Health suggestion report generation method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of wisdom medical treatment and artificial intelligence, and discloses a health suggestion report generation method and device, computer equipment and a storage medium, which are applied to a health assessment and health suggestion system of wisdom medical treatment, and the method comprises the following steps: collecting multi-dimensional time sequence health data of a user; preprocessing the multi-dimensional time sequence health data to generate a structured time sequence feature text; identifying an abnormal mode label based on the structured time sequence feature text through a lightweight language model; performing knowledge matching and knowledge adjustment on the basis of the abnormal model label through the associated medical knowledge base, and generating target health description information and target medical suggestion information; and generating a health suggestion report according to the abnormal mode label, the target health description information and the target medical suggestion information. The method has the advantages that the health assessment accuracy is improved, the personalized suggestion generation capability is enhanced, and the dynamic threshold adjustment is realized.

Description

Technical Field

[0001] The present application relates to the fields of smart medical care and artificial intelligence technology, and in particular to a method, apparatus, computer equipment, and storage medium for generating a health advice report. Background Art

[0002] Current technical solutions in the fields of smart healthcare and smart elderly care have significant limitations. While wearable devices can collect basic physiological data and perform simple cloud-based analysis, these systems typically only provide discrete metric monitoring and lack the ability to deeply integrate and intelligently analyze multi-dimensional, time-series health data. Methods for generating health advice reports can be applied to smart healthcare, specifically within health assessment and health advice systems. Health advice reports can provide users with health analysis and recommendations, enabling intuitive access to health information and improving their understanding of their own condition.

[0003] When processing health data with complex time series characteristics, traditional machine learning models have difficulty effectively capturing the long-term dependencies and nonlinear characteristics between data, resulting in insufficient accuracy and comprehensiveness in health risk assessments. Existing anomaly detection algorithms integrated into medical devices often use a fixed threshold judgment mechanism that cannot be dynamically adjusted based on individual user characteristics, making it prone to false positives or omissions. At the artificial intelligence application level, although language models perform well in natural language processing tasks, research on combining them with medical knowledge bases to process structured time series health data is still in its infancy, and a complete solution has yet to be formed. These technical deficiencies make it difficult for existing systems to generate personalized health advice reports with medical reference value, and are unable to meet users' needs for precise health management. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to propose a method, apparatus, computer device and storage medium for generating a health advice report to improve the accuracy of health assessment and enhance the ability to generate personalized health advice.

[0005] In order to solve the above technical problems, the present application provides a method for generating a health advice report, including:

[0006] Collect users' multi-dimensional time-series health data;

[0007] Preprocessing the multi-dimensional time series health data to generate structured time series feature text;

[0008] Identifying abnormal pattern labels based on the structured temporal feature text using a lightweight language model;

[0009] Performing knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information;

[0010] A health advice report is generated according to the abnormal pattern label, the target health description information and the target medical advice information.

[0011] In order to solve the above technical problems, the embodiment of the present application provides a device for generating a health advice report, including:

[0012] Data collection module, used to collect users' multi-dimensional time-series health data;

[0013] A data preprocessing module, configured to preprocess the multi-dimensional time series health data to generate structured time series feature text;

[0014] An abnormal pattern label recognition module, configured to recognize abnormal pattern labels based on the structured temporal feature text using a lightweight language model;

[0015] A knowledge matching module, configured to perform knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information;

[0016] A health advice report generating module is used to generate a health advice report based on the abnormal pattern label, the target health description information and the target medical advice information.

[0017] In order to solve the above technical problems, a technical solution adopted by the present invention is: to provide a computer device, including one or more processors; a memory for storing one or more programs, so that the one or more processors can implement the method for generating a health advice report as described in any one of the above.

[0018] To solve the above technical problems, a technical solution adopted by the present invention is: a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the method for generating a health advice report described in any one of the above.

[0019] An embodiment of the present invention provides a method, apparatus, computer equipment and storage medium for generating a health advice report. The method comprises: collecting multi-dimensional time-series health data of a user; pre-processing the multi-dimensional time-series health data to generate structured time-series feature text; identifying abnormal pattern labels based on the structured time-series feature text through a lightweight language model; performing knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information; and generating a health advice report based on the abnormal pattern labels, the target health description information and the target medical advice information. The embodiment of the present invention achieves in-depth analysis of complex health data and accurate advice generation by integrating multi-dimensional time-series health data, dynamically adjusting health standard thresholds, and combining medical knowledge bases for personalized knowledge matching. It has the advantages of improving the accuracy of health assessments, enhancing the ability to generate personalized advice, and realizing dynamic threshold adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the solutions in this application, a brief introduction will be given below to the drawings required for use in the description of the embodiments of this application. Obviously, the drawings described below are some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 This is a schematic diagram of an application environment of a method for generating a health advice report in an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of the implementation process of the method for generating a health advice report provided in an embodiment of the present application;

