A method and system for patient data collection based on mental disorder questionnaires and mapping manuals

Abstract

This invention belongs to the field of smart healthcare technology and discloses a method and system for patient data collection based on a mental disorder questionnaire and mapping manual. The collection method includes: acquiring individualized information and constructing a personal graph by combining it with a pre-built mental disorder knowledge graph; constructing a mental disorder assessment questionnaire and a corresponding symptom dimension mapping manual for evaluating symptom dimensions based on principal component analysis; outputting a personalized questionnaire based on the personal graph, the mental disorder assessment questionnaire, the symptom dimension mapping manual, and machine learning; obtaining the questionnaire results; and obtaining the assessment results based on the questionnaire results and transformation rules. This invention solves the problem that existing mental health questionnaires are not standardized and cannot effectively interpret patient symptoms.

Description

Technical Field

[0001] This invention belongs to the field of smart healthcare technology, and in particular relates to a method for collecting patient data based on a mental disorder questionnaire and mapping manual. Background Technology

[0002] Because of stereotypes about mental health disorders, some patients tend to seek answers to their problems online. However, search engine results are mostly related to the search domain, and the accuracy of the answers needs to be verified.

[0003] The concept of Knowledge Graph (KG) was formally proposed by Google in 2012, initially aimed at improving the accuracy of its search engine and thus enhancing user experience. Currently, Knowledge Graphs have been successfully applied in many fields, such as search engines, question-answering systems, and recommendation systems. As a hot research area in artificial intelligence, medicine is also one of the vertical fields where Knowledge Graphs are most widely applied, showing promising performance in smart healthcare areas such as intelligent assisted triage, disease risk prediction, clinical decision support, and medical question answering. For example, Li Huafang and her team used machine learning and big data technologies to analyze and mine data, and combined it with the individual experience of different experts to develop a predictive model for the treatment effect of schizophrenia. Based on this, they constructed a knowledge graph for predicting the treatment effect of schizophrenia, providing support for treatment decisions. Fang Hui and her team constructed a knowledge graph for diabetes intervention, built a questionnaire question bank, summarized the questionnaire responses from users, and recommended diabetes intervention measures to users according to preset conditions based on the responses to the questionnaire questions. Li Zhike and his team used smartphone terminals to collect users' conversations and filter their information. Based on the user data and the constructed mental health knowledge graph, they judged the users' mental health status and finally transmitted the analysis results to users with sub-optimal mental health.

[0004] However, the aforementioned technologies have several limitations: First, the concepts involved in mental disorders are complex, and there may be discrepancies between the words used by patients and their actual symptoms; however, existing knowledge graphs designed for psychiatry have failed to construct a complete corpus covering all types of psychiatric diseases. Second, there are currently no quantifiable biological indicators for mental disorders; the aforementioned intelligent algorithms for assisted diagnosis can only make diagnoses based on quantitative data provided by patients, and cannot interpret and transform the diverse symptom descriptions provided by psychiatric patients. Third, different scales are suitable for different mental illnesses, and determining the type of scale used by a patient and the specific symptoms corresponding to the scale items often requires doctors to spend a long time interpreting and evaluating. Currently, intelligent algorithms applied in psychiatry cannot provide personalized scale outputs for corresponding patients, and cannot standardize the conclusions of assessment questionnaires according to the International Classification of Diseases, 10th Revision (ICD-10). Summary of the Invention

[0005] The technical problem to be solved by this invention is to overcome the shortcomings and defects mentioned in the background art above. By collecting structured and unstructured data of mental symptoms from hospital systems and scientific research databases, a knowledge graph covering all mental illnesses is constructed. By constructing a mental disorder questionnaire and symptom dimension mapping manual, different questionnaire items and results are converted into ICD-10 symptom dimensions and standardized scores. By constructing an integrated system based on knowledge graph intelligent reading and symptom dimension mapping manual, personalized scales and semantic analysis results are output to patients and doctors, and intelligent prompts are generated on the terminal. A personalized patient data collection system for psychiatric outpatient clinics has been developed.

