predict brain imaging phenotypes associated with Wilson disease, revealing how gene expression affects changes in brain structure and function.

[0021] Furthermore, a comprehensive diagnostic model was established by combining the MRI data obtained from S1, the gene mutation data in S2, the Leipzig score and other relevant clinical manifestations and laboratory test results obtained from clinical practice. Through ROC curve analysis and sensitivity-specificity analysis, the threshold setting of the diagnostic model was optimized to balance the sensitivity and specificity of the diagnosis.

[0022] Furthermore, in the step S1, multiple magnetic resonance brain imaging scans are performed on the Wilson disease patient before and after the copper removal treatment to monitor the specific effects of the treatment on the patient's brain function and structure, which specifically includes the following steps:

[0023] Functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and three-dimensional T1-weighted imaging (3D-T1WI) were performed regularly to capture changes in brain structure and function before and after treatment. The MRI data before and after treatment were analyzed, including the effects of brain area volume, functional connectivity, and structural network properties. The effects of treatment on brain function recovery were evaluated by analyzing the changes in brain network properties before and after treatment, specifically including the improvement of copper excretion therapy on abnormal brain functional networks. The treatment effect was comprehensively evaluated in combination with clinical symptoms and laboratory test results.

[0024] Furthermore, the S4 step further includes the collection and analysis of long-term follow-up data, and the specific steps are as follows: regularly collect clinical symptoms, laboratory test results and multimodal brain imaging data of Wilson disease patients, including clinical symptoms, laboratory test results and brain imaging changes, construct multidimensional prognostic evaluation indicators, use survival analysis methods to construct a prognostic prediction model, predict the patient's disease progression rate and survival period, and evaluate the stability and reliability of the prognostic model through external validation and sensitivity analysis.

[0025] Furthermore, the gene mutation analysis in step S2 specifically includes sequencing common missense mutations, nonsense mutations and splice site mutations in the ATP7B gene, identifying new or known pathogenic gene mutations by high-throughput sequencing technology, and using bioinformatics tools to predict the function of mutations and analyze the effects of these gene mutations on protein function.

[0026] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0027] 1. The present invention can significantly improve the early diagnosis ability of Wilson disease by combining multimodal magnetic resonance imaging and genetic analysis. Multimodal MRI technology enables doctors to observe subtle changes in brain structure and function at the microscopic level. In addition, combined with the detection of ATP7B gene mutations, it can confirm the existence of the disease at the genetic level. Even before the symptoms are obvious, it can capture the changes in the tiny structure and function of the brain, thereby achieving effective identification of asymptomatic or mildly symptomatic patients. It is particularly suitable for cases with atypical manifestations or complex genetic backgrounds, and can greatly advance the diagnosis time point and provide timely intervention measures for patients;

[0028] 2. The present invention can not only predict the progression of the disease by accurately analyzing the patient's gene mutation characteristics and their correlation with brain imaging phenotypes, but also assist doctors in formulating more personalized treatment plans for each patient. By using machine learning technology to analyze large amounts of data, it can identify which gene mutations are closely related to brain function recovery and treatment response, thereby guiding the use of specific drugs and treatment methods to achieve precision medicine. By monitoring the patient's brain function and structural changes before and after treatment, the treatment effect can be effectively evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 It is a flow chart of the steps of the present invention;

[0031] Figure 2 This is a machine learning model diagram of the present invention;

[0032] Figure 3 This is a schematic diagram of the brain function-structure network image feature extraction of the present invention (BD);

[0033] Figure 4 Schematic diagram of statistical tests performed for the accuracy, specificity, sensitivity and area under the ROC curve of the machine learning model for predicting WD patients of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Example

[0037] like Figure 1 As shown, a method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis in this embodiment includes the following steps:

[0038] S1. Use functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI) and three-dimensional T1-weighted imaging (3D-T1WI) to perform a comprehensive scan of brain structure and function, and use imaging algorithms to automatically detect and correct image distortions caused by patient head movement. The specific methods are as follows: Use a set of head motion parameters derived from historical imaging data. The parameters involve common head movement patterns and amplitudes to set detection thresholds and correction strategies. Monitor the position and movement of the patient's head in real time during the MRI scan. Use fast imaging sequences to capture the head state. The time interval between each sequence is extremely short to track rapid movements. By comparing with preset parameters, any head movement that exceeds the normal range is identified. When abnormal movement is detected, the necessary correction parameters are immediately calculated, including adjusting the angle, focus and other scanning parameters of the MRI scan. The correction operation is based on the deviation between the current head position and the expected scanning position, and the quality of subsequent images is continuously monitored.

