Emotional disorder full-cycle assessment system, method, device, and storage medium

Abstract

The application provides a mood disorder full-cycle evaluation system, method, device and storage medium, and relates to the technical field of diagnosis and treatment assistance. The system comprises: a feature construction module, configured to construct a psychological pathophysiological feature vector based on a standardized feature vector; a main type hierarchical evaluation module, configured to output a main type label with the highest matching probability and a corresponding main type confidence; a subtype subdivision identification module, configured to take the main type label as a constraint condition, output a subtype label and a corresponding subtype confidence; an evaluation output module, configured to generate a mood disorder evaluation result; a closed-loop update module, configured to generate an updated main type label and an updated subtype label; and a dynamic monitoring module, configured to trigger the closed-loop update module, control the feature construction module, the main type hierarchical evaluation module and the subtype subdivision identification module to re-execute, and output a closed-loop communication feedback result. The system can improve the accuracy of mood disorder evaluation.

Description

Technical Field

[0001] This application relates to the field of diagnostic and treatment assistance technology, specifically to a system, method, device, and storage medium for the full-cycle assessment of mood disorders. Background Technology

[0002] In current technologies, the assessment of mood disorders mainly relies on clinical interviews, standardized scales, and physicians' subjective experience, and its accuracy has long been limited by multiple factors. First, the symptoms of different mood disorders (such as depression, bipolar disorder, generalized anxiety disorder, and post-traumatic stress disorder) highly overlap in the early stages. For example, sleep disturbances, depressed mood, and irritability are common in many diseases. However, existing diagnostic systems mostly use single-dimensional symptom counts or total score thresholds for judgment, lacking effective modeling of the characteristic differences between different disorders, resulting in low accuracy in differential diagnosis. Second, most existing assessment tools are based on cross-sectional, single-time data collection, such as completing a scale assessment or a structured clinical interview in an outpatient setting. This cannot capture the dynamic changes of patients' symptoms, physiological indicators, and behavioral characteristics over time. Since mood disorders themselves are characterized by fluctuation, relapse, and uncertain disease progression, static assessment results are difficult to reflect the true state of the condition, often leading to misdiagnosis or missed diagnosis.

[0003] Traditional methods are relatively crude in data processing. For example, they use the same norm reference values ​​for patients of different ages, genders, and comorbid backgrounds, ignoring the impact of individual differences on the interpretation of indicators, resulting in poor consistency of assessment results across different subgroups. Furthermore, most existing assessment models only output a general diagnostic category, failing to further differentiate subtypes with different biomarker characteristics and treatment response patterns within the same disease category, leading to a lack of precision in the assessment of complex cases. In summary, due to overlapping symptoms, static assessment, neglect of individual differences, and lack of subtype segmentation, existing mood disorder assessment technologies generally suffer from insufficient accuracy, making it difficult to meet the clinical needs for early identification and personalized treatment. Summary of the Invention

[0004] Based on this, this application provides a system, method, device and storage medium for the full-cycle assessment of mood disorders. By integrating hierarchical diagnosis and a full-cycle closed-loop approach, the accuracy of mood disorder assessment can be improved.

[0005] Firstly, this application provides a full-cycle assessment system for mood disorders, which may include:

[0006] The feature construction module is used to generate a standardized feature vector based on the received initial feature objects of the patient, and to construct a psychopathophysiological feature vector based on the standardized feature vector; The main type hierarchical assessment module is used to input the psychopathophysiological feature vector into a preset classification model or rule framework, calculate the matching probability of the patient belonging to each of the multiple candidate main types of emotional disorders, and output the main type label with the highest matching probability and the corresponding main type confidence. The subtype segmentation and identification module is used to identify the subtype to which the patient belongs in the feature subspace corresponding to the main type label, based on the psychopathophysiological feature vector, using the main type label as a constraint, and output the subtype label and the corresponding subtype confidence. An assessment output module is used to generate an emotional disorder assessment result based on at least one of the main type label, the subtype label, the updated main type label, and the updated subtype label; The closed-loop update module is used to generate an updated psychopathophysiological feature vector based on the updated data after receiving the patient's updated data, and to re-input the updated psychopathophysiological feature vector into the main type hierarchical assessment module and the subtype subdivision identification module to generate updated main type labels and updated subtype labels. The dynamic monitoring module is used to trigger the closed-loop update module when the monitoring cycle arrives or when updated data of the patient is received, and to control the feature construction module, the main type hierarchical assessment module, and the subtype subdivision identification module to be re-executed, and to output closed-loop communication feedback results.

[0007] Optionally, the feature construction module includes: The stratification correction unit is used to map the patient to a reference stratification group based on demographic information or clinical background information, and to perform baseline correction on the target indicators in the initial feature object according to the reference stratification group to generate a corrected feature object; The cleaning and standardization unit is used to perform missing value filling, outlier handling, unit unification and time alignment operations on the corrected feature object to generate the standardized feature vector. The psychopathophysiological mapping unit is used to weight and combine the symptom features, vital signs, biomarkers and behavioral features in the standardized feature vector according to a preset mapping relationship to generate the psychopathophysiological feature vector.

[0008] Optionally, the feature construction module further includes: A psychological abnormality risk index calculation unit is used to calculate the psychological abnormality risk index based on the aforementioned psychopathophysiological feature vector. The feature storage unit is used to structurally store the psychopathophysiological feature vector and the psychological abnormality risk index, and associate them with time tags and data source identifiers.

[0009] Optionally, the main type hierarchical evaluation module includes: The hierarchical model loading unit is used to load pre-trained hierarchical evaluation models or rule models. The main type inference unit is used to calculate the matching probability or score of different main types of emotional disorders based on the psychopathophysiological feature vector. A confidence scoring unit is used to calculate the confidence level of the main type based on the output probability distribution of each main type. The tag storage unit is used to store and output the hierarchical results of the main type and the confidence level of the main type.

[0010] Optionally, the subtype segmentation identification module includes: The subspace modeling unit is used to extract feature dimensions related to the main type label from the psychopathophysiological feature vector under the constraint of the main type label, and construct the feature subspace; The subtype inference unit is used to calculate the distance or probability between the patient and multiple preset subtype centers within the feature subspace, and to determine the subtype label based on the minimum distance or the maximum probability. The subtype confidence score unit is used to calculate the degree of matching corresponding to the subtype label as the subtype confidence score.

