Mental health auxiliary diagnosis and intervention system and method based on multi-source data fusion

Abstract

The invention relates to the technical field of artificial intelligence and psychological diagnosis, and particularly discloses a psychological health auxiliary diagnosis and intervention system and method based on multi-source data fusion, which ensure authority based on SCID-V and DSM-5 standards; diagnosis terms are automatically converted into a questioning mode suitable for the identity of an evaluated person, and the compliance and authenticity are improved; in combination with a subjective scale, an objective cognitive test and physiological index parameters, the diagnosis scientificity and accuracy are improved; outputting a disease possibility and an interpretable basis based on a machine learning / deep learning model; a graphical diagnosis report is generated and is convenient to understand; longitudinal effect evaluation after multiple evaluation and training is supported, and the method is suitable for scientific research and clinical tracking; localized deployment and privacy protection are supported, and medical data compliance is met.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and psychological diagnosis technology, specifically to a mental health auxiliary diagnosis and intervention system based on multi-source data fusion.

[0002] The present invention also relates to a method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion. Background Technology

[0003] Currently, the diagnosis of mental illnesses and psychological disorders, such as depression, anxiety disorders, attention deficit hyperactivity disorder, and bipolar disorder, mainly relies on interviews and assessments conducted by psychiatrists or psychotherapists.

[0004] The existing diagnostic process has the following problems: 1. High subjectivity: Traditional clinical assessment relies on the doctor's subjective judgment, which leads to inconsistencies in standards among different doctors.

[0005] 2. Poor compliance: Especially among children and adolescents, they often have difficulty understanding medical questions in standardized consultations such as SCID-V, leading to inaccurate answers or missing information.

[0006] 3. Single data source: Currently, diagnosis relies heavily on subjective scales or clinical conversations, lacking objective quantitative tests of cognitive function, which easily leads to missed diagnoses or misdiagnoses.

[0007] 4. Lack of continuous follow-up: Clinical assessments are usually one-off results, making it difficult to conduct longitudinal comparisons and track rehabilitation effects.

[0008] Therefore, there is an urgent need for an AI-assisted diagnostic system that integrates subjective and objective data, has multimodal analysis capabilities, supports child-friendly questioning, and allows for long-term tracking, in order to improve the scientific rigor, efficiency, and scalability of diagnosis. Summary of the Invention

[0009] The purpose of this invention is to address the aforementioned problems by providing a mental health auxiliary diagnosis and intervention system and method based on multi-source data fusion. By integrating the results of consultation dialogues, psychological scales, and cognitive tests, and using artificial intelligence models for comprehensive analysis, this invention outputs structured diagnostic reports and intervention suggestions, thereby solving the problems of strong subjectivity, single data source, and difficulty in tracking in existing technologies.

[0010] The technical solution adopted in this invention is as follows: A mental health auxiliary diagnosis and intervention system based on multi-source data fusion, the system comprising: The data acquisition module is used to collect medical history data, scale and cognitive data, and physiological indicator data of the evaluated subjects. The data processing module is used to preprocess the data collected by the data acquisition module and extract corresponding feature vectors respectively. The feature fusion module is used to fuse the feature vectors to obtain a comprehensive mental health feature vector; The assessment module is used to input the comprehensive mental health feature vector into the mental health assessment model to obtain the assessment results; The report generation module provides intervention recommendations and visualizes the obtained assessment results.

[0011] The second technical solution adopted in this invention is a method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion, employing a system for auxiliary diagnosis and intervention of mental health based on multi-source data fusion. The method specifically includes: Data is collected, including the patient's medical history, scales, cognitive data, physiological indicators, and previous examination reports. Data processing involves preprocessing the data and extracting the feature vectors corresponding to the data. Feature fusion involves inputting the feature vector into a multimodal fusion model to obtain a comprehensive mental health feature vector. The comprehensive mental health feature vector is input into a mental health assessment model to obtain the assessment results. Report generation: Based on the assessment results, an assessment report and corresponding intervention recommendations are generated.

[0012] Furthermore, the consultation data is acquired based on the standardized consultation system SCID-V or DSM-5, and the acquired consultation data is transformed into a user-friendly expression based on natural language processing.

[0013] Furthermore, the natural language processing includes sentiment dictionary analysis, keyword extraction, and semantic dependency analysis.

[0014] Furthermore, the scales and cognitive data include psychological scales and objective cognitive test scales for the subjects being assessed; The physiological indicator data includes the physiological examination data of the subject being evaluated.

[0015] The previous examination reports include the examination reports that the subject underwent at medical institutions in the past.

