A mental disorder core symptom multi-scale prediction model construction method and device

Abstract

The present application provides a mental disorder core symptom multi-scale prediction model construction method and device, comprising obtaining a symptom scale and functional magnetic resonance data; constructing a sparse symptom network according to the symptom scale, and calculating an importance index to screen to obtain a core symptom node; generating a whole brain functional connection matrix based on the functional magnetic resonance data and combining a training set to establish a connection group prediction sub-model, to obtain a positive prediction network and a negative prediction network; projecting the prediction network to a preset large-scale brain network and mapping with a human brain spatial gene expression atlas to respectively obtain brain connection representation and molecular annotation representation; and performing multi-scale fusion with the core symptom node as an index to generate a mental disorder core symptom multi-scale prediction model. The prediction model constructed by the present application can not only output each core symptom prediction parameter, but also provide cross-scale explanation features, avoiding the mixing of heterogeneous symptoms in traditional total score models, and improving the prediction accuracy and clinical targeting of mental disorder core symptoms.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of medical image processing and mental symptom network analysis technology, and in particular to a method and apparatus for constructing a multi-scale prediction model of core symptoms of mental disorders. Background Technology

[0002] Many mental disorders exhibit significant symptom heterogeneity and cross-diagnostic phenotype overlap, with schizophrenia being a typical example of a mental disorder.

[0003] Early identification and personalized diagnosis of mental illnesses is one of the important research areas in modern psychiatry, especially in the clinical management of mental illnesses such as depression, schizophrenia, bipolar disorder, and autism spectrum disorder. Constructing an efficient, accurate, and interpretable auxiliary diagnostic system has become a major challenge for both academia and industry.

[0004] Existing methods for identifying mental illness still face several technical bottlenecks. On the one hand, traditional methods often rely on static whole-brain connectivity matrices or single statistical features of local brain regions, making it difficult to capture the dynamic changes in the brain across different time scales. On the other hand, complex high-order interactions between brain regions are often overlooked, and time-domain, frequency-domain, and potential causal relationships have not been effectively modeled. Furthermore, while some studies have introduced deep learning models to improve identification accuracy, their generally black-box architecture lacks sufficient clinical interpretability and transparency, limiting their widespread deployment in real-world medical settings.

[0005] Traditional modeling methods often use the total disease score, subscale score, or a single diagnostic category as the prediction target. While this is convenient for overall description, it can easily mix the phenotypic structures of different symptom items, reducing the model's ability to explain the core symptom mechanisms.

[0006] Symptom network analysis can represent symptom items as nodes and conditional relationships between items as edges, thereby identifying core symptom nodes with high influence in the symptom system. However, standalone symptom network models typically remain at the phenotypic level, making it difficult to further construct predictive models supported by brain connectivity and molecular annotation. Connectome predictive modeling can utilize whole-brain functional connectivity edges from resting-state functional magnetic resonance imaging (fMRI) data to construct symptom prediction models. Spatial gene expression maps can further provide molecular-level annotations for the predictive network. Existing methods often implement the above analytical steps separately, lacking a method to integrate core symptom screening, connectome predictive modeling, and molecular annotation into a unified predictive model.

[0007] In view of this, the embodiments of this specification aim to provide a method and apparatus for constructing a multi-scale prediction model of core symptoms of mental disorders. Summary of the Invention

[0008] In view of the problems of coarsened prediction targets, fragmented cross-scale information, and insufficient model interpretability in existing technologies, the purpose of the embodiments in this specification is to provide a method and apparatus for constructing a multi-scale prediction model of core symptoms of mental disorders.

[0009] The specific technical solutions of the embodiments in this specification are as follows:

[0010] Firstly, embodiments of this specification provide a method for constructing a multi-scale prediction model for core symptoms of mental disorders, including:

[0011] The raw data is acquired and preprocessed. The raw data includes symptom scale data and resting-state functional magnetic resonance imaging data of multiple subjects. The symptom scale data includes scores of multiple symptom items.

[0012] Construct a sparse symptom network based on the symptom score matrix corresponding to the symptom item scores;

[0013] The importance index of each symptom node is calculated based on the sparse symptom network, and the core symptom nodes are obtained by filtering each symptom node according to the importance index and the preset symptom domain.

[0014] Based on the resting-state functional magnetic resonance imaging data and the preset brain region template, a whole-brain functional connectivity matrix is ​​generated for each subject. The elements in the whole-brain functional connectivity matrix represent the functional connectivity strength between two brain regions.

[0015] Using the whole-brain functional connectivity matrix and a pre-established cross-validation training set, a connectivity group prediction sub-model corresponding to each core symptom node is established to obtain the positive prediction network, the negative prediction network, and the configuration parameters.

[0016] The positive prediction network and the negative prediction network are projected onto a preset large-scale brain network to obtain brain connectivity representation, and the spatial distribution of the positive prediction network and the negative prediction network is mapped to the human brain spatial gene expression map to obtain molecular annotation representation.

[0017] Using the core symptom nodes as indexes, the importance indicators, connectome prediction sub-models, brain connectivity representations, and molecular annotation representations are fused to generate the multi-scale prediction model for core symptoms of mental disorders.

[0018] Secondly, embodiments of this specification provide a device for constructing a multi-scale prediction model for core symptoms of mental disorders, comprising:

[0019] The acquisition and preprocessing module is used to acquire raw data and perform preprocessing. The raw data includes symptom scale data and resting-state functional magnetic resonance imaging data of multiple subjects. The symptom scale data includes scores of multiple symptom items.