[0011] Optionally, the closed-loop update module includes: A re-acquisition triggering unit is used to detect whether the updated data has been received; The state fusion unit is used to weight and fuse the psychopathophysiological feature vector generated in the previous round of evaluation with the new input feature vector generated based on the updated data according to the historical state retention weight, so as to generate the updated psychopathophysiological feature vector. The re-evaluation scheduling unit is used to sequentially send the updated psychopathophysiological feature vectors into the main type hierarchical evaluation module and the subtype subdivision identification module.

[0012] Optionally, the dynamic monitoring module includes: A monitoring cycle control unit is used to receive and store the monitoring cycle; A re-collection triggering unit is used to generate a trigger signal when the monitoring period arrives or when the updated data is received; The re-evaluation scheduling unit is used to call the closed-loop update module according to the trigger signal, and sequentially schedule the feature construction module, the main type hierarchical evaluation module and the subtype subdivision identification module to be re-executed; A closed-loop communication feedback unit is used to output the closed-loop communication feedback result, which includes at least one of the updated main type label, the updated subtype label, and the mood disorder assessment output information.

[0013] Secondly, this application provides a method for assessing mood disorders throughout their entire lifecycle, including: A standardized feature vector is generated based on the initial feature objects of the received patient, and a psychopathophysiological feature vector is constructed based on the standardized feature vector. The psychopathophysiological feature vector is input into a preset classification model or rule framework to calculate the matching probability of the patient belonging to each of the multiple candidate emotional disorder main types, and outputs the main type label with the highest matching probability and the corresponding main type confidence. Using the main type label as a constraint, within the feature subspace corresponding to the main type label, the subtype to which the patient belongs is identified based on the psychopathophysiological feature vector, and the subtype label and the corresponding subtype confidence are output. After receiving the patient's updated data, an updated psychopathophysiological feature vector is generated based on the updated data, and an updated main type label and an updated subtype label are generated based on the updated psychopathophysiological feature vector. Based on at least one of the main type label, the subtype label, the updated main type label, and the updated subtype label, generate an emotional disorder assessment result; When the monitoring period arrives or when updated data for the patient is received, a closed-loop update mechanism is triggered, and the steps of re-executing the mood disorder assessment are controlled, as well as the output of closed-loop communication feedback results.

[0014] Thirdly, this application provides a computer device, the 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 methods described above.

[0015] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0016] Compared with existing technologies, the beneficial effects of this application are as follows: The full-cycle assessment system for mood disorders can first determine the patient's primary type, and then identify subtypes within the corresponding feature subspace using the primary type as a constraint, forming a coarse-to-fine two-layer diagnostic structure. This can distinguish between different mood disorders with overlapping symptoms, and further subdivide clinical manifestations within the same primary type, effectively reducing the misdiagnosis rate and supporting individualized intervention. Furthermore, a closed-loop update is automatically triggered when the monitoring cycle ends or new data is received, re-inputting the updated features into the hierarchical diagnostic process, achieving a complete closed loop from initial assessment to continuous tracking, and from static output to dynamic iteration. The full-cycle assessment system for mood disorders provided by this application can improve the accuracy of mood disorder assessment by integrating hierarchical diagnosis and a full-cycle closed-loop approach. Attached Figure Description

[0017] Figure 1 A schematic diagram of the architecture of the full-cycle assessment system for mood disorders provided in this application embodiment.

[0018] Figure 2 A schematic diagram illustrating the steps of the full-cycle assessment method for mood disorders provided in this application embodiment.

[0019] Figure labels: 10-Full-cycle assessment system for mood disorders; 11-Feature construction module; 12-Main type hierarchical assessment module; 13-Subtype subdivision identification module; 14-Assessment output module; 15-Closed-loop update module; 16-Dynamic monitoring module. Detailed Implementation

[0020] The present application will now be described in further detail with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the subject matter of the present application to the following embodiments. All technologies implemented based on the content of the present application fall within the scope of protection of the present application.

[0021] In the description of the embodiments in this application, "a plurality of" means two or more, unless otherwise expressly specified. The reference to "embodiment" herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] Please refer to Figure 1 , Figure 1 A schematic diagram of the architecture of the full-cycle assessment system for mood disorders provided in this application embodiment. The full-cycle assessment system for mood disorders 10 (hereinafter referred to as the system) may include: The feature construction module 11 is used to generate a standardized feature vector based on the received initial feature object of the patient, and to construct a psychopathophysiological feature vector based on the standardized feature vector; The main type hierarchical assessment module 12 is used to input psychopathophysiological feature vectors into a preset classification model or rule framework, calculate the matching probability of the patient belonging to each main type among multiple candidate emotional disorder main types, and output the main type label with the highest matching probability and the corresponding main type confidence. The subtype subdivision identification module 13 is used to identify the subtype to which the patient belongs based on the psychopathophysiological feature vector within the feature subspace corresponding to the main type label, using the main type label as a constraint, and output the subtype label and the corresponding subtype confidence. The assessment output module 14 is used to generate mood disorder assessment results based on at least one of the main type label, subtype label, updated main type label, and updated subtype label; The closed-loop update module 15 is used to generate an updated psychopathophysiological feature vector based on the updated data after receiving the patient's updated data, and to re-input the updated psychopathophysiological feature vector into the main type stratification assessment module 12 and the subtype subdivision identification module 13 to generate updated main type labels and updated subtype labels. The dynamic monitoring module 16 is used to trigger the closed-loop update module 15 when the monitoring cycle arrives or when updated patient data is received, and to control the feature construction module 11, the main type stratification assessment module 12, the subtype subdivision identification module 13 to re-execute, and to output closed-loop communication feedback results.

[0023] In this embodiment, the feature construction module 11 receives the patient's initial feature object, which is a unified feature set formed by processing multiple sources of data, including demographic information, clinical history data, text-based symptom description data, vital sign data, routine clinical blood test indicators, behavioral characteristic data, treatment data, psychological assessment data, and follow-up data, through field parsing, unit unification, and missing value marking. The initial feature object can be obtained by the feature construction module 11 based on the patient information, or it can be obtained by other systems processing the patient information and then sending it to the feature construction module 11. The standardized feature vector is a numerical vector with fixed dimensions obtained after performing operations such as missing value imputation, outlier handling, unit unification, and time alignment on the initial feature object. For example, it maps indicators with different units, such as blood pressure, uric acid, and pulse, to a unified numerical range and arranges them in a preset order. The psychopathophysiological feature vector is a multi-dimensional vector generated by weighting and combining symptom features, vital signs, biomarkers, and behavioral features based on the standardized feature vector through a preset mapping relationship. Each dimension corresponds to psychopathophysiological dimensions such as emotion, anxiety, agitation, inflammatory state, metabolic state, or neuroendocrine state.