[0006] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0007] A patient data collection method based on a mental disorder questionnaire and mapping manual includes the following steps: Step 1: Obtain individualized information and construct a personal graph by combining it with a knowledge graph of mental disorders; Step 2: Construct a mental disorder assessment questionnaire and corresponding symptom dimension mapping manual based on principal component analysis; Step 3: Based on the personal profile, mental disorder assessment questionnaire, symptom dimension mapping manual, and machine learning, a personalized questionnaire is generated. Step 4: Receive the questionnaire results and obtain symptom grading data based on the questionnaire results and conversion rules.

[0008] In the above-mentioned personalized patient data collection method, preferably, the individualized information includes structured information and unstructured information. The structured information includes age, gender, medical history diagnosis, test results, income, education, household registration, smoking and drinking history, and occupation. The unstructured information includes medical orders and medical record texts.

[0009] More preferably, step 2 includes: Step 201: Construct a preliminary mental disorder assessment questionnaire, establish a mapping relationship between the questionnaire items and symptom dimensions, and obtain a preliminary symptom dimension manual based on the mapping relationship; Step 202: Standardize the preliminary mental disorder assessment questionnaire and preliminary symptom dimension manual, construct a data matrix, select principal components, calculate the projection matrix of the selected principal components, perform dimensionality reduction, obtain the screened questionnaire items, and obtain the mental disorder assessment questionnaire and corresponding symptom dimension manual by combining ICD-10.

[0010] More preferably, in step 202, the standardization process is the z-score standardization method.

[0011] More preferably, step 3 includes: Step 301: Extract personal information from the personal profile, and obtain the corresponding symptom dimensions based on the personal information combined with the mental disorder assessment questionnaire and symptom dimension mapping manual; Step 302: Use personal information and corresponding symptom dimensions as a dataset, input them into machine learning training and validation, and output a personalized questionnaire.

[0012] More preferably, in step 302, the dataset is processed by synthesizing a minority of oversampled data and then used for machine learning training and validation; The machine learning algorithm used is a random forest algorithm based on the mapping relationship between questionnaire items and symptom dimensions, and optimized by grid search.

[0013] More preferably, when the questionnaire results are numerical, obtaining the evaluation results based on the questionnaire results and conversion rules includes: When the questionnaire result is on the scale of 0-4, the evaluation result is output directly. The 0-4 score range represents the range of the questionnaire results, with 4 being the highest score and 0 being the lowest score. When the questionnaire results are other score levels, the evaluation results are output by combining the linear mapping rule; The linear mapping rule is expressed as follows: ; in, This indicates that the assessment has yielded results; Indicates the score of the questionnaire results; and These represent the upper and lower limits of the target score, respectively. and These represent the upper limit and lower limit of the original score, respectively.

[0014] More preferably, when the questionnaire result is a language type, the step of obtaining the evaluation result based on the questionnaire result and the conversion rule includes: obtaining key information by combining the natural language recognition questionnaire result, and obtaining the evaluation result based on the key information and the scoring rule. The scoring rules are as follows: 0 points for no symptoms, 1 point for mild symptoms, 2 points for moderate symptoms, 3 points for obvious symptoms, and 4 points for severe symptoms.

[0015] Based on a general inventive concept, the present invention also provides a patient data collection system based on a mental disorder questionnaire and mapping manual, including a symptom knowledge graph module, a questionnaire and symptom mapping manual module, a machine learning output personalized questionnaire module, and a questionnaire analysis module. The symptom knowledge graph module stores a mental symptom knowledge graph constructed based on publicly available medical record data. The symptom knowledge graph module is used to construct a personal graph by combining the acquired individualized information with the mental disorder knowledge graph. The questionnaire and symptom mapping manual module is used to construct a mental disorder assessment questionnaire and a disorder dimension mapping manual based on principal component analysis. The machine learning output personalized questionnaire module is used to call up personal graphs, mental disorder assessment questionnaires, and disorder dimension mapping manuals to combine machine learning to output personalized questionnaires. The questionnaire analysis module is used to obtain questionnaire results and, based on the type of questionnaire results and different conversion rules, to obtain evaluation results.