[0039] S2. Collect ATP7B gene mutation data of WD patients, including gene sequence and expression level, and use bioinformatics tools to predict the specific effects of gene mutations on protein function. The specific methods are as follows: Collect mutation data on the ATP7B gene, integrate the location, type and known clinical manifestations of the mutation with relevant protein structure data and functional annotations in public databases, establish a comprehensive data set, use data on known gene mutations and their corresponding protein function effects to train a recurrent neural network model, optimize its predictive ability by learning the complex relationship between the biochemical properties of mutations and functional changes, predict the functional effects of under-studied ATP7B gene mutations, and output the protein function changes caused by each mutation.

[0040] S3. Collect and integrate gene expression data and brain MRI imaging data of WD patients, associate gene expression patterns with functional characteristics of brain area networks, and construct a multi-layer network, in which one layer represents gene expression data and the other layer represents brain MRI data; by analyzing the interaction and connection between the two layers of data, associate gene expression patterns with functional characteristics of brain area networks, and identify key biomarkers associated with the pathological process of WD.

[0041] At least 100 WD patients diagnosed by two senior neurologists with 10-15 years of work experience were collected and divided into two groups: liver type and brain type. In addition, in order to ensure the control validity of the research data, 50 healthy controls (HC) matched with WD patients in terms of gender, age and education level will be collected at the same time. Each participant will receive a comprehensive clinical evaluation, including the measurement of heart rate and blood pressure, as well as detailed blood and urine biochemical tests. The blood biochemical tests will cover blood routine, electrolyte levels, liver function, kidney function and coagulation indicators. In addition, special attention will be paid to the copper level and ceruloplasmin content in the serum, because these indicators are of great significance for the diagnosis of Wilson's disease. The urine test focuses on the quantitative analysis of 24-hour urine copper, which is one of the key tests for evaluating abnormal copper metabolism in the body. In addition to biochemical tests, each WD patient will also undergo a bilateral subcorneal Kayser-Fleischer ring examination and an ultrasonic examination of the digestive system to evaluate the impact of the disease on the eyes and digestive system.

[0042] S4. Integrate the data obtained before, predict the brain structure and functional abnormalities of Wilson disease patients by constructing a prediction model, and conduct in-depth analysis of the potential pathophysiological mechanisms. The specific steps are as follows: Use the multimodal MRI imaging data, gene mutation data and clinical data of the Leipzig scoring system collected in S1, S2 and S2 to integrate and construct a data set, analyze the gene expression and brain imaging phenotype across sampling points by spatial correlation analysis methods to extract key imaging genomics and genomics features, use recursive neural networks to process and analyze the integrated data, capture the complex nonlinear relationships in the data through recursive neural networks, and identify biomarkers that predict Wilson disease brain damage;

[0043] Using multimodal magnetic resonance imaging technology (conventional sequence, fMRI, DTI, 3D-T1WI) and imaging data analysis methods (structural network and functional network), we constructed brain functional network through BOLD-fMRI data and structural network through DTI data, and conducted correlation analysis between neuroimaging data and Leipzig score and genetic technology;

[0044] Resting-state BOLD-fMRI: The scanning sequence used a gradient echo planar echo sequence (GRE-EPI) and the latest MultiBandSENSE technology (MBFactor = 2). The specific parameters were as follows: MBFactor = 2, TR / TE = 800ms / 9.7ms, slice thickness = 3mm, number of scanning slices = 48, slice spacing = 0mm, FOV = 230mm×230mm, Flipangle = 52°, a total of 240 time points were scanned, and the duration was 3 minutes and 19 seconds;

[0045] DTI data acquisition: respiratory triggered scanning was used, with the following parameters: TR / TE = 3870 / 86ms, matrix = 112×109, FOV = 224×224mm2, voxel size = 2×2×2mm3, 1 B0 value, slice thickness = 2mm, number of slices = 66. Parallel acceleration factor = 2.5. The acquisition time was approximately 4 minutes and 20 seconds, and diffusion weighted gradients were acquired using 64 B = 1000mm2 directions.