[0024] The principal type stratification assessment module 12 is a component in the system used to determine the major category of mood disorder to which a patient belongs. The preset classification model or rule framework can be a pre-trained machine learning classification model (such as a softmax multi-classifier) ​​or a decision function composed of multiple threshold logics and weight rules. Candidate mood disorder principal types include at least the depressive spectrum, bipolar spectrum, post-traumatic stress disorder spectrum, generalized anxiety disorder spectrum, panic attack spectrum, and social anxiety spectrum. The matching probability refers to the likelihood that the psychopathophysiological feature vector belongs to each candidate principal type, usually represented by a probability value between 0 and 1, with the sum of the matching probabilities of all principal types being 1. The principal type label is the identifier of the candidate mood disorder principal type with the highest matching probability, such as "depressive spectrum" or "bipolar spectrum." The principal type confidence score indicates the reliability of the matching result, usually taken as the maximum matching probability value itself; for example, if the matching probability of the depressive spectrum is 0.85, then the principal type confidence score is 0.85.

[0025] The subtype segmentation and identification module 13 is a component in the system used to further differentiate different clinical manifestations within a main type. The feature subspace corresponding to the main type label refers to the low-dimensional space formed by extracting feature dimensions related to the main type from the psychopathophysiological feature vector after the main type is determined. For example, in the depression spectrum, the focus is on extracting feature dimensions such as depressed mood, sleep disturbances, and anxiety levels, while ignoring dimensions unrelated to the main type. A subtype refers to a more refined category within the same main type based on dimensions such as symptom structure, comorbidity pattern, biomarker pattern, treatment history, and disease course characteristics. For example, in the depression spectrum, it can be divided into high anxiety with sleep disturbance, typical depressed mood, or mixed agitation tendency. The subtype label is the identifier of the specific subtype identified, such as "high anxiety with sleep disturbance." Subtype confidence indicates the degree of matching of the subtype identification result, which can be the inverse distance between the psychopathophysiological feature vector and the preset subtype center or a conditional probability value.

[0026] The closed-loop update module 15 is the component in the system responsible for responding to new data and triggering re-evaluation. The updated psychopathophysiological feature vector is a new feature vector obtained by weighted fusion of the psychopathophysiological feature vector generated in the previous round of evaluation and the new input feature vector generated based on the updated data, according to the historical state retention weights. The closed-loop update module 15 re-feeds the updated psychopathophysiological feature vector into the main type hierarchical evaluation module 12 and the subtype subdivision identification module 13, causing these two modules to re-execute the matching probability calculation and subtype identification, thereby generating updated main type labels and updated subtype labels.

[0027] The assessment output module 14 is a component in the system that generates clinically usable information based on the stratified assessment results. The assessment results of mood disorders include at least one of the following: individualized intervention strategies, multi-time-window prognostic indices, disease course trend scores, or risk warning signals. For example, it can output intervention strategies such as "suggesting increased follow-up frequency, focusing on monitoring sleep indicators and anxiety-related symptoms," or output a trend judgment of "short-term risk rising, medium-term risk stable."

[0028] The dynamic monitoring module 16 is a component in the system that controls the scheduling and triggering of the entire closed-loop process. The monitoring cycle is a pre-set time interval, which can be a fixed duration for short-term (e.g., daily or weekly), medium-term (e.g., monthly), or long-term (e.g., quarterly). When the monitoring cycle arrives or when updated patient data is received, the dynamic monitoring module 16 triggers the closed-loop update module 15 to start working and controls the feature construction module 11, main type stratification assessment module 12, and subtype subdivision identification module 13 to re-execute the entire processing flow in sequence. The closed-loop communication feedback result refers to the information output by the dynamic monitoring module 16 used to prompt changes in the patient's status and suggest actions to the medical terminal, follow-up terminal, or management terminal. This result includes at least one of the updated main type label, updated subtype label, mood disorder assessment result, or risk warning. In this embodiment, the periodic triggering and data-driven triggering of the dynamic monitoring module 16 can enable the entire system to form a full-cycle intelligent management closed loop from initial feature construction, main type stratification, subtype identification, closed-loop update to assessment output and re-triggering.

[0029] For example, demographic information may include age, sex, marital status, occupation, ethnicity, and place of origin. Clinical history data may include the number of previous episodes, stage of illness, family history of mental illness, smoking history, alcohol consumption history, medication history, previous surgical history, and allergy history. Textual symptom description data may include the chief complaint, summary of present illness, admission and discharge diagnoses, past comorbidities, and scale scores. Vital signs data may include body temperature, pulse, respiration, systolic blood pressure, and diastolic blood pressure. Routine clinical blood tests may include complete blood count, biochemistry, endocrine, inflammation, and metabolic indicators. Behavioral characteristics may include objective / semi-objective behavioral indicators such as sleep, diet, activity level, impulsive behavior, and social behavior. Treatment data may include medical orders for antidepressants, antipsychotics, mood stabilizers, anti-anxiety medications, physical therapy, and psychotherapy. Psychological assessment data may include scores on various psychological scales for depression, anxiety, mania, personality, cognitive function, and social functioning. Follow-up data may include longitudinal follow-up time points, changes in indicators, treatment adherence, symptom fluctuations, adverse events / abnormal events, etc.

[0030] In some embodiments, the feature construction module 11 may include a cleaning and standardization unit, used to perform preprocessing operations such as missing value imputation, outlier handling, dimensional unification, and time alignment based on the initial feature object, generating a standardized feature vector with fixed dimensions; then, according to a preset mapping relationship, the symptom features, vital signs, biomarkers, and behavioral features in the standardized feature vector are weighted and combined to construct a psychopathophysiological feature vector that can comprehensively reflect the patient's mental state. The psychopathophysiological feature vector is a multi-dimensional vector, with each dimension corresponding to psychopathophysiological dimensions such as emotion, anxiety, agitation, inflammatory state, metabolic state, or neuroendocrine state.