[0016] The aforementioned patient data acquisition system, preferably, also includes a voice acquisition module and a semantic conversion module; The voice acquisition module is used to acquire medical consultation voice data and transmit the medical consultation voice data to the semantic conversion module; The semantic conversion module is used to combine the medical consultation voice data with semantic analysis to obtain the medical consultation semantic text, and transmit the medical consultation semantic text to the symptom knowledge graph module; The symptom knowledge graph module receives the semantic text of the medical visit as individualized information and combines it with the knowledge graph of mental disorders to construct a personal graph.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention gathers information resources on symptoms of mental illnesses, combines them with objective medical facts, and constructs a medical knowledge graph with symptom knowledge as its core. This consists of a mental illness symptom knowledge graph ontology model layer and a mental illness symptom knowledge graph data layer that provides actual data. It applies effective artificial intelligence technology to express professional knowledge of mental illnesses in a graph-like, semantic, and structured manner, enabling it to be recognized and calculated by computers. 2. Using patient information in the knowledge graph data layer of mental symptoms and symptom dimension information in the mapping manual as training data, based on machine learning technology, through model training and validation, we obtain the assessment questionnaire set corresponding to the symptom dimension that best matches the patient information, thereby realizing personalized push of patient assessment questionnaires and improving the efficiency of medical treatment. 3. Personalized questionnaires obtained through machine learning are more targeted, and the collection of patients' data-driven symptom results is faster and more standardized, effectively reducing the workload of doctors. Based on patient data, rapid classification and grading are possible, which facilitates the quick retrieval or use of data information in later diagnosis and helps improve the accuracy of subsequent diagnoses.

[0018] 4. By calling the mental disorder assessment questionnaire and symptom dimension mapping manual, and combining the questionnaire scoring rule conversion, different questionnaire items and results are converted into unified symptom dimensions and standardized scores. The various symptom dimensions and scores are summarized to improve the standardization of data, which is also conducive to data summarization and utilization. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the patient data collection method of the present invention; Figure 2 This is a schematic diagram of the construction mechanism of the patient data acquisition system of the present invention; Figure 3 This is a flowchart of the patient data acquisition system of the present invention; Figure 4 This is a schematic diagram of the constituent modules of the patient data acquisition system of the present invention. Detailed Implementation

[0021] To facilitate understanding of the present invention, the present invention will be described more fully and in detail below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present invention is not limited to the following specific embodiments.

[0022] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art. The technical terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of the invention.

[0023] Knowledge graphs are semantic network structures that represent relationships between entities through nodes and edges, providing a framework for integrating multi-source information. In mental health assessment, knowledge graphs integrate patient history, genetic information, symptom presentation, and medical literature, helping doctors comprehensively understand the patient's condition and facilitating personalized diagnosis and treatment. They help quickly identify correlations between factors, supporting doctors in uncovering potential causes, improving treatment accuracy, and providing a reference for predicting potential risks to patients. Knowledge graphs offer deeper and more comprehensive data analysis and applications in the field of mental health, and are expected to play a significant role in improving the assessment and treatment of mental disorders.

[0024] Example: Please see Figure 1 A patient data collection method based on a mental disorder questionnaire and mapping manual includes the following steps: S1. Construction of a knowledge graph of mental symptoms By aggregating information resources on symptoms of mental illness and combining them with objective medical facts, a medical knowledge graph centered on symptom knowledge is constructed, consisting of an ontology model layer for the mental illness knowledge graph and a data layer for the mental illness knowledge graph that provides actual data. S11. Collect patient information extensively. The information included in the patient database mainly comes from the hospital's medical record system and the individualized information corresponding to the patient in the medical record system. The sources of individualized information include scientific research information surveys in which the patient participated. All information includes structured information and unstructured information. The former includes patient age, gender, disease diagnosis, laboratory test results, questionnaire evaluation results, etc. from the electronic medical record system, as well as disease risk factor information collected from scientific research projects, such as family income, education, household registration, smoking and drinking history, occupation, etc.; the latter, based on the characteristics of psychiatric diagnosis and treatment, only includes text information such as medical orders and medical records. S12. Based on objective medical facts, construct a knowledge graph of mental symptoms, and establish the conceptual entity relationships of mental symptoms, in detail; Conceptual entities can be symptom types, patients, questionnaires, etc.; The relational entity can be the frequency, duration, location, nature, severity, diurnal characteristics, seasonal characteristics, triggering factors, relieving factors, etc. of the symptoms; Attributes can be information corresponding to relational entities; for example, day and night characteristics can be lighter in the morning and heavier in the evening.

[0025] S13. Integrate the structured and unstructured data of individuals to obtain a personal profile.