[0020] A sparse symptom network construction module is used to construct a sparse symptom network based on the symptom score matrix corresponding to the symptom item scores.

[0021] The core symptom node filtering module is used to calculate the importance index of each symptom node based on the sparse symptom network, and to filter each symptom node to obtain core symptom nodes based on the importance index and the preset symptom domain.

[0022] The connectivity matrix generation module is used to generate a whole-brain functional connectivity matrix for each subject based on the resting-state functional magnetic resonance data and a preset brain region template. The elements in the whole-brain functional connectivity matrix represent the functional connectivity strength between two brain regions.

[0023] The prediction network building module is used to build a connection group prediction sub-model corresponding to each core symptom node using the whole brain functional connectivity matrix and a pre-established cross-validation training set, so as to obtain a positive prediction network, a negative prediction network and configuration parameters.

[0024] The multi-scale annotation acquisition module is used to project the positive prediction network and the negative prediction network onto a preset large-scale brain network to obtain brain connectivity representation, and to map the spatial distribution of the positive prediction network and the negative prediction network to the human brain spatial gene expression map to obtain molecular annotation representation.

[0025] The multi-scale prediction model generation module is used to generate the multi-scale prediction model of the core symptoms of mental disorders by using the core symptom nodes as indexes and fusing the importance indicators, the connectome prediction sub-model, the brain connectivity representation, and the molecular annotation representation.

[0026] Thirdly, embodiments of this specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method provided by the above-described technical solution.

[0027] By employing the above technical solution, the embodiments of this specification provide a method and apparatus for constructing a multi-scale prediction model for core symptoms of mental disorders. This model integrates symptom networks, whole-brain functional connectivity, large-scale brain network projections, and spatial gene expression maps into a unified multi-scale prediction model. This allows the model to not only output prediction parameters for each core symptom but also provide cross-scale interpretive features such as symptom co-occurrence patterns, brain functional connectivity, brain connectivity representations, and molecular annotation representations. This ensures robust predictive performance and provides multi-dimensional interpretive data for understanding the neurobiological mechanisms behind core symptoms of mental disorders, enhancing the model's practical value in clinical auxiliary diagnosis and personalized intervention decision-making.

[0028] To make the above and other objects, features and advantages of the embodiments of this specification more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This diagram illustrates the steps of a method for constructing a multi-scale prediction model for core symptoms of mental disorders, as provided in an embodiment of this specification.

[0031] Figure 2 A flowchart illustrating the steps involved in generating the whole-brain functional connectivity matrix for each subject is shown.

[0032] Figure 3 A flowchart illustrating the steps for establishing a connection group prediction sub-model corresponding to each core symptom node is shown.

[0033] Figure 4 The flowchart illustrates the steps involved in mapping to obtain molecular annotation characterization.

[0034] Figure 5 This diagram illustrates the structure of a multi-scale prediction model construction device for core symptoms of mental disorders provided in an embodiment of this specification.

[0035] Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of this specification is shown.

[0036] Explanation of reference numerals in the attached figures:

[0037] 51. Acquisition and Preprocessing Module; 52. Sparse Symptom Network Construction Module; 53. Core Symptom Node Filtering Module; 54. Connection Matrix Generation Module; 55. Prediction Network Building Module; 56. Multi-Scale Annotation Acquisition Module; 57. Multi-Scale Prediction Model Generation Module;

[0038] 602. Computer equipment; 604. Processor; 606. Memory; 608. Drive mechanism; 610. Input / output module; 612. Input device; 614. Output device; 616. Presentation device; 618. Graphical user interface; 620. Network interface; 622. Communication link; 624. Communication bus. Detailed Implementation

[0039] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0040] It should be noted that the terms "first," "second," etc., used in this specification, claims, and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0041] To address the aforementioned issues, this specification provides a method and apparatus for constructing a multi-scale prediction model for core symptoms of mental disorders, which can solve the problems of coarsening of prediction targets, fragmentation of cross-scale information, and insufficient model interpretability in existing technologies. Figure 1 This diagram illustrates the steps of a method for constructing a multi-scale prediction model for core symptoms of mental disorders, as provided in the embodiments of this specification. This specification provides the operational steps of the method described in the embodiments or flowcharts, but based on conventional or non-creative labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or device products, the methods shown in the embodiments or accompanying drawings can be executed sequentially or in parallel. Specifically, as shown in the embodiments or accompanying drawings... Figure 1 As shown, the method may include the following steps:

[0042] S110: Acquire raw data and preprocess it. The raw data includes symptom scale data and resting-state functional magnetic resonance imaging data of multiple subjects. The symptom scale data includes scores of multiple symptom items.

[0043] It should be noted that the mental disorders applicable to this method include, but are not limited to, schizophrenia, bipolar disorder, depressive disorder, anxiety disorder, obsessive-compulsive disorder, trauma and stress-related disorders, as well as diseases or symptom clusters with symptom scales and functional magnetic resonance imaging data.

[0044] Symptom scale data includes multiple symptom items for mental disorders and scores used to characterize the severity of symptoms. For example, in the case of schizophrenia, the symptom scale data can be the Positive and Negative Syndrome Scale (PANSS), which includes 30 symptom items. Each symptom item is typically scored from 1 to 7, and the assessment is conducted by a psychiatrist based on clinical interviews, behavioral observations, and family statements. For other mental disorders, the corresponding symptom scale data can be replaced with standardized symptom scales for the corresponding disease or symptom cluster.