[0031] For example, the feature construction module 11 performs field parsing and unified mapping on the input data, mapping data fields from different sources and with different naming methods to a preset indicator set; at the same time, it unifies the units of execution for the test indicators, adds timestamps to time data, and identifies missing data. For textual symptom descriptions, the feature construction module 11 can extract keywords and convert them into structured features, forming a symptom feature vector. :

[0032] in, This indicates whether the corresponding symptom has occurred, with a value of 0 or 1. A value of 1 indicates that the symptom has occurred, and a value of 0 indicates that the symptom has not occurred. Together, they form a binary symptom feature vector, which is used to construct subsequent psychopathophysiological feature vectors. For indicator functions, when The function takes the value 1 when the condition is true and 0 when the condition is false. For the first A set of pre-defined symptom keywords, such as "depressed mood," "poor sleep," "anxiety," "loss of interest," or "fatigue." These keywords are symptom features extracted from clinical knowledge bases or historical medical records, forming a fixed symptom dictionary. For at a certain point in time The acquired data consists of textual descriptions of patient symptoms. This collection may include chief complaints, summaries of present medical history, progress notes, and textual descriptions from psychological assessment scales. That is to say, the first Does a preset symptom keyword appear in the patient's current text symptom description set? If it appears, the indicator function outputs 1; otherwise, it outputs 0.

[0033] Building upon this, the feature construction module 11 can map data fields from different sources and with different naming conventions to a preset indicator set through field parsing and unified mapping operations. It standardizes the units of measurement for the test indicators, such as unifying blood glucose levels to millimoles per liter and blood pressure to millimeters of mercury. It adds timestamps to time data, recording the collection time of each data point. It identifies missing data, marking which fields lack valid values. For text-based symptom description data, the feature construction module 11 extracts preset symptom keywords and converts them into structured features, forming a symptom feature vector.

[0034] By unifying and merging all feature items from demographic data, clinical history data, vital sign data, laboratory indicator data, behavioral characteristic data, psychological assessment data, follow-up data, and symptom feature vectors, an initial feature object is constructed:

[0035] in, is any original or parsed candidate feature term, representing the total number of features. It should be understood that the above formula is only used to illustrate that the system unifies multi-source heterogeneous information into the same patient feature space, and does not limit the specific number of features.

[0036] The feature construction module 11 contains a stratified correction unit, which maps patients to at least one reference stratification group based on demographic or clinical background information. The reference stratification group is constructed based on at least one combination of age range, gender, body mass index range, sampling period, smoking and drinking status, long-term drug exposure status, and significant comorbidities. For each indicator to be corrected in the initial feature object, the stratified correction unit retrieves the corresponding reference parameter from the parameter table or pre-defined stratification rules and performs baseline correction on the original indicator value. Specifically, for each feature item, the stratified correction unit calls the corresponding stratified reference parameter for standardization processing to generate a corrected feature object. :

[0037] in, The patient's stratification group. These are the original indicator values. and These respectively indicate the index in the stratified group. The reference mean and discrete parameters are as follows. This is a minimal constant to prevent the denominator from being zero. For indicators susceptible to comorbidities (including inflammation, metabolism, endocrine, liver and kidney function, and anemia-related indicators), the stratified correction unit further performs bias correction to make the corrected indicators closer to the target representation required for mental state assessment. The stratified correction unit ultimately outputs the corrected feature object. .

[0038] The feature construction module 11 also includes a cleaning and standardization unit, which performs missing value imputation, outlier handling, unit unification, and time alignment operations on the corrected feature object to generate the standardized feature vector. , It is a multidimensional vector composed of all preprocessed feature values ​​arranged in a fixed order.

[0039] In outlier handling, the system identifies outliers based on medically permissible ranges, distribution thresholds, contextual conflicts, or historical longitudinal variation magnitudes. Measurements exceeding physiologically reasonable ranges are truncated, marked, removed, or reverted to the most recent reliable value. For missing data imputation, rule-based imputation is used for static demographic fields, while nearest neighbor time points, sliding window mean, or trend interpolation are used for time-series physiological indicators. Fields that cannot be reliably inferred are marked as missing and incorporated into subsequent quality weights. In dimensional unification, the system performs a unified scale transformation on indicators with different dimensions, enabling features such as blood pressure, uric acid, white blood cell count, pulse, and symptom frequency to be included in the same computational framework. In time alignment, the system aggregates data from different time points according to a set time window, forming a fixed-dimensional vector corresponding to a specific evaluation time.

[0040] For indicators collected at a single time point (such as static demographic information or a single test result), the final value is denoted as... subscript Indicates the first Each characteristic indicator. Therefore, for One of the components in the normalized feature vector. Each element in the normalized feature vector can be represented as:

[0041] For features with data from multiple time points, such as pulse, blood pressure, and symptom scores in longitudinal follow-ups, the cleaning and standardization units aggregate data within a set time window.

[0042] in, To assess the time point The aggregated value is the time window centered or ending at. As this indicator at the evaluation time The representative value becomes One of the components.

[0043] The cleaning and standardization unit ultimately outputs a standardized feature vector. , It is a multidimensional vector composed of all preprocessed feature values ​​arranged in a fixed order. Specifically, Each component in the equation corresponds to a specific indicator (such as age, uric acid level, symptom keyword "depressed mood"), and the value of that component is the final value obtained after missing value imputation, outlier handling, unit unification, and time alignment.

[0044] The feature construction module 11 also includes a psychopathophysiological mapping unit. The psychopathophysiological mapping unit inputs the standardized feature vector into the feature mapping model and fuses symptom features, vital signs, biomarkers and behavioral features to construct psychopathophysiological feature vectors.

[0045] For example, feature mapping may include: mapping symptom features to psychological dimensions such as emotion, anxiety, and agitation; mapping vital signs and laboratory indicators to inflammation, metabolic, and neuroendocrine states; and mapping medical history and behavioral characteristics to disease stability and risk predisposition. Each psychopathophysiological dimension is obtained through a weighted combination:

[0046] in, It is the first A psychopathophysiological dimension It is a fundamental feature Mapping weights for this dimension This is the bias term. All dimensions together form the comprehensive feature vector. The psychopathophysiological mapping unit outputs the psychopathophysiological feature vector. .

[0047] Furthermore, the feature construction module 11 also includes a psychological abnormality risk index calculation unit. The psychological abnormality risk index calculation unit is based on psychopathophysiological feature vectors. Calculating the risk index of psychological abnormality The psychological abnormality risk index is used to describe the overall risk level of an individual currently in an abnormal mental state. The formula for calculating the psychological abnormality risk index is:

[0048] In the formula, For compression functions, such as the Sigmoid function, the output value is limited to between 0 and 1; It is the first Weighting coefficients for each psychopathophysiological dimension; This is a bias term. This psychological abnormality risk index can be used directly as an assessment output, or as one of the input features for subsequent main type identification, subtype judgment, and prognostic assessment.

