Sleep-mood-somatic health linkage risk assessment and intervention system and method

Abstract

The application discloses a sleep-emotion-somatic health linkage risk assessment and intervention system and method, relates to the technical field of medical informatization, and comprises the following steps: acquiring sleep monitoring, emotion evaluation and somatic health diagnosis dimension multi-source heterogeneous data, standardizing the multi-source heterogeneous data of each dimension, and acquiring corresponding target data flow; extracting associated characteristic indexes according to the target data flow of each dimension, and generating a sleep-emotion-somatic health association information triad graph according to the associated characteristic indexes; constructing a risk assessment model based on the information triad graph, user basic information and a clinical knowledge graph, performing clinical risk judgment on real-time multi-source heterogeneous data through the risk assessment model, outputting a clinical risk result and generating specific warning content and intervention suggestions; effectively solving the problem of intervention lag, improving the time series data processing efficiency, retaining only abnormal associations with clinical significance, and improving the risk identification accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical informatization, in particular to a sleep-emotion-physical health linkage risk assessment and intervention system and method. BACKGROUND

[0002] Currently, the management of sleep, emotion and physical health problems of psychiatric patients usually adopts the following several independent schemes: using intelligent mattress or bracelet and other devices, but the data of which is usually only displayed in the App of the device itself, reporting "sleep quality score" or "low blood oxygen event", lacking clinical interpretation. In the outpatient or regular follow-up, the doctor or nurse uses the GDS scale for manual assessment, and the results are recorded in the electronic medical record. This is a static and point-like assessment. The nursing staff records the patient's vital signs, complaints of discomfort (such as pain, shortness of breath) in the nursing record, which is in a different system from the sleep and emotion data.

[0003] However, sleep data, emotion scale data, and nursing record data are stored in different devices and information systems, with different formats and no connection, which cannot form a unified and overall view of the patient's health status, resulting in that the conventional physical health assessment is limited to a single dimension. For example, the bracelet reports "night sleep interruption", but cannot automatically associate it with whether it is caused by apnea (physical problem), and the resulting daytime emotional depression (emotional problem), which cannot realize the cross deep analysis of "sleep-emotion-physical" three-dimensional strong association, and relies on manual discovery of abnormal association, which is low in efficiency and easy to miss, and usually the problem has already significantly affected the patient's health, so that early intervention and the golden period are missed.

[0004] The patent "Chinese medicine intelligent diagnosis and treatment and health management system based on big data", publication number: CN120376162A, discloses a system including a multi-source data acquisition and integration unit, a big data analysis and mining unit, and an intelligent diagnosis and treatment and health management service unit. The built-in functions of the multi-source data acquisition and integration unit include data sources and data integration, wherein the data sources include clinical medical record data, health monitoring data, ancient medical literature data, and modern medical test data, and the data integration includes data cleaning, data standardization, and data fusion. The built-in functions of the big data analysis and mining unit include disease and syndrome type association analysis, prescription efficacy evaluation and optimization, personalized health management analysis, and drug research and development auxiliary analysis, wherein the operation steps of disease and syndrome type association analysis include feature extraction, machine learning modeling, model evaluation and optimization. However, this scheme mainly fuses multi-source data to construct a model and then diagnoses the patient's condition. Due to the difference between human bodies, the amount of data to be obtained is complex and massive. If the data quality is to be improved, the data analysis efficiency will be reduced, resulting in that the model cannot balance the data analysis accuracy and efficiency, and early intervention still cannot be realized. SUMMARY

[0005] The purpose of the present application is to address the problem that conventional medical information systems cannot identify abnormal associations and it is difficult to balance the analysis accuracy and efficiency of multi-source data fusion, resulting in unreliable analysis results and delayed intervention. The present application proposes a sleep-emotion-physical health linkage risk assessment and intervention system and method, which obtains multi-dimensional data of sleep, emotion and physical health status, analyzes the risk of user's physical health status through fusion association, and generates corresponding intervention suggestions to realize multi-dimensional abnormal association analysis, avoid intervention delay, and improve the efficiency of time series data processing by combining clinical knowledge graph construction model, while only retaining abnormal associations with clinical significance, and improving the accuracy of risk identification.

[0006] To achieve the above purpose, the technical solutions adopted by the embodiments of the present application are as follows:

[0007] In a first aspect, the embodiments of the present application provide a sleep-emotion-physical health linkage risk assessment and intervention method, which comprises:

[0008] Obtaining multi-source heterogeneous data of sleep monitoring, emotion evaluation and physical health diagnosis dimensions, standardizing the multi-source heterogeneous data of each dimension, and obtaining corresponding target data stream;

[0009] Extracting association feature indexes according to the target data stream of each dimension, and generating an information triad graph of sleep-emotion-physical health association according to the association feature indexes;

[0010] Constructing a risk assessment model based on the information triad graph, user basic information and clinical knowledge graph, performing clinical risk judgment on real-time multi-source heterogeneous data through the risk assessment model, outputting clinical risk results and generating specific warning content and intervention suggestions.

[0011] In the scheme, the data of sleep, emotion and physical health are processed and fused to generate a triad graph, the originally scattered and heterogeneous data is integrated into a unified health data view with the patient as the center and time as the axis, so that medical staff can efficiently evaluate the patient's state from three strongly correlated dimensions, comprehensively and holistically; by constructing a risk identification model based on the correlation of characteristic indicators and clinical knowledge graph, the traditional and passive "problem occurs, problem is solved" mode is changed into an active monitoring, risk assessment and early warning mode, which can timely warn before serious symptoms occur or in the early stage, gain valuable "time window" for clinical intervention, and solve the problem that traditional methods rely on the personal experience of medical staff, which is easy to miss the complex correlation of multi-factor, non-linear and sometimes lag, resulting in unreliable analysis results and delayed intervention. In addition, by analyzing the time series data of sleep, emotion and body, identifying specific abnormal correlation patterns between them, and combining the clinical knowledge graph model, the cross-period correlation and only clinically meaningful abnormal correlation features can be accurately captured, improving the efficiency of cross-dimensional data parallel fusion and the accuracy of risk identification.