[0045] S120: Construct a sparse symptom network based on the symptom score matrix corresponding to the symptom item scores.

[0046] For example, if the number of subjects included in the model is n and the number of symptom items is m, then the symptom scale data can be represented as an n×m symptom score matrix (denoted as X), where the i-th row of X represents the score of the i-th subject on each symptom item, and the j-th column represents the cross-subject score vector of the j-th symptom item.

[0047] The sparse symptom network is constructed by the conditional correlation between the entries of the symptom score matrix, thereby transforming the symptom entries into symptom nodes in the sparse symptom network.

[0048] S130: Calculate the importance index of each symptom node based on the sparse symptom network, and filter each symptom node to obtain the core symptom node based on the importance index and the preset symptom domain.

[0049] In the embodiments of this specification, the importance index can be calculated using the edge weight matrix of the sparse symptom network. The importance index includes at least one of expected influence value, strength, betweenness centrality, and proximity centrality. In some feasible embodiments, the importance index may preferably be the expected influence value to reflect the total influence of symptom nodes under the balance of positive and negative regulation.

[0050] When screening symptom nodes by combining importance indicators and preset symptom domains, symptom nodes with high influence and key roles can be selected. At the same time, it can avoid the core nodes from being concentrated in a single symptom domain, that is, retain representative core nodes of different symptom domain dimensions, thereby preserving the multidimensional phenotypic structure of different mental disorders preset symptom domains, achieving balanced coverage among symptom domains, which is conducive to improving the robustness and interpretability of model construction.

[0051] It should be noted that, in the embodiments of this specification, the core symptom nodes are determined based on the symptom scale data, the node importance index of the sparse symptom network, and the preset screening conditions. When screening core symptom nodes for each symptom domain based on the importance index, the screening conditions (i.e., the importance index) for each symptom domain can be the same or different; and when the same importance index is used for screening in different symptom domains, the threshold of the importance index can be the same or different; thus, for each symptom domain, the number of core symptom nodes screened can be the same or different, and each symptom domain will have at least one core symptom node screened.

[0052] Taking schizophrenia as an example, its pre-defined symptom domains include positive symptom domains, negative symptom domains, general psychopathology symptom domains, or other pre-defined symptom domains. When screening within each symptom domain corresponding to schizophrenia based on importance indicators (using expected impact value as an example), symptom nodes can be sorted from high to low according to their expected impact value. Then, one or more symptom nodes with the highest expected impact value within each symptom domain are selected as core symptom nodes. For example, the core symptom node selected in the positive symptom domain might be conceptual disorder, the core symptom node in the negative symptom domain might be emotional blunting, and the core symptom node in the general psychopathology symptom domain might be abnormal thought content. Under other sample data, other symptom domains, and other screening rules, the core symptom nodes can change accordingly.

[0053] S140: Based on the resting-state functional magnetic resonance imaging data and the preset brain region template, generate a whole-brain functional connectivity matrix for each subject, wherein the elements in the whole-brain functional connectivity matrix represent the functional connectivity strength between two brain regions.

[0054] S150: Using the whole-brain functional connectivity matrix and the pre-established cross-validation training set, establish a connectivity group prediction sub-model corresponding to each core symptom node to obtain the positive prediction network, the negative prediction network, and the configuration parameters.

[0055] S160: Project the positive prediction network and the negative prediction network onto a preset large-scale brain network to obtain brain connectivity representation, and map the spatial distribution of the positive prediction network and the negative prediction network to the human brain spatial gene expression map to obtain molecular annotation representation.

[0056] In the embodiments of this specification, the pre-defined large-scale brain network is determined by a pre-selected brain region template, functional network atlas, or brain region-to-functional system label table. This network is used to characterize the distribution of symptom prediction-related connections in brain functional systems such as the somatic motor network, subcortical network, default mode network, and frontoparietal network. The brain connectivity characterization includes at least one of node degree, edge count, edge weight strength, network strength, and centrality indices. These indices characterize the spatial distribution of symptom prediction-related functional connections in different brain functional systems, indicating which brain functional systems the prediction information for a particular core symptom is primarily located in, serving as one of the explanatory information output by the subsequent model.

[0057] Human brain spatial gene expression atlases can be derived from publicly available human brain gene expression databases, including gene expression intensities corresponding to multiple brain regions or spatial sampling points. Therefore, by spatially correlated or overlapping the spatial distribution of the prediction model with the human brain spatial gene expression atlas, a set of genes related to the spatial distribution of the prediction network can be obtained, ultimately leading to molecular annotation characterization.

[0058] S170: Using the core symptom items as an index, the importance index, the connectome prediction sub-model, the brain connectivity representation, and the molecular annotation representation are integrated to generate the multi-scale prediction model for the core symptoms of mental disorders.

[0059] The constructed multi-scale prediction model includes an input layer, a symptom network layer, a core node layer, a connectome prediction layer, a brain connectivity representation and molecular annotation representation layer (multi-scale annotation layer), and an output layer. The output layer includes prediction parameters for each core symptom node (including selection criteria for candidate functional connectivity edges), the positive and negative prediction networks (including at least the edge sets, edge weight parameters, network strength features, and calculation strategies of the positive / negative prediction networks), model coefficients (at least one of intercept, slope, etc.), brain connectivity representation, molecular annotation representation, and model interpretation information. The model interpretation information further includes at least one of the interpretable fields such as the symptom domain to which each core symptom node belongs, brain network contribution, network similarity parameters, enrichment analysis results, and model evaluation metrics.