[0049] Optionally, the feature construction module 11 may further include a feature storage unit. The feature storage unit stores psychopathological and physiological feature vectors. and psychological abnormality risk index The data is stored in a structured format, and each stored record is associated with a timestamp and a data source identifier. The timestamp identifies the point in time corresponding to the evaluation, and the data source identifier records the device or system from which the data comes (such as a medical information system, wearable device, follow-up terminal, etc.), so as to trace the historical status in subsequent dynamic monitoring and closed-loop updates.

[0050] In some optional embodiments, the main type hierarchical evaluation module 12 may include a hierarchical model loading unit, a main type inference unit, a confidence scoring unit, and a label storage unit.

[0051] The hierarchical model loading unit is used to load pre-trained hierarchical evaluation models or rule-based models. This model can be a machine learning classification model trained on a large number of samples (e.g., a softmax multi-classifier), or a decision function composed of multiple threshold logics and weight rules. The loaded model is then used for subsequent discrimination of psychopathological feature vectors. The main type inference unit is used to calculate the matching probability or score of different emotional disorder main types based on psychopathophysiological feature vectors. Specifically, it uses the psychopathophysiological feature vectors generated by the feature construction module. The data is input into the classification model loaded by the hierarchical model loading unit. The model targets each candidate major type of mood disorder. (For example, depression spectrum, bipolar spectrum, post-traumatic stress disorder spectrum, generalized anxiety disorder spectrum, panic attack spectrum, social anxiety spectrum) Output a matching probability. This probability can be calculated using the following softmax formula:

[0052] in, It is the input psychopathophysiological feature vector. Indicates the first Individual type tags, It is the first The discriminant function value (usually a linear discriminant function) corresponding to each main type is calculated. The denominator is the sum of the discriminant function values ​​of all candidate main types after taking the exponent, ensuring that the sum of the matching probabilities of all main types is 1. Based on the above calculation results, the main type inference unit selects the main type with the highest matching probability as the output, i.e., the main type label. .

[0053] The confidence score unit is used to calculate the main type confidence score based on the output probability distribution of each main type. The main type confidence score reflects the reliability of the inference result and is typically taken as the maximum matching probability value itself. For example, if the matching probability for the depression spectrum is 0.85, for the bipolar spectrum it is 0.10, and for other spectrums combined it is 0.05, then the primary type label is depression spectrum, with a confidence level of 0.85. This confidence level can be subsequently used for adjusting the weights of intervention strategies or determining early warning conditions.

[0054] Tag storage units are used for hierarchical results of the main type (i.e., main type tags). The data, along with the primary type confidence level, is structured and stored, and then output to the system's evaluation output module or dynamic monitoring module. During storage, the current evaluation timestamp and data source are associated to trace the historical primary type changes during closed-loop updates.

[0055] In some optional embodiments, the subtype segmentation identification module 13 includes a subspace modeling unit, a subtype inference unit, and a subtype confidence scoring unit.

[0056] The subspace modeling unit is used to extract feature dimensions related to the main type label from the psychopathophysiological feature vector under the constraint of the main type label, and construct a feature subspace. This feature subspace can use only the feature dimensions related to the current main type, or it can combine the full feature vector with the main type-specific weights for remapping, thereby reducing the interference of irrelevant features and improving the targeting of subtype identification.

[0057] The subtype inference unit is used to calculate the distance or probability between the patient and multiple pre-defined subtype centers within the feature subspace, and to determine the subtype label based on the minimum distance or maximum probability. The pre-defined subtype centers are typical feature vectors pre-clustered within the main type according to multiple dimensions such as symptom structure, comorbidity pattern, biomarker pattern, treatment history, adherence characteristics, and disease course characteristics. Specifically, the subtype inference unit can use the center distance method to calculate the psychopathophysiological feature vector. With the main type Under the conditions Subtype Center Calculate the squared Euclidean distance between them, and select the subtype with the smallest distance as the output:

[0058] Alternatively, subtype reasoning units can be calculated probabilistically within the main type. Psychopathological and physiological feature vector under constraints Belongs to the The conditional probability of each subtype:

[0059] in It is aimed at the first The discriminant function for each subtype, It represents the total number of subtypes, and selects the subtype with the highest probability as the output.

[0060] The subtype confidence score unit is used to calculate the degree of matching corresponding to the subtype label as the subtype confidence score. When using the center distance method, the subtype confidence score can be the inverse of the distance or a similarity score based on the distance mapping; when using the probability method, the subtype confidence score is directly taken as the maximum conditional probability value. The subtype segmentation and identification module 13 outputs subtype labels. The confidence scores of the corresponding subtypes are used for subsequent intervention strategy generation and prognostic assessment.

[0061] In some embodiments, the evaluation output module 14 may include an intervention strategy generation unit, which generates intervention strategy logic rules based on at least one of the main type label, subtype label, and psychopathophysiological feature vectors. The intervention strategy can be generated based on the current main type, current subtype, psychological abnormality risk index, key contributing factors, previous intervention history, recent response status, medication-related risks, and comorbidity constraints. The psychological abnormality risk index is an overall risk value between 0 and 1 output by the psychological abnormality risk index calculation unit; key contributing factors refer to several features that have the greatest impact on the current main type and subtype judgment, such as significantly elevated inflammatory markers or prominent anxiety symptoms; previous intervention history refers to the medications, physical or psychological treatment regimens and their durations previously received by the patient; recent response status refers to the patient's adherence to the most recent intervention strategy and the degree of symptom improvement; medication-related risks refer to the possibility of drug interactions, metabolic effects, or adverse reactions; and comorbidity constraints refer to the limitations imposed on intervention selection by the patient's coexisting physical diseases such as endocrine, metabolic, cardiovascular, or liver and kidney dysfunction.