[0012] Preferably, the multi-source heterogeneous data of sleep monitoring, emotion evaluation and physical health diagnosis dimensions is obtained, and the multi-source heterogeneous data of each dimension is standardized to obtain corresponding target data streams, including:

[0013] The first health monitoring data and the second health monitoring data of the user's sleep cycle are collected, and the target sleep monitoring data is calculated by fusing the first health monitoring data and the second health monitoring data;

[0014] Based on the user's electronic medical record, the state evaluation scale is extracted, analyzed and abnormally labeled to obtain structured target emotion evaluation data;

[0015] Based on the user's nursing information, the key entity recognition is performed according to the trained medical named entity recognition model, and the identified key entity is standardized and symptom attribute relationship mapped to form target physical health data.

[0016] Preferably, the first health monitoring data and the second health monitoring data of the user's sleep cycle are collected, and the target sleep monitoring data is calculated by fusing the first health monitoring data and the second health monitoring data, including:

[0017] The first health monitoring data of the user is obtained based on the first monitoring device, including at least heart rate data, blood oxygen saturation data and body movement data;

[0018] The heart shock signal of the specific human body region of the user is obtained based on the second monitoring device, and the respiration signal and heartbeat signal are obtained after signal preprocessing of the heart shock signal;

[0019] Fusion heart rate data, blood oxygen saturation data, body motion data, respiratory signal and heartbeat signal to calculate heart rate index, respiratory rate index and body motion index, and determine the sleep state type according to the index data;

[0020] The heart rate index, respiratory rate index, body motion index and sleep state type are packaged in a preset data frame format to obtain standardized target sleep monitoring data.

[0021] Preferably, the state evaluation scale based on the user electronic medical record is extracted, and the state evaluation scale is analyzed and abnormally marked to obtain structured target emotional evaluation data, including:

[0022] Based on the first communication link, the mental state evaluation scale of the user in the electronic medical record system or / and the psychological evaluation system is obtained;

[0023] Based on the mental state evaluation scale, at least the key evaluation field including the evaluation type, the evaluation name and the score thereof, and the evaluation date is extracted;

[0024] The key evaluation field is subjected to outlier check, and the outliers are abnormally marked or trigger the re-collection of the mental state evaluation scale. Based on the key evaluation field after the check, a unified structured format mapping is performed to obtain the target emotional evaluation data of the user.

[0025] Preferably, the key evaluation field extracted based on the mental state evaluation scale includes at least the evaluation type, the evaluation name and the score thereof, and the evaluation date, and further includes:

[0026] If the mental state evaluation scale is in the form of unstructured text, the mental state evaluation scale is subjected to semantic analysis based on natural language technology to obtain initial semantic information;

[0027] Based on the initial semantic information and the key evaluation information library, similarity matching is performed to extract the key evaluation field.

[0028] Preferably, based on the user nursing information, key entity recognition is performed according to the trained medical named entity recognition model, and the recognized key entity is subjected to standardization processing and symptom attribute relationship mapping to form target physical health data, including:

[0029] Based on the second communication link, the nursing course information in the hospital information system is obtained. For structured nursing course information, key entity data is extracted, including at least vital signs, medication records, laboratory test results and nursing assessment items. For unstructured nursing course information, medical named entity recognition model is used for key entity recognition, and the key entity recognition includes at least symptom entity, sign entity and treatment measure entity;

[0030] After the identified symptom entities, sign entities and treatment measure entities are normalized by standard medical terminology, a symptom-attribute relationship pair is constructed and entity data mapping is performed to obtain symptom attribute structured data;

[0031] The key entity data and the symptom attribute structured data constitute target physical health data.

[0032] Preferably, the target data flow according to each dimension is extracted to obtain a correlation characteristic index, and a sleep-emotion-physical health correlation information triad graph is generated according to the correlation characteristic index, including:

[0033] The target sleep monitoring data, the target emotion evaluation data and the target physical health data are spatiotemporally aligned;

[0034] The correlation characteristic index is extracted based on the spatiotemporally aligned target data flow of each dimension, and the correlation characteristic index is data-visualized to generate an information triad graph in which sleep monitoring, emotion evaluation and vital sign monitoring are juxtaposed.

[0035] Preferably, a risk assessment model is constructed based on the information triad graph, user basic information and a clinical knowledge graph, real-time multi-source heterogeneous data is clinically risk-judged by the risk assessment model, and a clinical risk result is outputted and specific warning content and intervention suggestions are generated, including:

[0036] Based on the correlation characteristic index, time-series correlation characteristics and clinically derived correlation characteristics are obtained, and a clinical knowledge graph is constructed based on the time-series correlation characteristics and the clinically derived correlation characteristics in combination with a clinical professional knowledge base;

[0037] A basic risk identification model is trained based on the correlation characteristic index and the clinical knowledge graph, and the basic risk identification model includes a time-series encoder, a knowledge-enhanced encoder and a generative model;

[0038] The time-series encoder, the knowledge-enhanced encoder and the generative model are optimized and trained based on the correlation characteristic index labeled by risk and intervention system tags in combination with the clinical knowledge graph and user basic information to obtain an optimized risk identification model;

[0039] Real-time multi-source heterogeneous data is clinically risk-judged based on the risk identification model, a clinical risk result is outputted and specific warning content and intervention suggestions are generated.