[0026] After preprocessing the patient's structured and unstructured data, conceptual entities, relationships, and attribute knowledge units are extracted from the data.

[0027] Using machine learning, the conceptual entities, relational entities, and attributes of mental symptoms are treated as a binary classification problem. Supervised learning techniques are employed, with pre-labeled data used to train the model, which then aligns, merges, and optimizes unlabeled data to form a personal atlas.

[0028] S2. Constructing a mental disorder assessment questionnaire and symptom dimension mapping manual. Symptom dimensions assessed by mental disorder questionnaires are usually reflected by a single item or multiple items. The correspondence between questionnaire items and symptom dimensions is defined as "mapping". S21. The establishment of the mental disorder assessment questionnaire and symptom dimension mapping manual will involve literature review and expert discussion to map and annotate the symptom dimensions corresponding to commonly used mental disorder questionnaire items. S22. For questionnaires and items with a large number of items and complex relationships between items, making it difficult to label the symptom dimensions through step S11, Principal Component Analysis (PCA) is used to reduce the dimensionality of the questionnaire items and extract feature components. S22.1 Data Preparation and Standardization: Based on the fact that mental symptoms are the knowledge graph data layer, symptom assessment information of patients with different diseases is extracted according to the disease type, and the z-score standardization method is used or the data is scaled to the same range. S22.2 Constructing a data matrix: Organize the data into a matrix, where rows represent samples and columns represent features; Calculate the covariance matrix, eigenvalues, and eigenvectors: Calculate the covariance matrix between features to describe the correlation or linear relationship between features; Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors. Eigenvectors represent the main directions or main components of the data. S22.3. Principal Component Selection: Select the number of principal components to retain based on the magnitude of their eigenvalues. Select the first few principal components with larger eigenvalues, as they contain the most important information about changes in the data. S22.4 Calculate the projection matrix: Construct a projection matrix using the selected principal components to project the original data into a new principal component space; S22.5 Dimensionality Reduction: By retaining the principal components, the original data is reduced to a new space with lower dimensions; S23, the diagnostic criteria of comparison and ICD-10 are used to standardize the symptom dimension labeling of the new division dimension.

[0029] S3, Personalized Assessment Questionnaire Output Module Using patient information from the knowledge graph of mental symptoms and symptom dimension information from the mapping manual as training data, and based on machine learning technology, the model is trained and validated to obtain the assessment questionnaire set corresponding to the symptom dimension that best matches the patient information, thereby realizing personalized push of patient assessment questionnaires. S31. Extract personalized patient data as feature information from the knowledge graph data layer of mental symptoms; S32. Use Scikit-Learning as the machine learning reference framework; S33. Considering the low prevalence of some mental disorders and the inherent imbalance of the data, the synthetic minority oversampling technique (SMOTE) is used to process the dataset. SMOTE avoids overfitting by synthesizing new samples along a straight line between a given sample and one of its neighboring samples. S34. Use grid search for hyperparameter tuning to optimize model performance; S35. Use the random forest classification algorithm to build the model; use cross-validation and machine learning models to predict the symptom dimensions, and use relevant features to calculate the weight of each feature; By using the random forest method, a large number of trees are built from the samples, and then an aggregate tree is generated by averaging all the trees.

[0030] S4. Questionnaire Results Analysis Module This module primarily calls upon the S1 Mental Disorders Questionnaire and Symptom Dimension Mapping Manual, combines the questionnaire scoring rules to convert different questionnaire items and results into unified symptom dimensions and standardized scores, and summarizes the results of each symptom dimension and score to output an assessment report. S41. Call the mental disorder questionnaire and symptom dimension mapping manual in S2 and mark the symptoms corresponding to the questionnaire items; S42. According to the data type of the questionnaire score, call different conversion rules to realize the standardization of questionnaire scores; (1) When the score is a numerical variable, for the numerical level result of 0-4 points, the original score result value is directly used as the final result value; for the numerical level result of 0-10 points or other numerical level results, the linear mapping rule is used. The principle is that the converted score = original score × (target score upper limit - target score lower limit) / (original score upper limit - original score lower limit) + target score lower limit, where the symptom dimension corresponding to the questionnaire item is based on the ICD-10 specification. (2) When the rating is a descriptive term, such as "depressed mood" or "elated mood", it is combined with Natural Language Processing (NLP) technology - a branch of artificial intelligence that focuses on enabling computers to understand, interpret and manipulate human language. By identifying the keyword information in the descriptive term, a standard rating level of 0-4 is defined as follows: 0 points are no symptoms (none, never, almost none, etc.), 1 point is mild symptoms (sometimes, occasionally, does not affect life, etc.), 2 points are moderate symptoms (about half the time, sometimes can be overcome, has a certain impact on work / study / social life, etc.), 3 points are obvious symptoms (most of the time, difficult to relieve on its own, significantly affects work / life / social life, feels pain, etc.), 4 points are severe symptoms (always, almost every day, difficult to take care of oneself, subjectively extremely painful). This rule system offers flexible configuration options, allowing users to customize the conversion method for symptom severity as needed. Users can adjust and optimize the parameters and rules for rule conversion based on the specific needs of an assessment questionnaire.