[0062] The system can score multiple candidate strategies based on a preset rule base, matching function, or learning-based strategy model, and select the highest-scoring strategy as the current output strategy. Candidate strategies include at least one of the following: monitoring frequency adjustment (e.g., changing the follow-up period from monthly to weekly), indicator re-collection suggestions (e.g., suggesting the addition of thyroid function or inflammatory marker testing), and intervention priority adjustment (e.g., prioritizing treatment of agitation symptoms over core depressive symptoms). The strategy scoring function is expressed as follows:

[0063] in, This represents a candidate strategy; It is a psychological abnormality risk index; It is a primary type constraint item, used to determine the primary type label. Provide the fit score of the strategy at the main type level; for example, the score of the antidepressant strategy will be reduced for bipolar spectrum. It is a subtype constraint term, used to determine the subtype label. Adjusting strategy scores, for example, may increase the score of the mood stabilizer strategy for those with a mixed agitation tendency; It is the current feature vector trigger term, used to determine the psychopathophysiological feature vector. Key factors in the process activate specific strategies, such as recommending adjuvant anti-inflammatory therapy when inflammation-related dimensions exceed a threshold. It is a correction term for the strategy based on historical feedback. This represents a summary of past intervention history and recent response characteristics; if a patient has a history of ineffectiveness or intolerance to a certain type of medication, the corresponding strategy score is reduced. These are preset weighting coefficients that can be dynamically adjusted according to the clinical scenario.

[0064] After calculating the scores of all candidate policies, the system selects the policy with the highest score as the current output policy.

[0065] This strategy The intervention strategy is output in the form of structured instructions or natural language suggestions. The evaluation output module 14 transmits this intervention strategy as part of the mood disorder assessment results to the clinical terminal or follow-up system. At the same time, the version of the intervention strategy and the subsequent patient response are recorded and fed back to the rule parameter update unit for dynamic adjustment when the next strategy is generated.

[0066] In some alternative embodiments, the evaluation output module 14 may also include a prognostic index calculation unit for calculating a multi-time-window prognostic index based on at least one of the main type label, subtype label, or psychopathophysiological feature vector.

[0067] The prognostic index calculation unit calculates risk values ​​for multiple time windows to reflect the different possible evolutionary directions of the patient's disease course. Time windows include short-term (e.g., ≤1 year), medium-term (e.g., 1 to 3 years), and long-term (e.g., ≥3 years). Each time window corresponds to a different risk type, including at least one of the following: readmission risk, symptom exacerbation risk, main type drift risk, risk of conversion from unipolar depression to bipolar or psychotic disorder, risk of decreased adherence, risk of self-harm, or risk of high-risk events.

[0068] The system targets each time window Calculate a risk value:

[0069] in, It is a compression function (such as the Sigmoid function) that limits the output value to between 0 and 1; It is for time windows The weight parameter vector; It is composed of psychopathophysiological feature vectors Main type tag Subtype tags A comprehensive feature vector composed of the current strategy version and historical feedback; It's a bias term. Different parameters are used for different time windows. and This is because the driving factors of short-term and long-term risks may be different. For example, short-term risks are more affected by the severity of current symptoms and acute biomarkers, while long-term risks are more affected by disease course characteristics, comorbidity patterns and treatment adherence.

[0070] The prognostic index calculation unit compares the risk value at the current assessment time with that at the previous assessment time and calculates the trend:

[0071] in, and These represent the time of the previous assessment and the time of the current assessment, respectively. and These are the risk values ​​for the corresponding time windows. If This indicates an overall increase in risk; if This indicates an overall decrease in risk; combining multiple time points allows for further calculation of fluctuation amplitude and stability coefficients. Based on this, the system outputs risk assessment results and trends with clinical value. Examples include "high short-term risk but controllable in the long term," "currently stable but with increased medium- to long-term risk of transformation," or "overall decrease, but monitoring of a specific subtype risk is still necessary." These risk assessment results, as part of the mood disorder assessment results, are output by the assessment output module 14 to the clinical terminal or follow-up system, and can generate risk warning signals when preset conditions are met.

[0072] In some optional embodiments, the closed-loop update module 15 includes a re-collection triggering unit, a state fusion unit, and a re-evaluation scheduling unit.

[0073] The re-collection trigger unit is used to detect whether updated patient data has been received. Updated data refers to at least one of the following generated during subsequent follow-ups: new outpatient records, inpatient records, laboratory results, vital signs, changes in chief complaints, treatment feedback, or follow-up data. When the re-collection trigger unit detects updated data, the system automatically sends the new data to the feature construction module 11, re-executes the hierarchical correction, cleaning and standardization, and psychopathophysiological feature vector construction process, and obtains a new input feature vector generated based on the updated data. .

[0074] The state fusion unit is used to weight and fuse the psychopathophysiological feature vector generated in the previous round of evaluation with the new input feature vector according to the weights retained in the historical state, to generate an updated psychopathophysiological feature vector. Specifically, the state fusion unit performs the following fusion operation:

[0075] in, This represents the psychopathophysiological feature vector generated in the previous round of assessment. This represents the updated psychopathophysiological feature vector obtained after fusion. The historical state is preserved by weights, with values ​​ranging from 0 to 1. This fusion method avoids the system from overreacting to single, occasional fluctuations, while gradually reflecting trends as the patient's condition continues to change, thus preserving the continuity of the disease course.

[0076] The re-evaluation scheduling unit is used to update the psychopathophysiological feature vector. The data is sequentially sent to the main type hierarchical evaluation module 12 and the subtype subdivision identification module 13, triggering a re-evaluation process. Specifically, this can be achieved by... Input the main type hierarchical evaluation module 12, which recalculates the matching probability and outputs the updated main type label. and the corresponding confidence level; use the updated main type label as a constraint, and... The input is to the subtype segmentation and identification module 13, which recalculates the subtype and outputs the updated subtype label. And the corresponding subtype confidence scores. Based on this, the intervention strategy rule engine module and the prognosis and trend assessment module are further invoked to regenerate the intervention strategy based on the updated principal type, subtype, and psychopathophysiological feature vectors. and prognostic risk value The entire recalculation process described above can be uniformly represented as:

[0077] in It represents a complete mapping function from psychopathophysiological feature vectors to main type, subtype, prognostic risk and intervention strategy.

[0078] When the recalculation results meet the preset warning conditions, a warning output can be triggered. The preset warning conditions may include one or more of the following: a rapid increase in short-term risk, a subtype changing from low-risk to high-risk, the simultaneous occurrence of multiple risk factor combinations, or the continuous ineffectiveness of historical intervention strategies. When the warning conditions are met, the closed-loop update module 15 sends corresponding prompt signals to the medical terminal, follow-up terminal, or management terminal through the evaluation output module 14 or the dynamic monitoring module 16, thereby realizing a complete dynamic closed-loop update process from data re-collection, status fusion, re-evaluation to warning feedback.

[0079] It should be understood that when the various modules of the system provided in the above embodiments are working, the division of each functional module in the above description is only used as an example. In actual applications, the above functions can be assigned to different functional modules as needed. That is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0080] The functional modules in the above embodiments 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. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of the embodiments of this application.