[0040] Preferably, the basic risk identification model is trained based on the correlation characteristic index and the clinical knowledge graph, and the basic risk identification model includes a time-series encoder, a knowledge-enhanced encoder and a generative model, including:

[0041] mapping each time-series association feature in the association feature indicators to a corresponding node of the clinical knowledge graph as first training data to train a time-series encoder for identifying association relationships of features in each dimension and aligning with clinical semantics;

[0042] combining the output of the time-series encoder with the triples of the clinical knowledge graph as second training data to train a knowledge-enhanced encoder for fusing time-series features in each dimension with clinical knowledge rules;

[0043] transforming the triples of the clinical knowledge graph into natural language descriptions, combining medical text corpora corresponding to sleep, emotion and body as third training data to train a generative model for judging risks and generating intervention suggestions.

[0044] In a second aspect, the embodiments of the present application provide a sleep-emotion-body health linkage risk assessment and intervention system, comprising:

[0045] a data monitoring module configured to acquire multi-source heterogeneous data in dimensions of sleep monitoring, emotion evaluation and body health diagnosis, perform standardized processing on the multi-source heterogeneous data in each dimension, and acquire corresponding target data streams;

[0046] a triad graph generation module configured to extract association feature indicators according to the target data streams in each dimension, and generate an information triad graph of sleep-emotion-body health association according to the association feature indicators;

[0047] a risk identification module configured to construct a risk assessment model based on the information triad graph, user basic information and a clinical knowledge graph, perform clinical risk judgment on real-time multi-source heterogeneous data through the risk assessment model, output a clinical risk result, and generate specific warning content and intervention suggestions.

[0048] The present application has the following beneficial effects:

[0049] 1. By processing and fusing data of sleep, emotion and body health and generating a triad graph, originally scattered and heterogeneous data is integrated into a unified health data view with a patient as the center and time as the axis, so that medical staff can efficiently evaluate the patient's state from three strongly associated dimensions in a comprehensive and overall manner;

[0050] 2. By associating feature indicators with clinical knowledge graphs to build risk identification models, the traditional and passive "problem occurs, then problem is solved" mode is changed to an active monitoring, risk assessment and early warning mode, which can timely warn before or at the early stage of serious symptoms, gain valuable "time window" for clinical intervention, and solve the problem that traditional methods rely on the personal experience of medical staff, which easily misses the complex correlation of multiple factors, nonlinearity and sometimes lag, leading to unreliable analysis results and delayed intervention, improving the pertinence of intervention suggestions and shortening the intervention response time;

[0051] 3. By analyzing the time series data of sleep, emotion and body, identifying specific abnormal correlation patterns between them, and combining clinical knowledge graphs to build a risk identification model containing time series encoders, knowledge-enhanced encoders and generative models, it can accurately capture cross-period correlations and only retain abnormal correlation features with clinical significance, improving the efficiency of cross-dimensional data parallel fusion while improving risk identification accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0052] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments, read in conjunction with the accompanying drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the application. Moreover, the same reference numerals are used in every drawing to denote the same components.

[0053] Figure 1 A sleep-emotion-body health linkage risk assessment and intervention method flowchart is provided for the embodiments of the present application;

[0054] Figure 2 An information triad diagram is provided for the embodiments of the present application;

[0055] Figure 3 A sleep-emotion-body health linkage risk assessment and intervention system module diagram is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present application more clear and apparent, the present application will be further described in detail below in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only one of the best embodiments of the present application, which are used to explain the present application and do not limit the protection scope of the present application. All other embodiments obtained by those skilled in the art without making creative efforts fall within the scope of protection of the present application.

[0057] Embodiment 1: As shown in the following table, a sleep-emotion-body health linkage risk assessment and intervention method includes steps S1-S3, wherein: Figure 1 S1: acquiring sleep, emotion and body health data of a user;

[0058] S1, acquire multi-source heterogeneous data of sleep monitoring, emotion evaluation and somatic health diagnosis dimensions, standardize the multi-source heterogeneous data of each dimension to obtain corresponding target data flow.

[0059] As an optional implementation, step S1 includes:

[0060] Collecting first health monitoring data and second health monitoring data of a user's sleep cycle, and fusing the first health monitoring data and the second health monitoring data to calculate target sleep monitoring data;

[0061] Based on the user's electronic medical record, a state evaluation scale is extracted, the state evaluation scale is analyzed and abnormally marked, and structured target emotion evaluation data is obtained;

[0062] Based on the user's nursing information, key entity recognition is performed according to a trained medical named entity recognition model, and the identified key entities are standardized and symptom attribute relationship mapped to form target somatic health data.

[0063] In this embodiment, by standardizing the data of each dimension, standardized, time-sequenced and correlatable basic data are provided for multi-source data fusion. Using a unified data format and time granularity, the originally isolated sleep, emotion and somatic data can be fused, avoiding the problem of low fusion efficiency caused by data heterogeneity in conventional systems.

[0064] As an optional implementation, the collection of first health monitoring data and second health monitoring data of a user's sleep cycle, the fusion of the first health monitoring data and the second health monitoring data to calculate target sleep monitoring data includes:

[0065] Based on the first monitoring device, first health monitoring data of the user is acquired, including at least heart rate data, blood oxygen saturation data and body motion data;

[0066] Based on the second monitoring device, the heart shock signal of a specific human body region of the user is acquired, and after signal preprocessing of the heart shock signal, the respiration signal and the heartbeat signal are acquired;

[0067] Fusing the heart rate data, the blood oxygen saturation data, the body motion data, the respiration signal and the heartbeat signal to calculate the heart rate index, the respiration rate index and the body motion index, and determining the sleep state type according to the index data;

[0068] The heart rate index, the respiration rate index, the body motion index and the sleep state type are packaged in a preset data frame format to obtain standardized target sleep monitoring data.