[0060] The method for constructing a multi-scale prediction model for core symptoms of mental disorders provided in this specification converts symptom scales into a sparse symptom network and filters out core symptom nodes as modeling targets. This avoids the mixing of heterogeneous symptoms in traditional total score models, thus improving the accuracy and clinical relevance of predictions for core symptoms of mental disorders. Furthermore, by integrating symptom networks, whole-brain functional connectivity, large-scale brain network projections, and spatial gene expression maps into a unified multi-scale prediction model, the model not only outputs prediction parameters for each core symptom but also provides cross-scale interpretive features such as symptom co-occurrence patterns, brain functional connectivity, brain connectivity representations, and molecular annotation representations. This ensures robust predictive performance and provides multi-dimensional interpretive data for understanding the neurobiological mechanisms behind core symptoms of mental disorders, enhancing the model's practical value in clinical auxiliary diagnosis and personalized intervention decision-making.

[0061] Furthermore, the model construction method provided in the embodiments of this specification can be extended according to different scales, brain region templates, threshold strategies, and spatial gene expression maps, etc., and is applicable to multi-center scientific research data platforms and stratified research on the mechanisms of mental illnesses, with a wide range of applications.

[0062] Specifically, in the embodiments of this specification, the preprocessing of the original data in step S110 may include:

[0063] Missing values ​​were processed for the symptom scale data and the covariate data, and image quality and head movement quality were controlled for the functional magnetic resonance imaging data.

[0064] This effectively improves the completeness and accuracy of the input data, thereby ensuring the accuracy of subsequent sparse symptom network construction, functional connectivity calculation, and predictive model training.

[0065] Furthermore, in this embodiment of the specification, before obtaining the symptom score matrix corresponding to the symptom item scores in step S120, the method further includes:

[0066] Nonparametric normalization is applied to the scores of symptom items that are not normally distributed, i.e., the scores of each symptom item are standardized. This reduces the interference of skewed distribution on partial correlation estimation and improves the accuracy and reliability of sparse symptom network construction.

[0067] Furthermore, a sparse symptom network is constructed based on the symptom scoring matrix, specifically as follows:

[0068] Using Gaussian graph models and sparsification estimation methods, the conditional correlations between symptom items in the symptom score matrix are calculated;

[0069] The sparse symptom network is constructed by using the symptom items as symptom nodes and the conditional correlations as weighted edges.

[0070] In other words, the weighted edges of a sparse symptom network are the partial correlations between two symptom entries after controlling for other symptom entries, and are used to determine the core symptom nodes in subsequent steps.

[0071] By constructing a sparse symptom network using the above methods, spurious associations can be effectively eliminated. This results in a sparse symptom network with sparse edge sets and interpretable weights, which helps reduce network redundancy and improve the accuracy of estimating the relationships between symptom items.

[0072] like Figure 2 As shown in the embodiments of this specification, step S140: generating a whole-brain functional connectivity matrix for each subject based on the resting-state functional magnetic resonance imaging data and a preset brain region template may include the following steps:

[0073] S210: Divide the resting-state functional magnetic resonance imaging data into several brain regions based on the preset brain region template.

[0074] S220: Extract the time series of mean blood oxygenation level dependent signals in each brain region from the resting-state functional magnetic resonance data of each subject;

[0075] S230: Calculate the correlation coefficient of the time series of the mean blood oxygenation level dependent signal between any two brain regions and perform Fisher's z-transform;

[0076] S240: The transformed correlation coefficients are used as functional connectivity strengths to generate whole-brain functional connectivity matrices for each subject.

[0077] The rows and columns of the generated whole-brain functional connectivity matrix correspond to two brain regions of the subject, and the matrix elements correspond to the functional connectivity strength between the two brain regions of the subject.

[0078] In the embodiments of this specification, a whole-brain functional connectivity matrix is ​​constructed by dividing brain regions, extracting BOLD (Blood Oxygen Level-Dependent) time series data, calculating the correlation coefficient between brain regions, and performing Fisher z-transform. This matrix can reduce the dimensionality of the original magnetic resonance imaging data into standardized, symmetrical, and approximately normally distributed brain region connectivity strength features, which not only preserves the core information of functional coupling between brain regions but also provides stable and comparable individualized input features for subsequent connectivity prediction models.

[0079] Furthermore, such as Figure 3 As shown in the embodiments of this specification, step S150: using the whole-brain functional connectivity matrix and the pre-established cross-validation training set, to establish a connectivity group prediction sub-model corresponding to each core symptom node, may specifically include the following steps:

[0080] S310: Calculate the correlation coefficient between each functional connection edge in the whole-brain functional connectivity matrix and the scores of each core symptom node in the cross-validation training set.

[0081] In the embodiments of this specification, the training data in the cross-validation training set includes scores from each subject for each core symptom node. For each core symptom node, the correlation between each functional connectivity edge and the score of that core symptom node is calculated. This allows the scores to be used as a supervisory signal to extract a subset of connections specific to symptoms from the whole-brain functional connectivity matrix.

[0082] In some preferred embodiments, the training data in the cross-validation training set also includes covariate data. These covariate data include at least one of the following: subject's age, gender, years of education, head movement parameters, and data collection sites. Before calculating the correlation coefficients between each functional connectivity edge and each core symptom node score, the covariate data can be processed using at least one of the following methods: missing value handling, partial correlation control, residual processing, and batch correction. Subsequently, the processed covariate data is used to perform covariate regression on the scores of each functional connectivity edge and / or each core symptom node to reduce the influence of non-symptom factors on functional connectivity edge selection / symptom scores, thereby improving the accuracy of subsequent predictive model construction.