[0081] The following is an example of the practical application of the mood disorder full-cycle assessment system provided in this application. In the first specific embodiment of the system assessment process, the mood disorder full-cycle assessment system receives multi-source data from a patient, including: age 35 years, female, no history of major physical illness; clinical test results including elevated uric acid level and slightly elevated blood pressure; vital signs showing a rapid pulse; chief complaint text including "depressed mood, poor sleep, and anxiety"; and one follow-up record.

[0082] The full-cycle assessment system for mood disorders performs field parsing and mapping on the aforementioned data, transforming data from different sources into a unified structured feature set X. Subsequently, based on demographic information, the system categorizes patients into the "middle-aged and young women group," calls the corresponding stratified reference parameters, and performs baseline correction on the original features to obtain corrected features X'. Building upon this, the system performs data cleaning and standardization on the corrected features, and aggregates data from multiple time points within a set time window to generate a standardized feature vector Z.

[0083] Next, the full-cycle assessment system for mood disorders inputs the feature vector into the psychopathophysiological feature construction module, which fuses and calculates symptom characteristics, biometrics, and vital signs to generate a psychopathophysiological feature vector H and calculates the psychological abnormality risk index R. Then, the system inputs the feature vector H into the disease stratification assessment module to calculate the matching probability of different mood disorder main types, determining that the patient belongs to the "depression spectrum" and outputting the corresponding confidence level. Based on this, the system performs subtype subdivision identification within the "depression spectrum" subspace, further identifying the patient as "high anxiety with sleep disorders." Subsequently, the system generates intervention strategies based on the main type label, subtype label, and risk index, including: increasing follow-up frequency, focusing on monitoring sleep indicators and anxiety-related symptoms, and outputting the corresponding strategy results. The full-cycle assessment system for mood disorders further calculates short-term and medium-term risks and generates risk trend values, determining that the patient has an upward risk trend in the short term.

[0084] When the full-cycle assessment system for mood disorders receives new follow-up data, it updates the feature vector and re-executes the above assessment process to achieve dynamic closed-loop monitoring.

[0085] In the second specific embodiment regarding system dynamic follow-up and risk evolution, when a patient first visits, the full-cycle assessment system for mood disorders receives their basic data, including: male, 28 years old, chief complaint of "depressed mood, loss of interest, and fatigue," vital signs are basically normal, and laboratory indicators are not significantly abnormal. After processing the data, the full-cycle assessment system for mood disorders generates a psychopathophysiological feature vector and stratifies it into a "depression spectrum," identifying the subtype as "typical depressed type," with a risk index at a moderate level. Based on the current assessment results, the full-cycle assessment system for mood disorders generates an initial intervention strategy, including routine monitoring and periodic follow-up.

[0086] Two weeks later, the full-cycle assessment system for mood disorders received new data, including changes in the chief complaint to "mood fluctuations, irritability, and reduced sleep," and also detected elevated pulse and abnormalities in some metabolic indicators.

[0087] The full-cycle assessment system for mood disorders will integrate and update new data with historical characteristics:

[0088] A re-administered stratified assessment revealed that the primary type remained "depressive spectrum," but the subtype shifted to "mixed agitation tendency," with a significant increase in the risk index. Based on this, the mood disorder full-cycle assessment system updated its intervention strategy, increasing monitoring frequency and focusing more intently on agitation and sleep changes.

[0089] Furthermore, during subsequent follow-up, the emotional disorder full-cycle assessment system continuously receives new data and updates the risk assessment results, calculates the risk values ​​and trend changes under multiple time windows, and triggers an early warning if the risk continues to rise.

[0090] Therefore, the mood disorder full-cycle assessment system provided in this application can first determine the patient's primary type, and then identify subtypes within the corresponding feature subspace using the primary type as a constraint, forming a coarse-to-fine two-layer diagnostic structure. This can distinguish between different mood disorders with overlapping symptoms, and further subdivide clinical manifestations within the same primary type, effectively reducing the misdiagnosis rate and supporting individualized intervention. Furthermore, a closed-loop update is automatically triggered when the monitoring period ends or new data is received, re-inputting the updated features into the hierarchical diagnostic process, achieving a complete closed loop from initial assessment to continuous tracking, and from static output to dynamic iteration. The mood disorder full-cycle assessment system provided in this application, by integrating hierarchical diagnosis and a full-cycle closed loop, can improve the accuracy of mood disorder assessment.

[0091] Based on the same concept, this application also provides a method for assessing the entire lifecycle of mood disorders; please refer to [link / reference]. Figure 2 , Figure 2 A schematic diagram illustrating the steps of the full-cycle assessment method for mood disorders provided in this application embodiment. The method may include the following steps: S1. Generate a standardized feature vector based on the received initial feature objects of the patient, and construct a psychopathophysiological feature vector based on the standardized feature vector.

[0092] S2. Input the psychopathophysiological feature vector into the preset classification model or rule framework, calculate the matching probability of the patient belonging to each of the multiple candidate emotional disorder main types, and output the main type label with the highest matching probability and the corresponding main type confidence.

[0093] S3. Using the main type label as a constraint, identify the patient's subtype based on the psychopathophysiological feature vector within the feature subspace corresponding to the main type label, and output the subtype label and the corresponding subtype confidence.

[0094] S4. After receiving the patient's updated data, generate an updated psychopathophysiological feature vector based on the updated data, and generate an updated main type label and an updated subtype label based on the updated psychopathophysiological feature vector.

[0095] S5. Generate mood disorder assessment results based on at least one of the main type label, subtype label, updated main type label, and updated subtype label.

[0096] S6. When the monitoring period arrives or when updated patient data is received, a closed-loop update mechanism is triggered, and the steps of re-executing the mood disorder assessment are controlled, as well as the closed-loop communication feedback results are output.

[0097] For specific instructions on using the above methods to assess mood disorders, please refer to the detailed explanation of the full-cycle assessment system for mood disorders in the above content, which will not be repeated here.

[0098] Based on the same concept, embodiments of this application also provide a computer device, which may include 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 methods described above.