[0016] Furthermore, the data processing specifically includes: Semantic features, keyword features, and emotional features were extracted from the consultation data, scale and cognitive data, physiological indicator data, and previous examination reports.

[0017] The text content of the previous inspection reports is extracted using AI image recognition.

[0018] Furthermore, the assessment analysis specifically includes obtaining the probability of mental illness occurring in the assessed individual and constructing the following discriminant: (1) In the formula, SDS is the comprehensive discrimination score.

[0019] Furthermore, the comprehensive discrimination score SDS specifically includes: Standardized calculation method for consultation content: DSM Score = Symptom duration weight score × Number of matching items weight score / Total number of items weight score; Standardized calculation method for scales: Scale Score = Scale score / Total scale score; SDS = 0.8 × DSM Score + 0.2 × Scale Comprehensive Score.

[0020] Furthermore, the evaluation analysis also includes using hierarchical Bayesian inference of the posterior probability of each mental illness based on the prior prevalence of each mental illness in different existing scenarios, in order to update the disease probability.

[0021] Furthermore, the posterior probability is calibrated using isotonic regression to obtain the final accurate probability value. The Bootstrap sampling method is then used to calculate the 95% confidence interval of the final probability, and the source of uncertainty is reported.

[0022] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention presents a multi-source data fusion-based auxiliary diagnostic and intervention system and method for mental health, featuring clinical-grade standardization based on SCID-V and DSM-5 standards to ensure authority; language friendliness: automatically translating diagnostic terms into questioning methods suitable for children and adolescents, improving compliance and authenticity; multi-source data fusion: combining subjective scales and objective cognitive tests to enhance the scientific rigor and accuracy of diagnosis; AI intelligent analysis: based on machine learning / deep learning models, outputting disease probabilities and interpretable evidence; visualization and reporting: generating graphical diagnostic reports for easy understanding by doctors, parents, and patients; continuous tracking: supporting multiple assessments and longitudinal effect evaluation after training, suitable for research and clinical follow-up; data security: supporting localized deployment and privacy protection, meeting medical data compliance requirements. Attached Figure Description

[0023] Figure 1 This is a flowchart of the mental health auxiliary diagnosis and intervention method based on multi-source data fusion according to the present invention. Detailed Implementation

[0024] The present invention will now be described in detail with reference to the accompanying drawings.

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0026] Example This embodiment provides a mental health auxiliary diagnosis and intervention system based on multi-source data fusion. The system specifically includes: The user terminal module is the client: It is used by users to complete medical consultations, answer psychological scale questions, conduct cognitive function tests, and record daily behaviors; it supports multi-form deployment, including mobile applications, mini-programs, and web clients, and has built-in voice input and natural language interaction interfaces to reduce the difficulty of user operation.

[0027] Data acquisition and preprocessing module: The system transforms user-input natural language text into structured consultation entries; standardizes the results of psychological scales such as PHQ-9 and GAD-7; and integrates cognitive tests such as TMT-B and Stroop tasks as supplementary features.

[0028] Multi-source data fusion module: By combining deep learning with a rule engine, multimodal data, including text, scale scores, reaction time data, and physiological signals, are mapped into a unified feature vector. An attention mechanism is introduced to weight different data sources, ensuring that the diagnostic process is interpretable.

[0029] Assisted Diagnosis and Intervention Engine: Diagnostic Model: A classification model trained based on the DSM-5 / ICD-11 standards, outputting the probability of occurrence of different mental illnesses; Interpretability Layer: Key questions and answers, scale items, and test indicators labeled with diagnostic conclusions; Intervention Recommendation Module: Matching personalized intervention suggestions based on diagnostic results, including psychological counseling, medication referral, cognitive training, or family education guidance.

[0030] Visualization and report generation module: It automatically generates structured diagnostic reports, including a symptom overview, disease probability distribution, key evidence, and intervention recommendations; the results are presented in chart format, suitable for doctors, parents, or schools to view; it supports longitudinal comparisons, showing the user's changing trends at different time points.

[0031] Backend management and doctor-side modules: Doctors / consultants can log in to the backend to view patient records and track the diagnostic process; it provides multi-user management, case retrieval, and data statistical analysis functions; and supports remote follow-up and intervention plan development.

[0032] The system provided in this embodiment also includes an encryption module, which has data encryption and access control mechanisms to ensure data security.

[0033] The specific implementation process of the system in this embodiment is as follows: After logging into the client, users enter the intelligent consultation module; the system guides users to answer questions in child / adolescent-friendly language and automatically corresponds to SCID-V items; at the same time, users complete psychological scale questions and cognitive function tests.