[0083] S320: Select the functional connection edges whose correlation coefficients meet the preset screening conditions as candidate functional connection edges.

[0084] The preset screening conditions can be the significance threshold, correlation strength threshold, or stability threshold of the candidate functional connection edges. From multiple candidate conditions, the threshold that best enables the model to predict the best or most stable performance is selected based on the model evaluation index in the cross-validation training set, the cross-validation evaluation index, or the model stability index, and then solidified as the final preset screening conditions. This reduces overfitting and improves the repeatability of the connection group prediction sub-model.

[0085] S330: Connect the candidate functions with positive correlation coefficients to form a positive prediction network for the core symptom node, and connect the candidate functions with negative correlation coefficients to form a negative prediction network for the core symptom node.

[0086] S340: Calculate the network strength characteristics based on the positive prediction network and the negative prediction network.

[0087] Specifically, one or a combination of processes such as summation, weighted summation, or regularization can be applied to the edge weights in the positive and negative prediction networks to obtain network strength features; network strength features can include positive network strength, negative network strength, and combinations thereof.

[0088] Taking the summation method as an example, for the i-th subject, the positive network strength feature can be represented as the sum of the functional connection strengths of all functional connections in the positive prediction network edge set for that subject, and the negative network strength feature can be represented as the sum of the functional connection strengths of all functional connections in the negative prediction network edge set for that subject. In the weighted summation method, when calculating the network strength feature, each functional connection strength is first multiplied by the corresponding edge weight parameter and then summed. In the regularization method, the network strength feature is obtained by linearly or nonlinearly combining the candidate functional connection edges using a regularization model.

[0089] S350: Using the network strength features and the core symptom node scores, the constructed initial prediction model is iteratively trained to obtain the connection group prediction sub-model.

[0090] An initial prediction model can be constructed using linear models, regularized regression models, support vector regression models, or other interpretable prediction algorithms; and trained using network strength features as input (i.e., independent variables) and core symptom node scores as output (i.e., dependent variables).

[0091] The output of the connection group prediction sub-model can be represented as the predicted value of the core symptom node score, which is equal to the intercept term plus the model coefficients corresponding to the positive network strength, negative network strength and their combination term. Iterative training and evaluation are performed by using the error or correlation index between it and the corresponding core symptom node scores (i.e., the true values) in the cross-training set.

[0092] The embodiments in this specification employ, as follows: Figure 3 The method described utilizes a cross-validation training set to filter functional connectivity edges and generate model parameters, thereby constructing and training the predictive model. This approach helps reduce the risk of information leakage and overfitting, and improves the reproducibility of the model construction process. Ultimately, the resulting predictive sub-model for each core symptom node outputs individualized predicted symptom scores and simultaneously provides configuration items such as the set of positive / negative edges related to that symptom and edge weight parameters. This provides direct and verifiable labeled data for understanding the neural connections of specific mental disorders.

[0093] In the embodiments of this specification, the configuration parameters obtained by constructing the connection group prediction sub-model may include the screening threshold of candidate functional connection edges, the edge set of the positive prediction network / negative prediction network, the edge weight parameters of the positive prediction network / negative prediction network, and the model coefficients.

[0094] Here, the edge set is the set of brain region connectivity pairs that enter the positive or negative prediction network, such as ordered or unordered connectivity pairs consisting of brain regions u and v. Edge weight parameters include the functional connectivity strength of the connectivity pair in each subject, the correlation coefficients used for feature selection in the training set and their directions, or the edge weights obtained during training when using a weighted model. Model coefficients include the intercept, slope, regression coefficients, regularization coefficients, or support vector regression parameters that map the positive network strength, negative network strength, or a combination of both to the core symptom node scores.

[0095] It should be noted that, in the embodiments of this specification, the information such as the subject's age, gender, and years of education included in the symptom scale and covariate data are all information that has been authorized and agreed upon by the user in the past. Furthermore, the acquisition, storage, use, and processing of the aforementioned symptom scale and covariate data in the technical solutions described in the embodiments of this specification all comply with the relevant provisions of national laws and regulations.

[0096] Furthermore, in the embodiments of this specification, molecular annotation characterization is obtained by mapping the spatial distribution of the prediction model to the spatial gene expression map of the human brain.

[0097] Furthermore, in the embodiments of this specification, the preset large-scale brain network may include at least one of the following: somatic motor network, subcortical network, default mode network, frontoparietal network, cingulate-insula network, attention network, auditory network, visual network, salience network, memory retrieval network, and cerebellar network.

[0098] Furthermore, in step S160, projecting the positive prediction network and the negative prediction network onto a preset large-scale brain network to obtain brain connectivity representation specifically involves: according to the large-scale brain network label to which each brain region belongs in the preset large-scale brain network, classifying the candidate functional connection edges in the positive prediction network and the negative prediction network into the corresponding brain system or inter-system connection, and statistically analyzing at least one of the following indicators for each brain system: node degree, number of edges, edge weight strength, network strength, and centrality index, to obtain brain connectivity representation.

[0099] Brain-system-level connectivity representations can intuitively reflect large-scale abnormal patterns in brain networks corresponding to different core symptoms, providing systemic representational evidence for the neural mechanisms underlying the heterogeneity of mental disorder symptoms.