[0031] For each specific questionnaire item and result, this system collaborates extensively with experts in the field of mental health to guide the implementation of conversion rules, design personalized mappings for questionnaire items, and record them in the system. This addresses situations where existing conversion rules fail to accurately reflect symptom severity and clinical usability.

[0032] This system has the ability to dynamically update the knowledge graph, meaning that the system can update and expand the transformation rules in a timely manner as new assessment questionnaires are introduced and domain knowledge progresses, in order to adapt to the ever-changing needs of the assessment questionnaires.

[0033] S43. For the scoring results, users can mark the current symptom item as "intervention needed". If the score is 2 points higher than the intervention threshold, the current symptom item is marked as "intervention needed"; if it is lower than the intervention threshold, the current symptom item is marked as "intervention not needed". S44. When summarizing all questionnaire items and results, symptom items and results belonging to the same category will be averaged and then compared with the intervention threshold to output the intervention markers and clinical significance of all symptom results.

[0034] Please see Figure 2-4 The present invention also provides a patient data collection system based on a mental disorder questionnaire and mapping manual, including a symptom knowledge graph module, a questionnaire and symptom mapping manual module, a voice collection module, and a semantic conversion module; The machine learning output module includes a personalized questionnaire module and a questionnaire analysis module. The symptom knowledge graph module stores a knowledge graph of mental symptoms built based on publicly available medical record data. The voice acquisition module is used to collect medical consultation voice data and transmit the medical consultation voice data to the semantic conversion module; The semantic conversion module is used to combine medical consultation voice data with semantic analysis to obtain medical consultation semantic text, and then transmit the medical consultation semantic text to the symptom knowledge graph module; The Symptom Knowledge Graph module is used to receive semantic text from medical visits as individualized information and combine it with the knowledge graph of mental disorders to construct a personal knowledge graph. The Questionnaire and Symptom Mapping Manual module is used to construct a mental disorder assessment questionnaire and a disorder dimension mapping manual based on principal component analysis. The machine learning output personalized questionnaire module is used to call up personal graphs, mental disorder assessment questionnaires, and disorder dimension mapping manuals to combine machine learning to output personalized questionnaires. The questionnaire analysis module is used to obtain questionnaire results and, based on the type of questionnaire results and different conversion rules, to obtain evaluation results.

[0035] Experimental Case 1: Patient information was collected to create a patient model, including gender, age, perinatal experiences, BMI, lifestyle habits, and economic status. The patient information obtained is as follows: female, 25 years old, perinatal woman, BMI 16, lifestyle habits include staying up late, low income; she consulted a psychiatrist and was given the PHQ-9 questionnaire (The Patient Health Questionnaire-9). Taking the first item of the questionnaire, "I feel no interest or enjoyment in anything," as an example, by referring to the mental disorder questionnaire and symptom dimension mapping manual, this item elicited the symptom "loss of interest," with a score of "3." The results of the assessment questionnaire were converted into the current assessment item's result value consistent with the standard rating scale, and then compared with the current symptom intervention threshold score of 2. This symptom item was then marked as "required / not required for intervention." In this embodiment, the first item of PHQ-9 results in "3 points," which is a continuous variable. The data type of the rating level is consistent with the standard rating. If the rating levels are inconsistent, a linear mapping rule is used. The principle is: Transformed Score = Original Score × (Upper Limit of Target Range - Lower Limit of Target Range) / (Upper Limit of Original Range - Lower Limit of Original Range) + Lower Limit of Target Range; the questionnaire item result value is transformed into the current questionnaire item result value consistent with the standard rating, i.e., 3. (4-0) / (3-0)+0=4.