[0034] Data collection and standardization: The system converts natural language responses into structured labels such as low mood and lack of concentration using NLP algorithms; psychological scales automatically calculate total scores and sub-factor scores; cognitive tests extract reaction time, error rate, and executive function indicators.

[0035] Feature fusion and modeling: The data fusion module inputs multi-source features into deep learning models such as the multimodal Transformer; outputs the classification probability distribution of mental disorders, such as depression 72% and generalized anxiety disorder 41%; and ensures clinical reliability through cross-validation with the rule engine and AI results.

[0036] Diagnostic and intervention recommendations: The system automatically generates diagnostic results and lists supporting evidence, including key answers, scale scores, and cognitive indicators; Provide intervention suggestions, such as "recommend psychological counseling + cognitive training" or "recommend further hospital examination".

[0037] Results Display and Tracking: Generates visual reports such as radar charts, trend curves, and comparison charts; users and doctors can compare results in subsequent tests to form longitudinal tracking data.

[0038] Example 2 This embodiment provides a method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion, using the mental health auxiliary diagnosis and intervention system based on multi-source data fusion from Embodiment 1, such as... Figure 1 As shown, it specifically includes: Data is collected based on diagnostic entries from SCID-V or DSM-5 standards. Combined with a natural language processing module, medical language is automatically translated into user-friendly expressions appropriate for the assessed individual. User responses are also collected. Data includes consultation data, scales, cognitive data, physiological indicators, and previous examination reports. The natural language processing module uses the Qwen3 series models and the Deepseek model, employing models with different parameter sizes to handle questions of varying difficulty. For example, the consultation model might use models from Qwen3-8B to 32B to ask different types of psychological questions; when processing multi-source data fusion analysis, Qwen3-235B or Deepseek is used.

[0039] Scales and cognitive data include psychological scales for the assessed subjects, such as depression, anxiety, and personality scales, as well as objective cognitive tests, such as attention tests and executive function tests; Physiological data includes data collected by the system after the user completes relevant examinations such as EEG, near-infrared spectroscopy (fNIRS), eye-tracking data, and wearable devices.

[0040] Previous examination reports include the examination reports that the person being evaluated has previously undergone at medical institutions.

[0041] Data processing involves extracting semantic features, keyword features, and emotional features from the consultation data, scale and cognitive data, physiological indicator data, and previous examination reports; and extracting text data from the previous examination reports using AI image recognition. Feature fusion involves inputting the feature vector into a multimodal fusion model to obtain a comprehensive mental health feature vector. The assessment and analysis process involves inputting the comprehensive mental health feature vector into a mental health assessment model to obtain the assessment results. The specific analysis process is as follows: Constructing a discriminative model specifically includes constructing a comprehensive discriminative score (SDS): Standardized calculation method for consultation content: DSMScore = Symptom duration weight score * Number of items conforming to DSM-5 diagnostic criteria weight score / Total weight score of items; Standardized calculation method for scales: ScaleScore = Scale score / Total scale score. Standardized scale score, such as PHQ-9 = 18 / 27 = 0.67; If there are multiple scales, the weights are as follows: The comprehensive scale score of the depression scale is calculated as follows: Depression scale score * 0.7 + Anxiety scale score * 0.2 + Psychological resilience scale score * 0.1. If no scale is available, it is scored as 0. The Anxiety Scale Scale Score is calculated as follows: Anxiety Scale Score * 0.7 + Depression Scale Score * 0.2 + Psychological Resilience Scale Score * 0.1. If no scale is available, it is scored as 0. The overall scale score of the Bipolar Disorder Scale is calculated as follows: Scale score of the Mood Disorder Scale * 0.8 + Scale score of the SCL90 Scale * 0.2. If no scale is available, it is scored as 0. In the above scale_score calculation, if any scale is missing data, that scale will be scored as 0. Fixed weights: DSM: 0.8; Scale: 0.2; SDS = 0.8 × DSM_Score + 0.2 × Scale_Score; If there is no scale data, SDS=DSM_Score.

[0042] Attention fusion mechanism calculation method: Attention assessment test CPTS score = CPT score / 15 points; Wechsler Intelligence Scale score = intelligence score / 160 points.

[0043] For cognitive assessment weights specific to ADHD: SDS = 0.8 × CPTScore + 0.2 × DSM_Score; The SDS for academic difficulties is calculated as follows: SDS = 0.6 × Wechsler Intelligence Scale Score + 0.3 × CPTS score + 0.1 × DSMS score. Comprehensive Attention and Cognitive Ability (SDS) = CPTS score × Wechsler Intelligence Scale total score × Psychological Resilience Scale score.