[0099] Based on the same concept, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0100] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A full-cycle assessment system for mood disorders, characterized in that, include: The feature construction module is used to generate standardized feature vectors based on the received initial feature objects of the patient, and to construct psychopathophysiological feature vectors based on the standardized feature vectors; wherein, the initial feature objects include a unified feature set formed by multi-source data such as demographic information, clinical history data, text-based symptom description data, vital sign data, routine clinical blood test indicators, behavioral feature data, treatment data, psychological assessment data, and follow-up data after field parsing, unit unification, and missing labeling; The main type hierarchical assessment module is used to input the psychopathophysiological feature vector into a preset classification model or rule framework, calculate the matching probability of the patient belonging to each of the multiple candidate main types of emotional disorders, and output the main type label with the highest matching probability and the corresponding main type confidence. The subtype segmentation and identification module is used to identify the subtype to which the patient belongs in the feature subspace corresponding to the main type label, based on the psychopathophysiological feature vector, using the main type label as a constraint, and output the subtype label and the corresponding subtype confidence. The closed-loop update module is used to generate an updated psychopathophysiological feature vector based on the updated data after receiving the patient's updated data, and to re-input the updated psychopathophysiological feature vector into the main type stratification assessment module and the subtype subdivision identification module to generate updated main type labels and updated subtype labels; wherein, the updated data refers to at least one of the following: new outpatient records, inpatient records, test results, vital signs, changes in chief complaints, treatment feedback, or follow-up data generated by the patient in subsequent follow-ups; An assessment output module is used to generate an emotional disorder assessment result based on at least one of the main type label, the subtype label, the updated main type label, and the updated subtype label; The dynamic monitoring module is used to trigger the closed-loop update module when the monitoring cycle arrives or when updated data of the patient is received, and to control the feature construction module, the main type hierarchical assessment module, and the subtype subdivision identification module to be re-executed, and to output closed-loop communication feedback results.

2. The full-cycle assessment system for mood disorders according to claim 1, characterized in that, The feature construction module includes: The stratification correction unit is used to map the patient to a reference stratification group based on demographic information or clinical background information, and to perform baseline correction on the target indicators in the initial feature object according to the reference stratification group to generate a corrected feature object; The cleaning and standardization unit is used to perform missing value filling, outlier handling, unit unification and time alignment operations on the corrected feature object to generate the standardized feature vector. The psychopathophysiological mapping unit is used to weight and combine the symptom features, vital signs, biomarkers and behavioral features in the standardized feature vector according to a preset mapping relationship to generate the psychopathophysiological feature vector.

3. The full-cycle assessment system for mood disorders according to claim 1, characterized in that, The feature construction module also includes: A psychological abnormality risk index calculation unit is used to calculate the psychological abnormality risk index based on the aforementioned psychopathophysiological feature vector. The feature storage unit is used to structurally store the psychopathophysiological feature vector and the psychological abnormality risk index, and associate them with time tags and data source identifiers.

4. The full-cycle assessment system for mood disorders according to claim 1, characterized in that, The main type hierarchical evaluation module includes: The hierarchical model loading unit is used to load pre-trained hierarchical evaluation models or rule models. The main type inference unit is used to calculate the matching probability or score of different main types of emotional disorders based on the psychopathophysiological feature vector. A confidence scoring unit is used to calculate the confidence level of the main type based on the output probability distribution of each main type. The tag storage unit is used to store and output the hierarchical results of the main type and the confidence level of the main type.

5. The full-cycle assessment system for mood disorders according to claim 1, characterized in that, The subtype segmentation identification module includes: The subspace modeling unit is used to extract feature dimensions related to the main type label from the psychopathophysiological feature vector under the constraint of the main type label, and construct the feature subspace; The subtype inference unit is used to calculate the distance or probability between the patient and multiple preset subtype centers within the feature subspace, and to determine the subtype label based on the minimum distance or the maximum probability. The subtype confidence score unit is used to calculate the degree of matching corresponding to the subtype label as the subtype confidence score.

6. The full-cycle assessment system for mood disorders according to claim 1, characterized in that, The closed-loop update module includes: A re-acquisition triggering unit is used to detect whether the updated data has been received; The state fusion unit is used to weight and fuse the psychopathophysiological feature vector generated in the previous round of evaluation with the new input feature vector generated based on the updated data according to the historical state retention weight, so as to generate the updated psychopathophysiological feature vector. The re-evaluation scheduling unit is used to sequentially send the updated psychopathophysiological feature vectors into the main type hierarchical evaluation module and the subtype subdivision identification module.

7. The full-cycle assessment system for mood disorders according to claim 1, characterized in that, The dynamic monitoring module includes: A monitoring cycle control unit is used to receive and store the monitoring cycle; A re-collection triggering unit is used to generate a trigger signal when the monitoring period arrives or when the updated data is received; The re-evaluation scheduling unit is used to call the closed-loop update module according to the trigger signal, and sequentially schedule the feature construction module, the main type hierarchical evaluation module and the subtype subdivision identification module to be re-executed; A closed-loop communication feedback unit is used to output the closed-loop communication feedback result, which includes at least one of the updated main type label, the updated subtype label, and the mood disorder assessment output information.

8. A method for assessing mood disorders throughout their entire lifecycle, characterized in that, include: A standardized feature vector is generated based on the initial feature objects of the received patients, and a psychopathophysiological feature vector is constructed based on the standardized feature vector; wherein, the initial feature objects include a unified feature set formed by multi-source data such as demographic information, clinical medical history data, text-based symptom description data, vital sign data, routine clinical blood test indicators, behavioral feature data, treatment data, psychological assessment data, and follow-up data after field parsing, unit unification, and missing data marking; The psychopathophysiological feature vector is input into a preset classification model or rule framework to calculate the matching probability of the patient belonging to each of the multiple candidate emotional disorder main types, and outputs the main type label with the highest matching probability and the corresponding main type confidence. Using the main type label as a constraint, within the feature subspace corresponding to the main type label, the subtype to which the patient belongs is identified based on the psychopathophysiological feature vector, and the subtype label and the corresponding subtype confidence are output. After receiving the patient's updated data, an updated psychopathophysiological feature vector is generated based on the updated data, and an updated main type label and an updated subtype label are generated based on the updated psychopathophysiological feature vector; wherein, the updated data refers to at least one of the following: new outpatient records, inpatient records, test results, vital signs, changes in chief complaints, treatment feedback, or follow-up data generated by the patient in subsequent follow-ups; Based on at least one of the main type label, the subtype label, the updated main type label, and the updated subtype label, generate an emotional disorder assessment result; When the monitoring period arrives or when updated data for the patient is received, a closed-loop update mechanism is triggered, and the steps of re-executing the mood disorder assessment are controlled, as well as the output of closed-loop communication feedback results.

9. A computer device, characterized in that, The computer device includes 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 of claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in claim 8.

Images