[0036] In this embodiment, the result value of the first item of PHQ-9 is 4, which is greater than the critical score. The clinical significance of the symptom result is: "The patient's loss of interest is severe and affects daily life. Timely intervention is required. This symptom is common in affective disorders such as depression and needs to be further analyzed in conjunction with other symptoms."

[0037] Repeat steps S22 to S24, analyzing each questionnaire item and result of all assessment questionnaires in sequence, obtaining the "intervention needed" mark and clinical significance analysis for each questionnaire item and result; summarize the "intervention needed" mark and clinical significance analysis of all questionnaire items and results, and output the test report.

[0038] In this experimental case, the patient was assessed using the PHQ-9. All results will be summarized and reported in the form of a spreadsheet. The analysis report includes the overall questionnaire assessment results and the questionnaire item analysis results (extracting symptoms and intervention markers). Clicking "Operation" will output the results and indicate the analytical significance of the corresponding symptoms or the overall questionnaire score.

[0039] The summary of all results is as follows: This patient is a perinatal woman, a high-risk group for depression, with a PHQ-9 score of 23, indicating possible severe depression. The patient's symptoms of "loss of interest," "depressed mood," "poor sleep," "decreased energy," "self-blame," and "decreased attention" are significant, requiring further questioning and evaluation by a doctor, and appropriate intervention measures. In addition, the patient has the following risk factors that need attention: low income, and attention should also be paid to the perinatal status to prevent adverse perinatal events.

[0040] Experimental Case 2: Patient information was obtained to create a patient model, including gender, age, BMI, lifestyle habits, appetite, sleep patterns, and physical symptoms. The patient information obtained is as follows: Female, 30 years old, BMI 17.2, excessive worry about work and family, poor appetite, difficulty falling asleep, tension headaches, palpitations, and upper abdominal discomfort. She sought treatment at a psychiatric department and was given the GAD-7 questionnaire (Validation of the Generalized Anxiety Disorder-7). Taking the first item of this questionnaire as an example, namely "feeling uneasy, worried, and irritable," by referring to the mental disorder questionnaire and symptom dimension mapping manual, the symptom elicited by this item is "panic," and the item result is "3 points." The results of the assessment questionnaire were converted into the current assessment item's result value with a standard rating level, and then compared with the current symptom intervention cutoff score of 2. The symptom item was then marked as "intervention needed / not needed". In this embodiment, the first item result of GAD-7 is "3 points", which is a continuous variable. The data type of the rating level is consistent with the standard rating. If the rating levels are inconsistent, a linear mapping rule is used. The principle is: Transformed Score = Original Score × (Target Range Upper Limit - Target Range Lower Limit) / (Original Range Upper Limit - Original Range Lower Limit) + Target Range Lower Limit. This transforms the questionnaire item result value into the current questionnaire item result value consistent with the standard rating, i.e., 3. (4-0) / (3-0)+0=4.

[0041] In this embodiment, the first item of GAD-7 has a result value of 4, which is greater than the critical score. The clinical significance of the symptom result is: "The patient has been troubled by worries about possible bad things in the future, feels uneasy, and it affects daily life. Timely intervention is needed. This symptom is common in generalized anxiety disorder and needs to be further analyzed in combination with other symptoms."

[0042] Repeat steps S22 to S24, and analyze each questionnaire item and result of all assessment questionnaires in turn, and obtain the "intervention required" mark and clinical significance analysis for each questionnaire item and result; Summarize all questionnaire items and results, identify "intervention needed" markers and their clinical significance, and output a test report.

[0043] In this experimental case, the patient underwent a GAD-7 assessment. A summary and detailed report of all results are shown in Table 1, which will be generated in spreadsheet format. The analysis report includes the overall questionnaire assessment results and the results of questionnaire item analysis (eliminating symptoms and intervention markers). Clicking "Operation" will output the results, indicating the analytical significance of the corresponding symptoms or the overall questionnaire score. An example summary of all results is as follows: The patient is a middle-aged woman, in a high-risk group for anxiety, with a GAD-7 score of 13, suggesting possible moderate anxiety disorder; the patient exhibits significant symptoms of "panic" and "motor tension," requiring further questioning and evaluation by a physician, and appropriate intervention measures.