[0023] Figure 3 yes Figure 2 A schematic flow chart of a specific implementation of step S2;

[0024] Figure 4 yes Figure 2 A schematic flow chart of a specific implementation of step S3;

[0025] Figure 5 yes Figure 4 A schematic flow chart of a specific implementation of step S33;

[0026] Figure 6 yes Figure 5 A schematic flow chart of a specific implementation method after step S331;

[0027] Figure 7 yes Figure 2 A schematic flow chart of a specific implementation of step S4;

[0028] Figure 8 yes Figure 2 A schematic flow chart of a specific implementation of step S5;

[0029] Figure 9 Schematic diagram of a device for generating a health advice report provided in an embodiment of the present application;

[0030] Figure 10 It is a schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of the application are for the purpose of describing specific embodiments only and are not intended to limit this application. The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0032] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0033] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0034] The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0035] It should be noted that the method for generating a health advice report provided in the embodiment of the present application is generally executed by a server, and accordingly, the device for generating a health advice report is generally configured in the server.

[0036] The method for generating a health advice report provided by the embodiment of the present invention can be applied in the following situations: Figure 1In an application environment, the client communicates with the server through a network. The server can receive the multi-dimensional time-series health data of the client; and generate a health advice report based on the multi-dimensional time-series health data. The server in the present invention sends the health advice report to the client. The client can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server can be implemented with an independent server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific embodiments.

[0037] The data generation method based on dynamic multi-level prompt words provided in the embodiment of the present application can be applied to financial market dynamic simulation and risk prediction scenarios, as well as to multimodal medical image-assisted diagnosis generation scenarios.

[0038] In existing technologies, the smart healthcare field generally uses wearable devices to collect physiological data and utilizes traditional machine learning models for analysis and early warning. These methods struggle to effectively capture long-term dependencies and nonlinear characteristics when processing complex time-series health data due to limitations in model structure, resulting in insufficient anomaly detection accuracy. For example, in the daily monitoring of elderly patients with chronic diseases, traditional systems may miss abnormal correlations between blood pressure fluctuations and sleep quality, leading to misjudgment of health risks.

[0039] In order to solve the above problems, it was found during the research and development process that the traditional time series analysis model cannot effectively handle the spatiotemporal correlation between multi-source heterogeneous health data, and the generation of medical advice lacks a personalized adaptation mechanism. By analyzing the difficulties of multi-dimensional data fusion, it is considered to use natural language processing technology to transform data features, use lightweight language models to capture time series patterns, and combine with medical knowledge bases to realize dynamic knowledge adjustment. Therefore, this application proposes a method for generating a health advice report, including collecting multi-dimensional time series health data of users; preprocessing the multi-dimensional time series health data to generate structured time series feature text; identifying abnormal pattern labels based on the structured time series feature text through a lightweight language model; performing knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information; generating a health advice report based on the abnormal pattern labels, target health description information and target medical advice information. This application effectively solves the problem of insufficient feature correlation capture in traditional methods in time series health data analysis, and achieves high-precision abnormal pattern recognition through natural language feature conversion and lightweight model collaborative processing. At the same time, relying on the two-way interactive mechanism of the medical knowledge base, the dynamic generation of personalized health guidance plans is completed on the basis of ensuring the scientific nature of the recommendations.

[0040] See also Figure 2 , Figure 2 A specific implementation of a method for generating a health advice report is shown.

[0041] It should be noted that the method of the present invention is not limited to the method of Figure 2 The process sequence shown is limited to the following steps:

[0042] S1: Collect users’ multi-dimensional time-series health data.

[0043] Specifically, various edge devices, such as smart watches, wristbands, and portable medical sensors, collect users' multi-dimensional time-series health data. This multi-dimensional time-series health data includes heart rate, blood pressure, blood oxygen, exercise data, sleep data, environmental data, user age, user gender, and biochemical data. This data can fully reflect the user's health status.

[0044] S2: Preprocess the multi-dimensional time series health data to generate structured time series feature text.

[0045] Specifically, multi-dimensional time series health data often contains noise and missing values, so it needs to be preprocessed. In the data cleaning stage, the embodiment of the present application adopts complex algorithms based on isolation forests or local anomaly factors to improve the accuracy of outlier detection, and selects interpolation, smoothing or deletion methods to process outliers according to the situation. In terms of feature engineering, the embodiment of the present application extracts rich time series features such as frequency domain, statistics and time domain, such as heart rate variability, blood pressure change trends, etc., to more accurately capture data patterns. Then, using predefined template rules, the numerical time series data is converted into textual descriptions, and the semantics and syntax of the description are optimized. By preprocessing multi-dimensional time series health data to ensure the data quality of the input SLM, the foundation is laid for its subsequent analysis.