[0046] High-resolution 3D-T1WI data acquisition: 3D fast field echo sequence was used for high-resolution T1 weighted image acquisition. The specific scanning parameters were as follows: TR / TE=7.8 / 3.6ms, layer thickness / layer spacing=1 / 0mm, number of scanning layers=185 layers, FOV=256mm×256mm, Flipangle=8°, and scanning time was 5 minutes and 56 seconds.

[0047] The DTI data were preprocessed using PANDA software. First, the brain extraction tool BET was used to remove non-brain tissues such as the scalp, skull, and cerebrospinal fluid. Next, the Mcflirt toolkit was used to correct the DTI data for head motion. Finally, the eddy_correct toolkit was used to correct the DTI data for eddy current.

[0048] Brain white matter fiber tracking, using the preprocessed PANDA DTI images to calculate each diffusion tensor coefficient, establish the diffusion tensor coefficient distribution of each voxel, and then use the deterministic tracking algorithm to track the brain white matter fibers between brain regions;

[0049] The brain structure network was constructed by combining the white matter fiber tracking results and defining the points of the structure network according to the Power_264 standard template. The edges of the structure network were defined by the fiber connectivity (FN) between brain regions, the average fractional anisotropy (FA) and the average volume of two connected nodes. The specific calculation formula of the edge is as follows: (Wij = FN*FA / volume), the product of the fiber connectivity (FN) between brain regions and the average fractional anisotropy (FA), and normalized by the average volume of two connected nodes.

[0050] Brain structure network analysis: GRETNA software was used to analyze the above multimodal network, mainly including global network attribute analysis (small-world metric, network efficiency, modularity), local node attribute analysis (node ​​degree, node efficiency, node betweenness) and local connection strength analysis. 10,000 nonparametric permutation tests and FDR correction were used to compare the differences between the two groups of parameters, or to calculate the brain structure covariation network: by calculating the Pearson correlation coefficient of the gray matter volume of the voxels contained in any two brain regions in the brain, the structural covariation matrix of each subject was obtained. Finally, the corresponding brain structure covariation network was obtained by binarization.

[0051] IBM SPSS statistical software was used to compare demographic and clinical characteristics, and normality test, Mann-WhitneyU test, chi-square test and one-way ANOVA analysis were performed. Multivoxel pattern analysis (MVPA) based on machine learning was also used to predict neuroimaging and laboratory indicators of WD patients.

[0052] IBM SPSS 16.0 statistical software was used to compare demographic and clinical characteristics. Normality test was performed by Shapiro-Wilk test. Education duration and neuropsychological cognitive scale of the two groups were statistically analyzed by Mann-Whitney U test. Chi-square X2 test was used to compare qualitative variables - gender differences between the two groups. One-way ANOVA was used to determine whether there were differences between the two groups. 10,000 nonparametric permutation tests were performed to determine the intergroup differences in global and local network parameters between the two groups. After determining the intergroup differences, Spearman correlation was used to calculate the correlation between various indicators and clinical scores and gene expression in the WD brain injury patient group. All correlation tests defined age, gender and education level as covariates for regression analysis. Benjamin-Hochberg correction was used for multiple comparison correction of data. In all statistical analyses of network parameters, P values ​​less than 0.05 were considered to be statistically significant. Statistical analysis of the above network attribute parameters was performed using MATLAB software, SPSS software and GRETNA release software.