[0100] Furthermore, step S160, mapping the spatial distributions of the positive and negative prediction networks to the human brain spatial gene expression map to obtain molecular annotation characterization, may further include:

[0101] S410: Process the spatial gene expression map of the human brain to form a brain region-gene expression matrix corresponding to the preset brain region template.

[0102] The processing of the human brain spatial gene expression map may include at least one of the following: brain region registration, probe or gene symbol conversion, expression value normalization, and brain region scale summarization.

[0103] S420: Map the positive prediction network and the negative prediction network to the brain region-gene expression matrix respectively to obtain a set of genes related to the spatial distribution of the prediction network.

[0104] Specifically, the node degree, edge weight strength, spatial weight distribution, etc. of the positive prediction network / negative prediction network can be mapped to the human brain-gene expression matrix, wherein the mapping process can include at least one of spatial correlation, overlapping mapping, and ranking enrichment.

[0105] S430: Perform ontology enrichment analysis, pathway enrichment analysis, or a combination thereof on the gene set to generate the molecular annotation characterization.

[0106] The resulting molecular annotations are used to reflect normalized human brain expression patterns that match the spatial distribution of the predicted network. This allows for the revelation at the transcriptomic level of which functional genes or signaling pathways are enriched in the brain regions involved in the predicted network, providing molecular evidence for the biological basis of symptoms. Furthermore, the molecular annotations, along with connectome predictive sub-models and brain connectivity representations, will be incorporated into a multi-scale predictive model, enhancing its cross-scale interpretability and facilitating the identification of potential biomarkers and therapeutic targets. This will improve the scientific depth and clinical translational value of the core symptom prediction model for mental disorders.

[0107] In this embodiment of the specification, step S170 involves outputting the prediction parameters, positive prediction network, negative prediction network, brain connectivity representation, molecular annotation representation, and model interpretation information corresponding to each core symptom node. The prediction parameters include at least one of the following: a screening threshold for candidate functional connectivity edges, the edge set of the positive or negative prediction network, edge weight parameters, network strength feature calculation method, model coefficients, and intercept. The model interpretation information includes at least one of the following: the symptom domain to which the core symptom node belongs, brain network contribution, network similarity parameters, enrichment analysis results, and model evaluation metrics.

[0108] As shown in Table 1, the multi-scale prediction model constructed in the embodiments of this specification includes an input layer (not shown), a symptom network layer, a core node layer, a connectome prediction layer, a brain connectivity representation and molecular annotation representation layer (multi-scale annotation layer), and an output layer.

[0109] Table 1

[0110]

[0111] The output of the output layer includes prediction parameters for each core symptom node, positive prediction network, negative prediction network, brain connectivity representation, molecular annotation representation, and model interpretation information.

[0112] The prediction parameters include at least the following interpretable fields: First, a screening threshold, used to record the significance, correlation strength, or stability thresholds used when candidate functional connection edges enter the positive or negative prediction network; Second, an edge set, used to record the brain region number, brain region name, the large-scale brain network to which the selected functional connection edge belongs, and the positive or negative network label; Third, edge weight parameters, used to record the functional connection strength, feature screening correlation coefficient, correlation direction, and weights in the training model of the selected functional connection edge; Fourth, the network strength feature calculation method, used to record summation, weighted summation, regularized combination, or other feature aggregation methods; Fifth, model coefficients, used to record the intercept, slope, regression coefficient, regularization parameter, or support vector regression parameter.

[0113] The model interpretation information includes at least one of the following: the symptom domain to which the core symptom node belongs, node importance index, brain network contribution, distribution of the predicted network in the large-scale brain network, gene set and enrichment terms related to the spatial distribution of the predicted network, model evaluation index, and prediction performance index in cross-validation or independent validation.

[0114] Finally, the method for constructing a multi-scale prediction model of core symptoms of mental disorders provided in the embodiments of this specification can not only output parameters for reproducing experiments and deploying predictions, but also output explanatory information for explaining the shared and specificity of different core symptoms at the levels of symptom topology, brain connectivity and molecular annotation, providing structured, multi-scale data support for subsequent clinical auxiliary diagnosis and mechanism research.

[0115] In some preferred embodiments, after step S150: establishing the connection group prediction sub-model corresponding to each core symptom node, the method further includes:

[0116] Node degree vectors are generated based on the positive prediction network and the negative prediction network;

[0117] The network similarity parameters between the prediction sub-models of each connection group are calculated based on the node degree vector. The network similarity parameters include at least one of cosine similarity, correlation coefficient, and distance metric.

[0118] Therefore, in step S170, the multi-scale prediction model for the core symptoms of mental disorders is generated, which is further described as follows:

[0119] By integrating the network similarity parameters with the importance index, the predicted network parameters, the brain connectivity representation, and the molecular annotation representation, a multi-scale prediction model for the core symptoms of mental disorders is generated, so as to enrich the explanatory information of the model.

[0120] Based on the above-described method for constructing a multi-scale prediction model of core symptoms of mental disorders, this specification also provides a corresponding device for constructing such a model. The device may include a system (including a distributed system), software (application), module, component, server, client, etc., using the method described in this specification, combined with necessary hardware implementation. Based on the same innovative concept, the devices in one or more embodiments provided in this specification are as described in the following embodiments. Since the implementation schemes and methods for solving the problem are similar, the implementation of specific devices in this specification can refer to the implementation of the aforementioned method, and repeated details will not be elaborated further. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0121] like Figure 5 As shown, the device for constructing a multi-scale prediction model for core symptoms of mental disorders includes:

[0122] The acquisition and preprocessing module 51 is used to acquire raw data and perform preprocessing. The raw data includes symptom scale data and resting-state functional magnetic resonance imaging data of multiple subjects. The symptom scale data includes scores of multiple symptom items.