[0044] Table 1 Summary of Patient Assessment Questionnaire Analysis Results

[0045] Experimental Case 3: Patient information was obtained to create a patient model, including information such as gender, age, lifestyle habits, social situation, and sleep patterns. The obtained patient information is as follows: Male, 32 years old, enjoys spending money, prone to arguments with family members, experiences reduced sleep but remains energetic during the day, consulted a psychiatrist, and was given the Young Mania Rating Scale (YMRS) questionnaire. Taking the fifth item of the questionnaire, "Subjectively feeling irritable," as an example, by referring to the mental disorder questionnaire and symptom dimension mapping manual, this item elicited the symptom "irritability," and the result was "2 points." The results of the assessment questionnaire were converted into the current assessment item's result value with a standard rating level, and then compared with the current symptom intervention cutoff score of 2. The symptom item was then marked as "intervention needed / not needed". In this embodiment, the fifth item of YMRS results in "2 points," which is a continuous variable. The data type of the rating level is consistent with the standard rating. If the rating levels are inconsistent, a linear mapping rule is used. The principle is: Transformed Score = Original Score × (Upper Limit of Target Range - Lower Limit of Target Range) / (Upper Limit of Original Range - Lower Limit of Original Range) + Lower Limit of Target Range. This transforms the questionnaire item result value into the current questionnaire item result value consistent with the standard rating, i.e., 2. (4-0) / (8-0)+0=1.

[0046] In this embodiment, the result value of the first YMRS entry is 1, which is less than the critical score. The clinical significance of the symptom result is "The patient's irritability symptoms are not obvious and no treatment is needed for the time being".

[0047] Repeat steps S22 to S24, and analyze each questionnaire item and result of all assessment questionnaires in turn, and obtain the "intervention required" mark and clinical significance analysis for each questionnaire item and result; Summarize all questionnaire items and results, identify "intervention needed" markers and their clinical significance, and output a test report.

[0048] In this experimental case, the patient underwent a YMRS assessment. All results and detailed reports are summarized in Table 2 and will be generated in spreadsheet format. The analysis report includes the overall questionnaire assessment results and the results of questionnaire item analysis (eliminating symptoms and intervention markers). Clicking "Operation" will output the results, indicating the analytical significance of the corresponding symptoms or the overall questionnaire score. An example summary of all results is as follows: The patient is a middle-aged male with a YMRS score of 26, suggesting possible bipolar disorder. The patient exhibits significant symptoms of "elevated mood" and "increased activity and energy," requiring further questioning and evaluation by a physician, and appropriate intervention measures.

[0049] Table 2 Summary of Patient Assessment Questionnaire Analysis Results

[0050] In summary, this invention constructs a knowledge graph of mental symptoms, expressing clinical professional knowledge in a graphical, semantic, and structured manner. It integrates literature review, expert discussion, and statistical analysis to build a mapping manual between mental disorder questionnaire items and symptom dimensions. Using machine learning, it matches patients' individualized information with their symptom dimensions, combining the mental disorder questionnaire items with the symptom mapping manual to find the most suitable assessment questionnaire for each patient, facilitating personalized questionnaire delivery to different patients. Based on this, a method for analyzing mental disorder questionnaire items was developed, achieving normalization and standardization of questionnaire scores and data standardization. Through the utilization of diverse information, it enables effective interaction between doctors and patients and flexible data management.