[0069] In some embodiments, the first monitoring device comprises a smart watch / bracelet, and the first health monitoring data represents aggregated data of the user collected by the smart watch / bracelet in a certain time period. The data includes: heart rate data, including resting heart rate, average heart rate, heart rate variability, and heart rate curve data; blood oxygen saturation data, including nighttime blood oxygen saturation baseline value, blood oxygen saturation nighttime measurement record list, and identified blood oxygen drop events; body movement data, including body movement frequency and intensity time series; activity data, including steps, calories consumed, and activity duration.

[0070] The second monitoring device comprises a device for monitoring data of a specific region of the human body, such as a smart mattress. A high-frequency photoelectric sensor array is arranged in a specific region of the mattress corresponding to the chest and neck region of the human body. The sensor array is highly sensitive to the micro-body vibrations caused by heartbeats, blood flow, and respiratory movements, i.e., ballistocardiogram signals. The raw analog voltage signals output by the sensor array are amplified and filtered by a signal conditioning circuit, which suppresses low-frequency body turning noises and high-frequency environmental electronic noises. The signals are then converted into digital signals by an analog-to-digital converter. In the controller, the digital signals are further digitally filtered and processed to separate out signal components corresponding to different physiological activities, including: respiratory signals and heartbeat signals. The respiratory signals are obtained by extracting ultra-low frequency signal components in the 0.1 Hz ~ 0.5 Hz frequency band, and the heartbeat signals are obtained by extracting medium-low frequency signal components in the 0.5 Hz ~ 20 Hz frequency band.

[0071] Specifically, the heart rate data, blood oxygen saturation data, body movement data, respiratory signals, and heartbeat signals are fused to calculate heart rate indicators, respiratory rate indicators, and body movement indicators, and the sleep state type is determined based on the indicator data, including:

[0072] The separated heartbeat signals are subjected to peak detection, and the first health monitoring data is combined to identify consecutive heartbeat intervals and calculate instantaneous heart rate and average heart rate, thereby obtaining the heart rate indicators.

[0073] The separated respiratory signals are subjected to peak or trough detection, and the first health monitoring data is combined to calculate respiratory period and respiratory rate, thereby obtaining the respiratory rate indicators.

[0074] Body movement events are identified based on the amplitude variance or energy mutation of the first health monitoring data and the original ballistocardiogram signals or specific frequency band signals, thereby obtaining the body movement indicators.

[0075] Further, based on the body movement frequency, heart rate, respiration rate and their variability of the user in the whole night sleep state, the sleep state type is divided into accurate falling asleep, rapid eye movement period, light sleep and deep sleep, and the sleep state type corresponding to the time window of each index is determined in combination with the calculated heart rate index, respiration rate index and body movement index. The heart rate index, respiration rate index, body movement index and sleep state type are summarized according to the preset time interval, and the target sleep monitoring data is obtained.

[0076] In the embodiment, the sleep-related data is collected by dual sources, which can cover peripheral physiology, core cardiopulmonary activity and body movement behavior, provide more comprehensive data support for sleep staging and apnea / hypopnea event recognition, and realize uninterrupted whole-night monitoring, improve the integrity and authenticity of the data; the measurement deviation of a single monitoring device is effectively filtered through multi-source data fusion calculation index, and the data credibility is further improved. The integrated data not only contains isolated physiological values, but also associates the sleep state of the corresponding period, realizes the correlation analysis of physiological indexes and sleep state, and provides a prerequisite for subsequent sleep and body abnormality correlation.

[0077] As an optional implementation, the state evaluation scale based on the user electronic medical record is analyzed and abnormally marked to obtain structured target emotion evaluation data, including:

[0078] The mental state evaluation scale of the user in the electronic medical record system or / and the psychological evaluation system is obtained based on the first communication link.

[0079] The key evaluation field including at least the evaluation type, the evaluation name and the score thereof and the evaluation date is extracted based on the mental state evaluation scale;

[0080] The abnormal value of the key evaluation field is checked, and the abnormal value is abnormally marked or triggers the re-collection of the mental state evaluation scale, the uniform structured format mapping is performed based on the checked key evaluation field, and the target emotion evaluation data of the user is obtained.

[0081] As an optional implementation, the key evaluation field including at least the evaluation type, the evaluation name and the score thereof and the evaluation date extracted based on the mental state evaluation scale further includes:

[0082] If the mental state evaluation scale is in the form of unstructured text, the mental state evaluation scale is semantically analyzed based on natural language technology to obtain initial semantic information;

[0083] The similarity matching is performed based on the initial semantic information and the key evaluation information library to extract the key evaluation field.

[0084] In some embodiments, the first communication link refers to a connection established through a medical information exchange standard interface with an electronic medical record system and a dedicated psychological assessment system of a hospital to access emotional structured assessment report data and unstructured assessment report text. Among them, for emotional structured assessment report data, key assessment fields such as assessment scale name, scale total score, assessment execution date and time, and assessor are accurately extracted through database query or API call; for unstructured assessment report text, such as assessment results in nursing records or doctor's notes, natural language processing technology is used for analysis, which is first converted into machine-readable initial semantic information, and then key entities such as assessment name, total score, and assessment date are identified and extracted, so as to convert unstructured text information into structured data.

[0085] Further, the data extracted from different sources and in different formats are verified for reasonableness (for example, the total score of the old-age depression assessment should be between 0-30), and the abnormal values are marked or triggered for re-collection. The verified data is uniformly mapped to a standard format for storage, and a timestamp (assessment date and time) is attached to each assessment record to facilitate time sequence correlation analysis with physical health data.

[0086] As an optional implementation, based on user care information, key entity recognition is performed according to a trained medical named entity recognition model, and the recognized key entities are standardized and mapped to symptom attribute relationships to form target physical health data, including:

[0087] Based on the second communication link, nursing course information in the hospital information system is obtained, and for structured nursing course information, key entity data is extracted, including at least vital signs, medication records, laboratory test results, and nursing assessment items; for unstructured nursing course information, a medical named entity recognition model is used for key entity recognition, including at least symptom entities, sign entities, and treatment measure entities;

[0088] After the identified symptom entities, sign entities, and treatment measure entities are standardized by medical terminology, a symptom-attribute relationship pair is constructed and entity data mapping is performed to obtain symptom attribute structured data;

[0089] Among them, the key entity data and the symptom attribute structured data constitute the target physical health data.