[0053] Multi-voxel pattern analysis (MVPA) based on machine learning predicts the neuroimaging and laboratory indicators of WD patients, and then explores the analysis method of the association between ATP7B gene mutation data and abnormal expression data of brain function-structure imaging in WD patients from a new perspective of "imaging-gene-Leipzig score". The gene expression module related to brain network changes in WD patients is obtained. Finally, the above imaging data is converted into individual characteristic data that can be quantitatively analyzed through machine learning methods, thereby trying to explain the neural mechanism of WD disease occurrence and development by combining micro-scale and macro-scale. The data analysis process of MVPA is as follows:

[0054] MVPA analysis mainly includes four steps: feature selection, creating examples, classifier training and generalization testing. The specific steps are as follows:

[0055] The multi-factor feature matrix of WD patients was used to train a mathematical model for early diagnosis of WD patients. WD patients were judged according to the Leipzig score, with confirmed WD diagnosis set to 1 and unconfirmed WD diagnosis set to 0. The values ​​were assigned to the model of the support vector machine of supervised machine learning. The 2D feature matrix of WD patients' sample number × feature number was split into a training sample set and a validation sample set. The principal component analysis method was used to reduce the dimension of the features. The trained model was used to predict the validation samples. The classifier was evaluated according to the Roc and Auc curves. The predicted categories (predicted scores) and the actual cognitive impairment were compared and analyzed. The established model for early prediction of WD patients was statistically tested. Non-parametric statistical methods were selected to perform statistical tests on the accuracy, specificity, sensitivity and area under the ROC curve of the model for predicting WD patients. The machine learning model was attached Figure 2 , Figure 3 and Figure 4 shown.

[0056] The method also includes, in the step of analyzing the magnetic resonance brain imaging data, performing a time series analysis on the BOLD-fMRI data to identify the activity pattern of the brain functional network in the resting state, specifically the basal ganglia, thalamus and cerebellum, which are closely related to the neuropsychiatric symptoms of WD patients, calculating the functional connection matrix, analyzing the functional coupling strength between different brain regions, identifying brain region pairs with abnormal functional connections, specifically including abnormal connections between the basal ganglia and the prefrontal cortex; then applying a graph theory analysis method to evaluate the small-world properties of the brain functional network, including network efficiency, modularity and node centrality, to quantify the degree of abnormality of the brain functional network of WD patients, compare the brain functional network characteristics of WD patients and healthy controls, reveal the specific changes of the brain functional network of WD patients, and use the DPARSFA software package to preprocess the data;

[0057] First, the first 10 images of each sequence were removed to eliminate the influence of machine shimming. Then, the images collected at different time points were spatially corrected. The data of subjects with head movement greater than 1.5 mm were no longer included in the next step of analysis. After that, the brain of each subject was standardized to the Montreal Neurological Institute (MNI) space using spatial normalization. After that, the images were smoothed using a Gaussian filter with a half-width of 10 mm. Finally, the linear drift of the image sequence was removed and the images were filtered using a bandpass filter with a frequency of 0.01 to 0.08 Hz.

[0058] Calculate BOLD time series according to ROI First, divide the brain into 264 ROI brain regions according to the Power_264 template of the previous brain atlas. Then, calculate the average BOLD signal time series of each ROI by averaging the signals of all voxels in the region.

[0059] Construction of brain functional network Next, the obtained average time series were correlated with each other to generate a 264 × 264 correlation matrix for each participant. In order to remove the spurious inter-regional correlations in the resulting correlation matrix, the statistical threshold selection criteria were: P < 0.05, FDR-corrected connectivity, and the whole-brain functional network was constructed. The GRETNA software was used to analyze the above multimodal network, mainly including global network property analysis (small-world metric, network efficiency, modularity), local node property analysis (node ​​degree, node efficiency, node betweenness) and local connection strength analysis. 10,000 nonparametric permutation tests and FDR correction were used to compare the differences in the two groups of parameters.

[0060] In this embodiment, specifically, in the magnetic resonance brain imaging data analysis step, fiber tracking is performed on the DTI data to construct a brain white matter structural network. Specific brain regions include the corpus callosum, internal capsule and thalamus. By analyzing the topological properties of the structural network, including path length, clustering coefficient and global efficiency, abnormally connected brain regions in the structural network are identified, specifically including regions associated with cognitive and motor dysfunction in WD patients. By comparing the brain structural network properties of WD patients and healthy controls, specific changes in the brain structural network of WD patients are revealed.

[0061] In this embodiment, specifically, in the statistical analysis and machine learning application steps, the ATP7B gene mutation data is associated with the analysis results of the brain function and structural network to identify brain imaging phenotypic characteristics associated with specific gene mutations, including brain area volume, functional connection strength and structural network properties. A prediction model is constructed using a machine learning algorithm, and the genotype is used as input to predict the patient's brain imaging phenotype. The prediction accuracy and generalization ability of the model are evaluated through cross-validation and external test sets.