[0046] See also Figure 3 , Figure 3 A specific implementation of step S2 is shown, which is described in detail as follows:

[0047] S21: Use the isolation forest algorithm or the local anomaly factor to detect outliers in the multi-dimensional time series health data, and process the outliers to obtain basic time series health data.

[0048] S22: Extracting time series feature data from the basic time series health data, wherein the time series feature data includes time domain feature data, frequency domain feature data and statistical feature data.

[0049] S23: Convert the time series feature data into natural language description information to generate the structured time series feature text.

[0050] Specifically, multi-dimensional time-series health data is first detected for anomalies using an isolation forest algorithm or a local anomaly factor. The former quickly locates global outliers by randomly partitioning the feature space, while the latter identifies local outliers by calculating local density differences. Outliers are replaced with interpolated values ​​of adjacent data points or directly deleted. The processed basic time-series health data is extracted using sliding window statistics in the time domain to capture short-term fluctuation trends, decomposed into periodic components in the frequency domain using a fast Fourier transform, and calculates distribution morphology parameters in the statistical dimension. Finally, all feature values ​​are entered into a preset semantic template. For example, the time domain mean feature is converted into "the average heart rate over the past seven days has remained at 72 beats / minute," and the frequency domain dominant frequency feature is converted into "daily blood pressure fluctuations present two periodic peaks," forming a structured text that is both machine-readable and human-understandable. The embodiments of the present application can effectively remove noise interference from multi-dimensional time-series data and retain valid information. By fully mining the potential patterns in the data through multi-dimensional feature extraction, the generated structured text retains both numerical accuracy and semantic coherence, providing high-quality input data for subsequent language model processing, thereby significantly improving the accuracy of abnormal pattern recognition in health advice reports.

[0051] The isolation forest algorithm detects outliers in data by constructing a random forest. This can be achieved by recursively partitioning the dataset and calculating path length differences. It is suitable for processing global outliers in high-dimensional data. The local anomaly factor calculates the degree of anomaly of a data point based on local density. This can be achieved by comparing the density ratio of the reachable distances between the target point and its neighbors. It is suitable for identifying observations with local density anomalies. Time domain feature data reflects the trend characteristics of health indicators over time. This can be achieved by calculating the mean, variance, or difference series using a sliding window. Frequency domain feature data reveals the regular characteristics of cyclical fluctuations in health indicators. This can be achieved by extracting the main frequency components using a Fourier transform. Statistical feature data are quantitative indicators that describe the distribution characteristics of data. This can be achieved by calculating kurtosis, skewness, or quantiles. Natural language description information converts numerical features into textual expressions with clear semantics. This can be achieved by formatting and concatenating feature names and values ​​using predefined templates.

[0052] S3: Identify abnormal pattern labels based on the structured temporal feature text through a lightweight language model.

[0053] Specifically, the present application adopts a lightweight language model (SLM) to identify abnormal pattern labels based on structured time series feature text. Furthermore, in order to enable large language models to run efficiently on resource-constrained edge devices, the present application prunes and quantizes the pre-trained lightweight language model. In the pruning stage, a variety of strategies such as structured pruning and unstructured pruning are adopted to remove unimportant parameters and connections to reduce the complexity of the model. In the quantization stage, post-training quantization or quantization-aware training and other methods are used to quantize the weights and activation values ​​of the model into a low-precision format to further compress the model size. The present application constructs an efficient and lightweight SLM that can quickly and accurately analyze time series health data on edge devices.

[0054] In a specific embodiment, the structured temporal feature text input into the lightweight language model is a textual description of temporal health features + environmental features + user personalized information. For example, the structured temporal feature text is: "User basic information: Gender: Male, Age: 45 years old", "Average heart rate in the past 7 days: 85bpm, fluctuation range: 78-95bpm", "Blood pressure trend: rising trend, average systolic blood pressure in the past 3 days: 145mmHg", "Blood oxygen level: normal, average SpO2: 97%".

[0055] See also Figure 4 , Figure 4 A specific implementation of step S3 is shown, which is described in detail as follows:

[0056] S31: Inputting the structured temporal feature text into the lightweight language model, and using natural semantic processing technology to extract health indicator features and numerical features corresponding to the health indicator features in the structured temporal feature text.

[0057] S32: Identify temporal trend features in the structured temporal feature text.

[0058] S33: Calculate abnormal weight information based on the time series trend feature, and identify the abnormal pattern label based on the abnormal weight information, the health indicator feature and the numerical feature.