[0123] The sparse symptom network construction module 52 is used to construct a sparse symptom network based on the symptom score matrix corresponding to the symptom item scores.

[0124] The core symptom node screening module 53 is used to calculate the importance index of each symptom node according to the sparse symptom network, and to screen each symptom node according to the importance index and the preset symptom domain to obtain core symptom nodes.

[0125] The connectivity matrix generation module 54 is used to generate a whole-brain functional connectivity matrix for each subject based on the resting-state functional magnetic resonance data and a preset brain region template. The elements in the whole-brain functional connectivity matrix represent the functional connectivity strength between two brain regions.

[0126] The prediction network building module 55 is used to build a connection group prediction sub-model corresponding to each core symptom node using the whole brain functional connectivity matrix and a pre-established cross-validation training set, so as to obtain a positive prediction network, a negative prediction network and configuration parameters.

[0127] The multi-scale annotation acquisition module 56 is used to project the positive prediction network and the negative prediction network onto a preset large-scale brain network to obtain brain connectivity representation, and to map the spatial distribution of the positive prediction network and the negative prediction network to the human brain spatial gene expression map to obtain molecular annotation representation.

[0128] The multi-scale prediction model generation module 57 is used to generate the multi-scale prediction model of the core symptoms of mental disorders by using the core symptom nodes as indexes and fusing the importance indicators, the connectome prediction sub-model, the brain connectivity representation, and the molecular annotation representation.

[0129] The beneficial effects obtained by the apparatus provided in the embodiments of this specification are consistent with the beneficial effects obtained by the methods described above, and will not be repeated here.

[0130] like Figure 6 The illustration shows a computer device provided in an embodiment of this specification. The multi-scale prediction model construction apparatus for core symptoms of mental disorders in this specification can be the computer device in this embodiment, executing the methods described above. The computer device 602 may include one or more processors 604, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. The computer device 602 may also include any memory 606 for storing information of any kind, such as code, settings, data, etc. Non-limitingly, for example, memory 606 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory can represent a fixed or removable component of the computer device 602. In one case, when processor 604 executes associated instructions stored in any memory or combination of memories, the computer device 602 can perform any operation of the associated instructions. The computer device 602 also includes one or more drive mechanisms 608 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.

[0131] Computer device 602 may also include an input / output module 610 (I / O) for receiving various inputs (via input device 612) and providing various outputs (via output device 614). A specific output mechanism may include a presentation device 616 and an associated graphical user interface (GUI) 618. In other embodiments, the input / output module 610 (I / O), input device 612, and output device 614 may be omitted, and the device may function solely as a computer device within a network. Computer device 602 may also include one or more network interfaces 620 for exchanging data with other devices via one or more communication links 622. One or more communication buses 624 couple the components described above together.

[0132] Communication link 622 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 622 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0133] Corresponding to, for example Figures 1 to 4 In addition to the method shown, embodiments of this specification also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described method.

[0134] This specification also provides computer-readable instructions, wherein when a processor executes the instructions, the program therein causes the processor to perform the following... Figures 1 to 4 The method shown.

[0135] This specification also provides a computer program product, including at least one instruction or at least one program segment, wherein the at least one instruction or the at least one program segment is loaded and executed by a processor to achieve the following: Figures 1 to 4 The method shown.

[0136] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.

[0137] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0138] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.

[0139] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0140] In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.

[0141] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described in this specification, depending on actual needs.

[0142] Furthermore, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0143] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0144] This specification uses specific embodiments to illustrate the principles and implementation methods of this specification. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this specification. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this specification. Therefore, the content of this specification should not be construed as a limitation of this specification.

Claims

1. A method for constructing a multi-scale prediction model for core symptoms of mental disorders, characterized in that, include: The raw data is acquired and preprocessed. The raw data includes symptom scale data and resting-state functional magnetic resonance imaging data of multiple subjects. The symptom scale data includes scores of multiple symptom items. Construct a sparse symptom network based on the symptom score matrix corresponding to the symptom item scores; The importance index of each symptom node is calculated based on the sparse symptom network, and the core symptom nodes are obtained by filtering each symptom node according to the importance index and the preset symptom domain. Based on the resting-state functional magnetic resonance imaging data and the preset brain region template, a whole-brain functional connectivity matrix is ​​generated for each subject. The elements in the whole-brain functional connectivity matrix represent the functional connectivity strength between two brain regions. Using the whole-brain functional connectivity matrix and a pre-established cross-validation training set, a connectivity group prediction sub-model corresponding to each core symptom node is established to obtain the positive prediction network, the negative prediction network, and the configuration parameters. The positive prediction network and the negative prediction network are projected onto a preset large-scale brain network to obtain brain connectivity representation, and the spatial distribution of the positive prediction network and the negative prediction network is mapped to the human brain spatial gene expression map to obtain molecular annotation representation. Using the core symptom nodes as indexes, the importance indicators, connectome prediction sub-models, brain connectivity representations, and molecular annotation representations are fused to generate the multi-scale prediction model for core symptoms of mental disorders.