[0090] In the present embodiment, the second communication link refers to a communication link for communicating with a nursing information system, a vital sign monitoring system, a laboratory information system and a medical order system through a medical information exchange standard interface; to query and extract nursing medical history information, such as vital signs: body temperature, blood pressure (systolic pressure / diastolic pressure), pulse, manually measured blood oxygen saturation, respiratory rate, with accurate measurement time stamps; medication records extracted from the medical order system, including: drug name, dose, administration route (oral, intravenous injection, etc.), administration time and frequency; laboratory test results: blood routine, electrolyte, liver function and other key indicators; nursing assessment items: such as pain score (numerical rating scale NRS), fall risk assessment (Morse score) and other standardized scores.

[0091] In some embodiments, for free text in nursing records and doctor medical history records, a medical named entity recognition model constructed based on natural language processing technology is used for semantic analysis and extraction, such as identifying symptom entities (such as “chest pain”, “dyspnea”, “dizziness”, “nausea”, “nighttime paroxysmal dyspnea”) from the text, physical sign entities (such as “edema of both lower extremities”, “wet rales in the lungs”) and treatment measure entities (such as “oxygen inhalation”, “physical cooling”); and normalizing the identified symptoms with different expressions but the same meaning into standard medical terms, for example, normalizing “breathless” and “short of breath” into “dyspnea”, and associating them with the industry standard term set. Further extract the severity of the symptoms (such as “severe headache”), the body part (such as “left lower limb pain”) and the time modifier (such as “nighttime cough aggravation”), construct “symptom-attribute” relationship pairs, and form structured data.

[0092] In some embodiments, all data sources, i.e. structured data, symptoms and signs after natural language processing, device data, etc., are aligned and fused with patient ID and time stamp as the unique index; finally, each piece of physical health data contains patient ID, observation item type (such as “symptom”, “vital sign”, “medication”), specific content (such as “dyspnea”, “blood pressure”), numerical value / state, unit and occurrence time.

[0093] S2, extracting associated feature indicators according to the target data flow of each dimension, and generating a sleep-emotion-physical health associated information triplot according to the associated feature indicators.

[0094] As an optional implementation, step S2 includes:

[0095] spatiotemporally aligning the target sleep monitoring data, the target emotion assessment data and the target physical health data;

[0096] The correlation characteristic indexes of the dimensional target data streams are extracted based on spatiotemporal alignment, and the correlation characteristic indexes are subjected to data visualization to generate an information triad chart of sleep monitoring, emotion evaluation and vital sign monitoring in parallel.

[0097] Specifically, the correlation characteristic indexes of the dimensional target data streams at least include:

[0098] Sleep dimension characteristics: sleep efficiency (total sleep time / in-bed time), deep sleep proportion, sleep latency, number of night awakenings, apnea hypopnea index, average night blood oxygen;

[0099] Emotion dimension characteristics: total score of the Geriatric Depression Scale and its key sub-item scores, total score of the Generalized Anxiety Disorder Scale and its key sub-item scores, wherein the scale scores are converted into continuous numerical sequences, and for non-continuous evaluation points, interpolation or nearest neighbor method is used for smoothing processing to form a time series;

[0100] Somatic dimension characteristics: numerical type, including average heart rate, respiratory rate, pain score; event type, converting symptoms (such as “dyspnea”, “dizziness”) in symptom attribute structured data and medication records (such as “taking painkillers”) into binary events (occurrence = 1, non-occurrence = 0) occurring on the time axis.

[0101] In some examples, as shown in FIG. 1, Figure 2 The correlation characteristic indexes share the coordinate axis, and a parallel triad chart is constructed: the horizontal axis represents time, and the three parallel vertical axis regions include an upper region for drawing a sleep stage curve, wherein N1 represents the initial stage from wakefulness to sleep, N2 represents light sleep stage, N3 represents deep sleep stage, and REM represents rapid eye movement sleep stage, and the intensity and distribution of apnea events and body movement events are displayed in the form of a shaded bar chart; a middle region for drawing a score change curve of emotion scales such as GDS (Geriatric Depression Scale) and GAD-7 (Generalized Anxiety Disorder) in the form of a line chart or an area chart; and a lower region for drawing a vital sign curve (such as heart rate, blood oxygen), and annotating the occurrence of pain events, dyspnea symptoms and medication time points in the form of event marker points or bar charts.

[0102] In this embodiment, by constructing the information triad graph, it is helpful to quickly locate the time sequence linkage phenomenon of sleep, emotion and vital signs, so that the explicit correlation can be directly converted into the core correlation characteristics of the subsequent risk identification model, that is, to screen high-value characteristics, reduce the model calculation complexity, reduce the data processing amount, and thus improve the model training efficiency. At the same time, the traditional original data table or feature matrix is difficult for clinical experts to understand, while the visual chart with shared coordinate axes can directly present the data rule, so that the expert can quickly point out that the emotion score interpolation result in a certain period does not conform to the clinical actuality (for example, the patient's anxiety score should show a downward trend due to drug adjustment, but not an upward trend), so as to correct the data preprocessing bias.

[0103] S3, constructing a risk assessment model based on the information triad graph, user basic information and clinical knowledge graph, performing clinical risk judgment on real-time multi-source heterogeneous data through the risk assessment model, and outputting clinical risk results and generating specific warning content and intervention suggestions.