[0044] Discriminant probability inference: Converting SDS into discriminant probabilities using a logistic function: (1) In the formula, SDS is the comprehensive discrimination score.

[0045] The evaluation analysis also includes using hierarchical Bayesian inference of the posterior probability of each mental illness based on its prior prevalence in different existing scenarios, to update the disease probabilities, as follows: Use the prior prevalence data for each disease in different scenarios from the following list; Table 1. Data on the prior prevalence of various diseases in different scenarios.

[0046] Prior probability ratio = Prior probability / (1 - Prior probability); Posterior probability ratio = (P_disc / (1-P_disc)) × prior probability ratio; Each piece of evidence is assigned a direction, namely support / refutation and strength, and the disease probability is updated progressively.

[0047] Probabilistic fusion and strong calibration: Calibration: On a reserved validation set, the generated posterior probabilities are calibrated using the isotonic regression method to obtain the final accurate probability values; Uncertainty: Using the Bootstrap sampling method, calculate the 95% confidence interval of the final probability and report the sources of uncertainty; Previous medical examination reports should only be used as a reference for professional judgment.

[0048] The large model is based on the qwen3 series models and deepseek, and was formed after fine-tuning and training on its own dataset.

[0049] The report is generated based on the above assessment results, including an assessment report and corresponding intervention recommendations, such as: core symptoms, risk stratification, visualized cognitive results, and treatment and intervention recommendations.

Claims

1. A mental health auxiliary diagnosis and intervention system based on multi-source data fusion, characterized in that, The system includes: The data acquisition module is used to collect medical history data, scale and cognitive data, and physiological indicator data of the evaluated subjects. The data processing module is used to preprocess the data collected by the data acquisition module and extract corresponding feature vectors respectively. The feature fusion module is used to fuse the feature vectors to obtain a comprehensive mental health feature vector; The assessment module is used to input the comprehensive mental health feature vector into the mental health assessment model to obtain the assessment results; The report generation module provides intervention recommendations and visualizes the obtained assessment results.

2. A method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion, employing the auxiliary diagnosis and intervention system for mental health based on multi-source data fusion as described in claim 1, characterized in that: The method specifically includes: Data is collected, including the patient's medical history, scales, cognitive data, physiological indicators, and previous examination reports. Data processing involves preprocessing the data and extracting the feature vectors corresponding to the data. Feature fusion involves inputting the feature vector into a multimodal fusion model to obtain a comprehensive mental health feature vector. The comprehensive mental health feature vector is input into a mental health assessment model to obtain the assessment results. Report generation: Based on the assessment results, an assessment report and corresponding intervention recommendations are generated.

3. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 2, characterized in that, The consultation data is acquired based on the standardized consultation system SCID-V or DSM-5, and the acquired consultation data is transformed into a user-friendly expression based on natural language processing.

4. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 3, characterized in that, The natural language processing includes sentiment dictionary analysis, keyword extraction, and semantic dependency analysis.

5. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 2, characterized in that, The scales and cognitive data include psychological scales and objective cognitive test scales for the subjects being assessed. The physiological indicator data includes the physiological examination data of the subject being evaluated; The previous examination reports include the examination reports that the subject underwent at medical institutions in the past.

6. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 2, characterized in that, The data processing specifically includes: Semantic features, keyword features, and emotional features were extracted from the consultation data, scale and cognitive data, physiological indicator data, and previous examination reports, respectively. The text content of the previous inspection reports is extracted using AI image recognition.

7. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 2, characterized in that, The assessment analysis specifically includes obtaining the probability of mental illness occurrence in the assessed individuals and constructing the following discriminant formula: （1） In the formula, SDS is the comprehensive discrimination score.

8. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 7, characterized in that, The comprehensive discriminant score SDS specifically includes: Standardized calculation method for consultation content: DSM Score = Symptom duration weight score × Number of matching items weight score / Total number of items weight score; Standardized calculation method for scales: Scale Score = Scale score / Total scale score; SDS = 0.8 × DSM Score + 0.2 × Scale Comprehensive Score.

9. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 2, characterized in that, The evaluation analysis also includes using hierarchical Bayesian inference of the posterior probability of each mental illness based on the prior prevalence of each mental illness in different existing scenarios, in order to update the disease probability.

10. The method for auxiliary diagnosis and intervention of mental health based on multi-source data fusion according to claim 9, characterized in that, The posterior probability is calibrated using isotonic regression to obtain the final accurate probability value. The Bootstrap sampling method is then used to calculate the 95% confidence interval of the final probability, and the source of uncertainty is reported.

Images