Claims

1. A method for collecting patient data based on a mental disorder questionnaire and mapping manual, characterized in that, Includes the following steps: Step 1: Obtain individualized information and construct a personal graph by combining it with a knowledge graph of mental disorders; Step 2: Construct a mental disorder assessment questionnaire and corresponding symptom dimension mapping manual based on principal component analysis; Step 2 includes: Step 201: Construct a preliminary mental disorder assessment questionnaire, establish a mapping relationship between the questionnaire items and symptom dimensions, and obtain a preliminary symptom dimension manual based on the mapping relationship; Step 202: Standardize the preliminary mental disorder assessment questionnaire and the preliminary symptom dimension manual, construct a data matrix, select principal components, calculate the projection matrix of the selected principal components, perform dimensionality reduction, obtain the screened questionnaire items, and obtain the mental disorder assessment questionnaire and the corresponding symptom dimension manual by combining ICD-10. Step 3: Based on the personal profile, mental disorder assessment questionnaire, symptom dimension mapping manual, and machine learning, a personalized questionnaire is generated. Step 3 includes: Step 301: Extract personal information from the personal profile, and obtain the corresponding symptom dimensions based on the personal information combined with the mental disorder assessment questionnaire and symptom dimension mapping manual; The personal information includes: demographic characteristics, conceptual entities and their attributes; Step 302: Use personal information and corresponding symptom dimensions as a dataset, input them into machine learning training and validation, and output a personalized questionnaire; Step 4: Receive the questionnaire results and obtain symptom grading data based on the questionnaire results and conversion rules.

2. The patient data collection method based on the mental disorder questionnaire and mapping manual according to claim 1, characterized in that, The individualized information includes structured and unstructured information. The structured information includes age, gender, medical diagnosis, test results, income, education, household registration, smoking and drinking history, and occupation. The unstructured information includes medical orders and medical record texts.

3. The patient data collection method based on the mental disorder questionnaire and mapping manual according to claim 1, characterized in that, In step 202, the standardization process is the z-score standardization method.

4. The patient data collection method based on the mental disorder questionnaire and mapping manual according to claim 1, characterized in that, In step 302, the dataset is processed by synthesizing a minority of oversampled data and then used for machine learning training and validation; The machine learning algorithm used is a random forest algorithm based on the mapping relationship between questionnaire items and symptom dimensions, and optimized by grid search.

5. The patient data collection method based on the mental disorder questionnaire and mapping manual according to any one of claims 1-4, characterized in that, When the questionnaire results are numerical, obtaining the evaluation results based on the questionnaire results and conversion rules includes: When the questionnaire result is on the scale of 0-4, the evaluation result is output directly. When the questionnaire results are other score levels, the evaluation results are output by combining the linear mapping rule; The linear mapping rule is expressed as follows: ； in, This indicates that the assessment has yielded results; Indicates the score of the questionnaire results; and These represent the upper and lower limits of the target score, respectively. and These represent the upper limit and lower limit of the original score, respectively.

6. The patient data collection method based on the mental disorder questionnaire and mapping manual according to any one of claims 1-4, characterized in that, When the questionnaire result is a language type, obtaining the evaluation result based on the questionnaire result and conversion rules includes: obtaining key information by combining the natural language recognition questionnaire result, and obtaining the evaluation result based on the key information and scoring rules. The scoring rules are as follows: 0 points for no symptoms, 1 point for mild symptoms, 2 points for moderate symptoms, 3 points for obvious symptoms, and 4 points for severe symptoms.

7. A patient data collection system based on a mental disorder questionnaire and mapping manual, performing the method of claim 1, characterized in that, It includes a symptom knowledge graph module, a questionnaire and symptom mapping manual module, a machine learning output personalized questionnaire module, and a questionnaire analysis module. The symptom knowledge graph module stores a mental symptom knowledge graph constructed based on publicly available medical record data. The symptom knowledge graph module is used to construct a personal graph by combining the acquired individualized information with the mental disorder knowledge graph. The questionnaire and symptom mapping manual module is used to construct a mental disorder assessment questionnaire and a disorder dimension mapping manual based on principal component analysis. The machine learning output personalized questionnaire module is used to call up personal graphs, mental disorder assessment questionnaires, and disorder dimension mapping manuals to combine machine learning to output personalized questionnaires. The questionnaire analysis module is used to receive questionnaire results and obtain evaluation results based on the questionnaire result type and different conversion rules.

8. The patient data collection system based on the mental disorder questionnaire and mapping manual according to claim 7, characterized in that, It also includes a voice acquisition module and a semantic conversion module; The voice acquisition module is used to acquire medical consultation voice data and transmit the medical consultation voice data to the semantic conversion module; The semantic conversion module is used to combine the medical consultation voice data with semantic analysis to obtain the medical consultation semantic text, and transmit the medical consultation semantic text to the symptom knowledge graph module; The symptom knowledge graph module receives the semantic text of the medical visit as individualized information and combines it with the knowledge graph of mental disorders to construct a personal graph.

Images