[0104] As an optional implementation, step S3 includes:

[0105] obtaining time sequence correlation characteristics and clinical derivative correlation characteristics based on the correlation characteristic indicators, and constructing a clinical knowledge graph based on the time sequence correlation characteristics and the clinical derivative correlation characteristics in combination with a clinical professional knowledge base;

[0106] training a basic risk identification model based on the correlation characteristic indicators and the clinical knowledge graph, the basic risk identification model including a time sequence encoder, a knowledge enhancement encoder and a generative model;

[0107] optimizing and training the time sequence encoder, the knowledge enhancement encoder and the generative model based on the correlation characteristic indicators labeled by the risk and intervention system tags in combination with the clinical knowledge graph and user basic information, to obtain an optimized risk identification model;

[0108] performing clinical risk judgment on real-time multi-source heterogeneous data based on the risk identification model, and outputting clinical risk results and generating specific warning content and intervention suggestions.

[0109] Specifically, obtaining time sequence correlation characteristics and clinical derivative correlation characteristics based on the correlation characteristic indicators includes:

[0110] performing data cleaning for different data types, wherein, performing outlier elimination and missing value completion on numerical data in the target sleep monitoring data and the target somatic health data, performing logical verification and missing value processing on the target emotion evaluation data, including returning abnormal data for re-recording, marking invalid for single scale missing value or filling with historical mean value, and performing deduplication on event type data in the target somatic health data;

[0111] According to the cleaned data, basic features are extracted, including numerical features, time sequence features, and clinical derived features, which represent the interaction between the body, sleep, and emotions and the corresponding risk states.

[0112] Based on the dimensional feature association influence, the dimensional phenomenon types and feature thresholds are determined according to the clinical guidelines, for example, the phenomenon types of emotions are severe depression and moderate depression, and the feature thresholds of severe depression and moderate depression are total score greater than 12 and total score greater than 8, respectively.

[0113] According to the basic features, the time sequence association and clinical derived association between multi-dimensional features are determined, and then time sequence association features and clinical derived association features are obtained. The time sequence association features include time sequence lag association features (such as the change of B-dimensional features after the occurrence of A-dimensional a phenomenon) and co-occurrence features (such as the simultaneous occurrence frequency of a phenomenon and B-dimensional y phenomenon), and the clinical derived association features include risk stratification features (such as the composite feature constructed by combining patient age and the sum of emotion assessment types) and trend association features (such as the correlation between the decline slope of sleep efficiency and the rise slope of anxiety score).

[0114] Specifically, based on the time sequence association features and clinical derived association features, a clinical knowledge graph is constructed by combining a clinical professional knowledge base, including:

[0115] The clinical professional knowledge base is constructed based on clinical guideline knowledge, expert consensus knowledge, or medical terminology knowledge.

[0116] The clinical professional knowledge base, time sequence association features, and clinical derived association features are integrated to determine graph relationships, relationship attributes, and graph nodes, and then a clinical knowledge graph is constructed according to the graph relationships, relationship attributes, and graph nodes.

[0117] It should be noted that the clinical knowledge graph covers the full-link knowledge of phenomena, attributes, associations, risks, and interventions, which is used to convert scattered clinical rules, data association features, and medical terminologies into a structured node-relationship-attribute graph, ensuring clinical rationality and data adaptability, and providing interpretable, quantifiable, and iterative clinical knowledge support for risk identification models.

[0118] As an optional implementation, the basic risk identification model is trained based on the association feature indicators and the clinical knowledge graph, and the basic risk identification model includes a time sequence encoder, a knowledge enhanced encoder, and a generative model, including:

[0119] Each time sequence association feature in the association feature indicators is mapped to the corresponding node of the clinical knowledge graph, which is used as the first training data to train a time sequence encoder for identifying the association relationship between each dimensional feature and aligning with clinical semantics;

[0120] The triplets of the clinical knowledge graph are combined with the output of the time series encoder as second training data to train a knowledge-enhanced encoder for fusing time series features of each dimension with clinical knowledge rules;

[0121] The triplets of the clinical knowledge graph are converted into natural language descriptions, and medical text corpora corresponding to sleep, emotion, and body are combined as third training data to train a generative model for judging risk and generating intervention suggestions.

[0122] In the present embodiment, the time series encoder is mainly used to convert the original features of sleep, emotion, and body at the hour level into vector representations with time series dependencies, providing basic features at the data level for subsequent knowledge fusion; the knowledge-enhanced encoder is mainly used to fuse the time series features with the clinical association rules in the knowledge graph, thereby solving the problem that pure data-driven models do not understand clinical logic, and providing fused features of data rules and clinical knowledge for the risk identification module of the subsequent generative model; the generative model generates structured risk judgment results (including risk level information) and converts the results into natural language warnings and intervention suggestions that can be directly used by doctors, realizing effective landing from data judgment to clinical application.

[0123] In some embodiments, the associated feature indicators labeled based on the risk and intervention system are combined with the clinical knowledge graph and user basic information to optimize the training of the time series encoder, the knowledge-enhanced encoder, and the generative model, and an optimized risk identification model is obtained, including:

[0124] The hour-level feature sequence output by the pre-trained time series encoder is taken as a basic time series sequence, and the symptom attribute relationship pairs corresponding to the current user sample are extracted from the clinical knowledge graph to obtain knowledge association features;

[0125] The knowledge association features are input into the attention layer of the knowledge-enhanced encoder as attention bias, and the basic time series sequence is input into the knowledge-enhanced encoder, and the fusion and optimization training of the time series encoder and the knowledge-enhanced encoder are realized through the attention mechanism;

[0126] The fusion feature vector output by the optimized training of the knowledge-enhanced encoder is combined with the associated feature indicators labeled with risk level and intervention suggestion labels as input of the generative model for optimization training of the model, and an optimized generative model is obtained.