[0062] In this embodiment, specifically, in the patient grouping and screening steps, a comprehensive diagnostic model is established by combining the Leipzig score, clinical manifestations, laboratory test results and gene mutation analysis, and the threshold setting of the diagnostic model is optimized through ROC curve analysis and sensitivity-specificity analysis to balance the sensitivity and specificity of the diagnosis.

[0063] In this embodiment, specifically, in the magnetic resonance brain imaging data acquisition step, magnetic resonance brain imaging scans are performed on patients before and after treatment to monitor changes in brain function and structure, specifically the effects of treatment on brain area volume, functional connectivity and structural network properties. By analyzing the changes in brain network properties before and after treatment, the effect of treatment on brain function recovery is evaluated, specifically including the improvement of copper excretion therapy on abnormal brain functional networks. Combined with clinical symptoms and laboratory test results, the treatment effect is comprehensively evaluated.

[0064] In this embodiment, specifically, in the statistical analysis and machine learning application steps, long-term follow-up data of patients, including clinical symptoms, laboratory test results, and brain imaging changes, are collected, multidimensional prognostic evaluation indicators are constructed, and a survival analysis method is used to construct a prognostic prediction model to predict the patient's disease progression rate and survival period. The stability and reliability of the prognostic model are evaluated through external validation and sensitivity analysis.

[0065] In this embodiment, specifically, the gene mutation analysis step includes sequencing common missense mutations, nonsense mutations and splice site mutations in the ATP7B gene, identifying new or known pathogenic gene mutations by high-throughput sequencing technology, and using bioinformatics tools to predict the function of mutations and analyze the effects of these gene mutations on protein function.

[0066] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0067] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting brain damage in Wilson's disease based on multimodal MRI and genetic analysis, characterized in that: The following steps are involved: S1. Comprehensive scans of brain structure and function were performed using functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and three-dimensional T1-weighted imaging (3D-T1WI). Imaging algorithms were used to automatically detect and correct image distortions caused by patient head movement. The specific methods are as follows: Using a set of head motion parameters derived from historical imaging data, which involve common head movement patterns and amplitudes, to set detection thresholds and correction strategies; Monitor the position and movement of the patient's head in real time during the MRI scan, using rapid imaging sequences to capture the head state with extremely short intervals between each sequence to track rapid movements and identify any head movement outside the normal range by comparing it with preset parameters; When abnormal movement is detected, the necessary correction parameters are immediately calculated, including adjusting the angle, focus and other scanning parameters of the MRI scan. The correction operation is based on the deviation between the current head position and the expected scanning position, and the quality of subsequent images is continuously monitored. S2. Collect ATP7B gene mutation data of WD patients, including gene sequence and expression level, and use bioinformatics tools to predict the specific effect of gene mutation on protein function. The specific method is as follows: Collect mutation data on the ATP7B gene, integrate the location, type and known clinical manifestations of the mutation with relevant protein structure data and functional annotations in public databases, build a comprehensive data set, use the data on known gene mutations and their corresponding protein function effects to train a recurrent neural network model, and optimize its predictive ability by learning the complex relationship between the biochemical properties of mutations and functional changes; Predict the functional impact of understudied ATP7B gene mutations and output the protein function changes caused by each mutation; S3. Collect and integrate gene expression data and brain MRI imaging data of WD patients, associate gene expression patterns with functional characteristics of brain area networks, and construct a multi-layer network, one layer of which represents gene expression data and the other layer represents brain MRI data; by analyzing the interaction and connection between the two layers of data, associate gene expression patterns with functional characteristics of brain area networks, and identify key biomarkers associated with the pathological process of WD; S4. Integrate the previously obtained data to construct a prediction model to predict brain structural and functional abnormalities in patients with Wilson disease and conduct in-depth analysis of the potential pathophysiological mechanisms. The specific steps are as follows: The multimodal MRI imaging data, gene mutation data, and clinical data of the Leipzig scoring system collected in S1, S2, and S2 were integrated to construct a dataset, and the gene expression and brain imaging phenotypes across sampling points were analyzed by spatial correlation analysis to extract key imaging omics and genomics features; Recurrent neural networks are used to process and analyze the integrated data. The recurrent neural networks capture the complex nonlinear relationships in the data and identify biomarkers that predict brain damage in Wilson disease.

2. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: The multimodal MRI data obtained in the S1 include functional magnetic resonance imaging (fMRI) data, on which time series analysis is performed to identify the activity patterns of brain functional networks in the resting state. Specific brain regions include the basal ganglia, thalamus and cerebellum, which are closely related to the neuropsychiatric symptoms of WD patients. The functional connectivity matrix is ​​calculated, the functional coupling strength between different brain regions is analyzed, and pairs of brain regions with abnormal functional connectivity are identified, specifically including abnormal connections between the basal ganglia and the prefrontal cortex. Graph theory analysis methods are then applied to evaluate the small-world properties of the brain functional network, including network efficiency, modularity and node centrality, to quantify the degree of abnormality of the brain functional network of WD patients, compare the characteristics of the brain functional network of WD patients and healthy controls, and reveal specific changes in the brain functional network of WD patients.

3. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: Diffusion tensor imaging (DTI) data used in the S1 were used for fiber tracking to construct a brain white matter structural network. Specific brain regions included the corpus callosum, internal capsule, and thalamus. The topological properties of the structural network, including path length, clustering coefficient, and global efficiency, were analyzed to identify abnormally connected brain regions in the structural network, specifically including regions associated with cognitive and motor dysfunction in WD patients. The specific changes in the brain structural network of WD patients were revealed by comparing the brain structural network properties of WD patients with those of healthy controls.

4. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: Using the multimodal MRI data and gene data obtained in S1 and S3, the following analysis and prediction process is performed: Combining the functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) data obtained from S1 and the ATP7B gene mutation data collected in S3, statistical and machine learning techniques are used to identify brain imaging phenotypic characteristics associated with specific gene mutations. A predictive model is constructed using a machine learning algorithm, and the genotype data is used as input to predict brain imaging phenotypes associated with Wilson disease, revealing how gene expression affects changes in brain structure and function.

5. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: A comprehensive diagnostic model was established by combining the MRI data obtained from S1, the gene mutation data in S2, the Leipzig score and other relevant clinical manifestations and laboratory test results obtained from clinical practice. The threshold setting of the diagnostic model was optimized through ROC curve analysis and sensitivity-specificity analysis to balance the sensitivity and specificity of the diagnosis.

6. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: In the step S1, multiple magnetic resonance brain imaging scans are performed on the Wilson disease patient before and after the copper removal treatment to monitor the specific effects of the treatment on the patient's brain function and structure, which specifically includes the following steps: Functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and three-dimensional T1-weighted imaging (3D-T1WI) were performed regularly to capture changes in brain structure and function before and after treatment. The MRI data before and after treatment were analyzed, including the effects of brain area volume, functional connectivity, and structural network properties. The effects of treatment on brain function recovery were evaluated by analyzing the changes in brain network properties before and after treatment, specifically including the improvement of copper excretion therapy on abnormal brain functional networks. The treatment effect was comprehensively evaluated in combination with clinical symptoms and laboratory test results.

7. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: The S4 step further includes the collection and analysis of long-term follow-up data, and the specific steps are as follows: regularly collect clinical symptoms, laboratory test results and multimodal brain imaging data of Wilson disease patients, including clinical symptoms, laboratory test results and brain imaging changes, construct multidimensional prognostic evaluation indicators, use survival analysis methods to construct a prognostic prediction model to predict the disease progression rate and survival period of patients, and evaluate the stability and reliability of the prognostic model through external validation and sensitivity analysis.

8. The method for predicting brain damage in Wilson's disease based on multimodal MRI and gene analysis according to claim 1, characterized in that: The gene mutation analysis in step S2 specifically includes sequencing common missense mutations, nonsense mutations and splice site mutations in the ATP7B gene, identifying new or known pathogenic gene mutations by high-throughput sequencing technology, and using bioinformatics tools to predict the function of mutations and analyze the effects of these gene mutations on protein function.

Images