2. The method according to claim 1, characterized in that, Based on the resting-state functional magnetic resonance imaging (fMRI) data and preset brain region templates, a whole-brain functional connectivity matrix is ​​generated for each subject, further including: Based on the preset brain region template, the resting-state functional magnetic resonance imaging data is divided into several brain regions; For each subject's resting-state functional magnetic resonance imaging (fMRI) data, the mean blood oxygen level dependent signal time series in each brain region was extracted; Calculate the correlation coefficient of the time series of the mean blood oxygen level dependent signal between any two brain regions and perform Fisherz transform; The transformed correlation coefficients were used as the functional connectivity strength to generate the whole-brain functional connectivity matrix for each subject.

3. The method according to claim 2, characterized in that, Using the whole-brain functional connectivity matrix and a pre-established cross-validation training set, a connectivity group prediction sub-model corresponding to each core symptom node is constructed, including: Calculate the correlation coefficient between each functional connection edge in the whole-brain functional connectivity matrix and the scores of each core symptom node in the cross-validation training set; Select functional connection edges whose correlation coefficients meet the preset screening conditions as candidate functional connection edges; The candidate functions with positive correlation coefficients are connected to form a positive prediction network for the core symptom node, and the candidate functions with negative correlation coefficients are connected to form a negative prediction network for the core symptom node. Calculate the network strength characteristics based on the positive prediction network and the negative prediction network; The initial prediction model is iteratively trained using the network strength features and the core symptom node scores to obtain the connection group prediction sub-model.

4. The method according to claim 3, characterized in that, The brain connectivity representation includes one or more combinations of node degree, edge data, edge weight strength, network strength, and centrality index of the positive prediction network and the negative prediction network on the preset large-scale brain network. The pre-defined large-scale brain network includes at least one of the following: somatic motor network, subcortical network, default mode network, frontoparietal network, cingulate-insula network, attention network, auditory network, visual network, salience network, memory retrieval network, and cerebellar network.

5. The method according to claim 4, characterized in that, Molecular annotation characterization is obtained by mapping the spatial distributions of the positive and negative prediction networks to the human brain spatial gene expression map, including: The spatial gene expression map of the human brain is processed to form a brain region-gene expression matrix corresponding to the preset brain region template; The positive prediction network and the negative prediction network are respectively mapped to the brain region-gene expression matrix to obtain a set of genes related to the spatial distribution of the prediction network; The molecular annotation characterization is generated by performing ontology enrichment analysis, pathway enrichment analysis, or a combination thereof on the gene set.

6. The method according to claim 3, characterized in that, After establishing the connection group prediction sub-model corresponding to each core symptom node, the method further includes: Node degree vectors are generated based on the positive prediction network and the negative prediction network; The network similarity parameters between the prediction sub-models of each connection group are calculated based on the node degree vector. The network similarity parameters include at least one of cosine similarity, correlation coefficient, and distance metric. By integrating the importance indicators, prediction networks, brain connectivity representations, and molecular annotation representations, a multi-scale prediction model for the core symptoms of mental disorders is generated, further comprising: By integrating the network similarity parameters with the importance index, the predicted network parameters, the brain connectivity representation, and the molecular annotation representation, a multi-scale prediction model for the core symptoms of mental disorders is generated.

7. The method according to claim 1, characterized in that, Before obtaining the symptom score matrix corresponding to the symptom item scores, the method further includes: Nonparametric normalization transformation was performed on the non-normally distributed symptom item scores; Constructing a sparse symptom network based on the symptom score matrix includes: Using Gaussian graph models and sparsification estimation methods, the conditional correlations between symptom items in the symptom score matrix are calculated; The sparse symptom network is constructed by using the symptom items as symptom nodes and the conditional correlations as weighted edges.

8. The method according to claim 1, characterized in that, The importance indicators include at least one of the following: expected influence value, strength, betweenness centrality, and proximity centrality; Based on the aforementioned importance indicators and preset symptom domains, core symptom nodes are obtained by filtering each symptom node, including: Within each symptom domain, each symptom node is sorted according to the aforementioned importance index, and one or more nodes in each symptom domain that meet the preset sorting conditions are identified as core symptom nodes.

9. A device for constructing a multi-scale prediction model for core symptoms of mental disorders, characterized in that, include: The acquisition and preprocessing module is used to acquire raw data and perform preprocessing. The raw data includes symptom scale data and resting-state functional magnetic resonance imaging data of multiple subjects. The symptom scale data includes scores of multiple symptom items. A sparse symptom network construction module is used to construct a sparse symptom network based on the symptom score matrix corresponding to the symptom item scores. The core symptom node filtering module is used to calculate the importance index of each symptom node based on the sparse symptom network, and to filter each symptom node to obtain core symptom nodes based on the importance index and the preset symptom domain. The connectivity matrix generation module is used to generate a whole-brain functional connectivity matrix for each subject based on the resting-state functional magnetic resonance data and a preset brain region template. The elements in the whole-brain functional connectivity matrix represent the functional connectivity strength between two brain regions. The prediction network building module is used to build a connection group prediction sub-model corresponding to each core symptom node using the whole brain functional connectivity matrix and a pre-established cross-validation training set, so as to obtain a positive prediction network, a negative prediction network and configuration parameters. The multi-scale annotation acquisition module is used to project the positive prediction network and the negative prediction network onto a preset large-scale brain network to obtain brain connectivity representation, and to map the spatial distribution of the positive prediction network and the negative prediction network to the human brain spatial gene expression map to obtain molecular annotation representation. The multi-scale prediction model generation module is used to generate the multi-scale prediction model of the core symptoms of mental disorders by using the core symptom nodes as indexes and fusing the importance indicators, the connectome prediction sub-model, the brain connectivity representation, and the molecular annotation representation.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 8.

Images