[0127] In some embodiments, the risk and intervention system label includes a first-level label, a second-level label, and a third-level label. The first-level label represents whether the correlation pattern is established, i.e., whether any phenomenon of a dimension changes with any phenomenon of another dimension. The first-level label is marked by a binary label. If yes, mark 1. If no, mark 0. The second-level label represents the clinical risk level of the correlation pattern, including no risk, low risk, high risk, and the like. The third-level label represents the warning content and intervention suggestion corresponding to the risk, which is marked by text, for example, high risk: low sleep efficiency with severe depression, prone to cause night arrhythmia; intervention suggestion: adjust sleep medicine, increase the frequency of psychological counseling.

[0128] Specifically, through optimized training, it is ensured that the encoder can output a feature sequence with time sequence correlation and basic clinical semantics. The feature sequence of the time sequence encoder will be input into the knowledge-enhanced encoder. At this time, the knowledge-enhanced encoder will call the embedding vector and the correlation rule of the corresponding phenomenon in the knowledge graph, so as to upgrade the pure data features to the fusion features of data and knowledge, and provide accurate judgment basis for the risk identification of the generative model. When the fusion features of the knowledge-enhanced encoder are input into the generative model, the structured instructions of the correlation judgment result and the risk level are first obtained, and then the instructions and the intervention rules in the knowledge graph are input into the generative model to generate warnings and intervention suggestions.

[0129] In the embodiment, through model pre-training and optimized training, it is ensured that the final risk identification model completely learns the expression paradigm of clinical terms and intervention suggestions, so as to accurately and reliably convert the structured judgment result into natural language clinical suggestions. The three basic models work together, which not only solves the problem of low efficiency of time sequence data processing in conventional systems, but also makes up for the business demand of no clinical knowledge and no landing output, provides efficient processing, accurate fusion, and practical output for risk identification, and finally realizes reliable risk judgment and takes into account the judgment efficiency, improves the intervention response efficiency, and overcomes the problem that the conventional system only outputs data reports and cannot simultaneously and accurately give intervention schemes and intervention lag.

[0130] In other possible embodiments, when the risk identification model identifies a novel correlation that has not been included in the preset rule library, it is added to the optimized preset rule library to optimize the model in reverse, for example, the lag effect of "significant decrease in daytime activity does not immediately affect mood, but will cause sleep interruption after 2 days", so that the model is continuously optimized and updated, ensuring that the accuracy and clinical relevance of the analysis continue to improve, further continuously improving the risk identification accuracy of the model and the reliability of the intervention suggestion generation.

[0131] Embodiment 2, as shown in Figure 3 The sleep-emotion-somatic health linkage risk assessment and intervention system provided by the embodiments of the present application comprises:

[0132] a data monitoring module, configured to acquire multi-source heterogeneous data of sleep monitoring, emotion evaluation and somatic health diagnosis dimensions, to perform standardization processing on the multi-source heterogeneous data of each dimension, and to acquire corresponding target data streams;

[0133] a triplex graph generation module, configured to extract associated feature indexes according to the target data streams of each dimension, and to generate an information triplex graph of sleep-emotion-somatic health association according to the associated feature indexes;

[0134] a risk identification module, configured to construct a risk assessment model based on the information triplex graph, user basic information and a clinical knowledge graph, to perform clinical risk judgment on real-time multi-source heterogeneous data through the risk assessment model, to output a clinical risk result and to generate specific warning content and intervention suggestions.

[0135] In the embodiment, the data of sleep, emotion and somatic health are processed and fused to generate a triplex graph, the originally scattered and heterogeneous data is integrated into a unified health data view with a patient as the center and time as the axis, so that medical staff can efficiently evaluate the patient's state from three strongly associated dimensions in a comprehensive and overall manner; a risk identification model is constructed through associated feature indexes and a clinical knowledge graph, the traditional and passive mode of "problem occurs, then problem is solved" is changed into an active monitoring, risk assessment and early warning mode, and thus a valuable "time window" can be obtained for clinical intervention before serious symptoms occur or in an early stage, and the problem that the traditional method relies on the personal experience of medical staff, easily misses the complex association of multi-factor, nonlinearity and sometimes lag, and leads to unreliable analysis results and delayed intervention is solved. In addition, by analyzing the time series data of the three dimensions of sleep, emotion and somatic health, identifying specific abnormal association patterns therebetween, and combining a clinical knowledge graph construction model, cross-period association and abnormal association features with only clinical significance can be accurately captured, and the efficiency of cross-dimension data parallel fusion and the accuracy of risk identification are improved.

[0136] The specific embodiments above are the preferred embodiments of the present application, and the specific implementation range of the present application is not limited thereto, the scope of the present application includes but is not limited to the specific embodiments, and equivalent changes made according to the shape, structure and method of the present application are within the protection scope of the present application.

Claims

1. A method for risk assessment and intervention linking sleep, mood, and physical health, characterized by: Includes the following steps: Acquire multi-source heterogeneous data from sleep monitoring, emotion assessment, and physical health diagnosis dimensions; standardize the multi-source heterogeneous data from each dimension to obtain the corresponding target data stream. Based on the target data stream of each dimension, relevant feature indicators are extracted, and a tripartite diagram of the correlation between sleep, mood and physical health is generated based on the relevant feature indicators. A risk assessment model is constructed based on an information tripartite graph, user basic information, and a clinical knowledge graph. This model assesses clinical risk from real-time, multi-source, heterogeneous data, outputs clinical risk results, and generates specific warnings and intervention recommendations. Based on the aforementioned correlation feature indicators, temporal correlation features and clinically derived correlation features are obtained, and a clinical knowledge graph is constructed by combining the temporal correlation features and clinically derived correlation features with a clinical professional knowledge base. A basic risk identification model is trained based on the aforementioned correlation feature indicators and the aforementioned clinical knowledge graph. The basic risk identification model includes a temporal encoder, a knowledge-enhanced encoder, and a generative model. Based on the associated feature indicators labeled with risk and intervention system tags, combined with the clinical knowledge graph and user basic information, the temporal encoder, knowledge enhancement encoder and generative model are optimized and trained to obtain an optimized risk identification model; Based on the risk identification model, clinical risk assessment is performed on real-time multi-source heterogeneous data, and clinical risk results are output, along with specific warning content and intervention suggestions.

2. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 1, characterized in that: The process involves acquiring multi-source heterogeneous data across sleep monitoring, emotion assessment, and physical health diagnosis dimensions, standardizing the multi-source heterogeneous data for each dimension, and obtaining the corresponding target data stream, including: Collect first and second health monitoring data of the user's sleep cycle, and calculate the target sleep monitoring data by fusing the first and second health monitoring data; Based on the user's electronic medical record, a state assessment scale is extracted, and the state assessment scale is parsed and anomaly annotation is performed to obtain structured target emotion assessment data. Based on user care information, key entities are identified using a trained medical named entity recognition model. The identified key entities are then standardized and mapped to symptom attribute relationships to form target physical health data.

3. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 2, characterized in that: The process of collecting first and second health monitoring data from the user's sleep cycle, and then fusing the first and second health monitoring data to calculate the target sleep monitoring data, includes: The user's first health monitoring data is obtained based on the first monitoring device, including at least heart rate data, blood oxygen saturation data, and body movement data; Based on the second monitoring device, cardiac impact signals of specific human body areas of the user are obtained, and respiratory and heartbeat signals are obtained after signal preprocessing of the cardiac impact signals. By integrating heart rate data, blood oxygen saturation data, body movement data, respiratory signals, and heartbeat signals, heart rate indicators, respiratory rate indicators, and body movement indicators are calculated, and the sleep state type is determined based on the data of each indicator; The heart rate, respiratory rate, body movement, and sleep state types are encapsulated in a preset data frame format to obtain standardized target sleep monitoring data.

4. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 2, characterized in that: The process involves extracting a state assessment scale from the user's electronic medical record, parsing and annotating the state assessment scale, and obtaining structured target emotion assessment data, including: Based on the first communication link, obtain the mental state assessment scale of the user in the electronic medical record system and / or psychological testing system; Based on the mental state assessment scale, extract key assessment fields including at least the assessment type, assessment name and its score, and assessment date; Outlier validation is performed on the key assessment fields, and outliers are marked as anomalies or trigger the re-collection of the mental state assessment scale. Based on the validated key assessment fields, a unified structured format is mapped to obtain the user's target emotion assessment data.

5. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 4, characterized in that: The key assessment fields extracted based on the mental state assessment scale include at least the assessment type, assessment name and its score, and assessment date, and also include: If the mental state assessment scale is in unstructured text form, then the mental state assessment scale is semantically parsed based on natural language processing technology to obtain initial semantic information; Based on the initial semantic information, a similarity match is performed with the key evaluation information database to extract key evaluation fields.

6. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 2, characterized in that: The process involves identifying key entities based on user care information using a trained medical named entity recognition model, standardizing the identified key entities, and mapping symptom attribute relationships to form target physical health data, including: Nursing progress information in the hospital information system is obtained based on the second communication link. For structured nursing progress information, key entity data is extracted, including at least vital signs, medication records, laboratory test results, and nursing assessment items. For unstructured nursing progress information, a medical named entity recognition model is used to identify key entities, including at least symptom entities, sign entities, and treatment measure entities. After normalizing the identified symptom entities, sign entities, and treatment measure entities using standard medical terminology, symptom-attribute relationship pairs are constructed and entity data is mapped to obtain structured symptom attribute data. The key entity data and the symptom attribute structured data together constitute the target physical health data.

7. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 2, characterized in that: The step of extracting correlation feature indicators from the target data stream of each dimension and generating a tripartite graph of the correlation between sleep, mood, and physical health based on the correlation feature indicators includes: The target sleep monitoring data, the target emotion assessment data, and the target physical health data are spatiotemporally aligned. Based on spatiotemporal alignment, the relevant feature indicators of the target data streams in each dimension are extracted and visualized to generate a three-part information graph that combines sleep monitoring, emotion assessment, and vital sign monitoring.

8. The method for risk assessment and intervention linking sleep, emotion, and physical health according to claim 1, characterized in that: The basic risk identification model trained based on the associated feature indicators and the clinical knowledge graph includes a temporal encoder, a knowledge-enhancing encoder, and a generative model, comprising: Each temporal correlation feature in the correlation feature index is mapped to the corresponding node of the clinical knowledge graph, and used as the first training data to train a temporal encoder for recognizing the correlation relationships of features in each dimension and aligning with clinical semantics. The triples based on the clinical knowledge graph, combined with the output of the temporal encoder, are used as the second training data to train a knowledge-enhancing encoder that integrates temporal features of each dimension with clinical knowledge rules. The triples based on the clinical knowledge graph are converted into natural language descriptions, and combined with medical text corpora corresponding to sleep, emotions, and physical conditions as third training data to train a generative model for risk assessment and intervention recommendations.

9. A sleep-emotion-physical health linkage risk assessment and intervention system, applicable to the sleep-emotion-physical health linkage risk assessment and intervention method as described in any one of claims 1-8, characterized in that: include: The data monitoring module is used to acquire multi-source heterogeneous data from sleep monitoring, emotion assessment, and physical health diagnosis dimensions, and to standardize the multi-source heterogeneous data from each dimension to obtain the corresponding target data stream. The triptych generation module is used to extract correlation feature indicators from the target data stream of each dimension, and generate a triptych of information on the correlation between sleep, mood and physical health based on the correlation feature indicators. The risk identification module is used to build a risk assessment model based on the information tripartite graph, user basic information, and clinical knowledge graph. The risk assessment model is used to make clinical risk judgments on real-time multi-source heterogeneous data, output clinical risk results, and generate specific warning content and intervention suggestions.

Images