[0059] Specifically, after structured temporal feature text is input into a lightweight language model, natural semantic processing technology first parses the text for health indicator names and values, such as heart rate, blood pressure, and other indicators and their measured values. Subsequently, temporal trend features are identified by analyzing the numerical changes between adjacent time points, such as an upward trend in blood pressure values ​​for three consecutive days. Abnormal weight information is generated based on the calculated trend features, such as assigning a higher weight when the slope of the blood pressure trend exceeds a threshold. Finally, the health indicator features, numerical features, and abnormal weight information are comprehensively evaluated. For example, a hypertension risk label is generated when the blood pressure value exceeds the standard range and the trend is abnormal. This process, by combining temporal dynamic analysis with the degree of numerical deviation, avoids the limitations of traditional methods that rely solely on static threshold judgment. This application can accurately identify abnormal pattern labels in health data, such as generating timely warnings when both blood pressure and heart rate show abnormal trends. Furthermore, the use of a lightweight language model reduces computing resource consumption, enabling the solution to run in real time on edge devices such as smartwatches. By combining the analysis of temporal trend features with numerical features, the solution avoids false or missed detections caused by single threshold detection and improves the reliability of abnormal pattern recognition.

[0060] Among them, a lightweight language model refers to a neural network model with a small number of parameters and high computational efficiency. Specifically, it can be implemented using a pruned or quantized pre-trained model, which can efficiently process text data in a low-resource environment. Natural semantic processing technology refers to a method of converting text data into machine-parseable semantic information. Specifically, it can be implemented using word embedding or attention mechanisms to accurately extract the names of health indicators and their corresponding values. Temporal trend features refer to the changing patterns of health indicators in time series. Specifically, they can be implemented using sliding window statistics or Fourier transforms to capture periodic fluctuations or long-term trend changes in indicators. Abnormal weight information refers to a quantitative value that reflects the deviation of an indicator from the normal range. Specifically, it can be implemented using standard deviation calculation or outlier detection algorithms to measure the degree of abnormality of different indicators.

[0061] See also Figure 5 , Figure 5 A specific implementation of step S33 is shown, which is described in detail as follows:

[0062] S331: Compare the time series trend characteristics with the preset parameter range to calculate the abnormal weight information.

[0063] S332: Compare the numerical feature with a preset health standard threshold to determine an abnormal health indicator feature.

[0064] S333: Determine the correlation between the abnormal health indicator features, and generate the abnormal pattern label based on the correlation and the abnormal weight information.

[0065] Specifically, the time series trend of health indicators is extracted through a sliding window. For example, the trend of blood glucose values ​​for seven consecutive days is fitted to obtain the slope of change. The trend slope is compared with the age-adapted reference range. For example, when the slope exceeds the preset threshold, a higher abnormal weight is assigned. At the same time, the current measured blood glucose value is compared with the standard threshold. If it exceeds the standard for three consecutive times, it is marked as an abnormal indicator. The correlation coefficient between abnormal indicators is further calculated. For example, when abnormal blood glucose is positively correlated with abnormal blood pressure, it is determined to be a metabolic syndrome association pattern based on the medical knowledge base, and finally a composite abnormality label is generated based on the abnormal weight.

[0066] Among them, the time series trend feature refers to the rising, falling or periodic regular features extracted from the time series changes of health data. Specifically, it can be achieved by sliding window mean calculation combined with differential operation, and is used to dynamically evaluate the changing trend of health indicators. Among them, the abnormal weight information refers to the quantitative value reflecting the degree of deviation of a single health indicator from the normal state. Specifically, it can be achieved by weighted calculation of the deviation amplitude of the trend slope and the preset parameter range, and is used to distinguish the abnormal priority of different indicators. Among them, the correlation refers to the interaction relationship between multiple abnormal health indicators in physiological mechanisms or pathological manifestations. Specifically, it can be achieved by Pearson correlation coefficient calculation or association path analysis based on medical knowledge graph, and is used to identify complex abnormal patterns.

[0067] See also Figure 6 , Figure 6 A specific implementation after step S333 is shown, which is described in detail as follows:

[0068] S333A: Compare the user age information in the structured temporal feature text with a preset age range to determine a target age range for the user.

[0069] S333B: Adjust the preset health standard threshold according to the target age range and the gender information in the structured time series feature text to obtain an adjusted health standard threshold.

[0070] S333C: Adjust the abnormal pattern label according to the adjusted health standard threshold.

[0071] Specifically, after detecting abnormal health indicator characteristics, the user age value is first extracted from the structured data and matched with multiple preset age intervals, for example, a 55-year-old user is classified into the late middle-aged interval. Then, the corresponding health standard adjustment rules are called in combination with the gender information, for example, the lower limit of the bone density reference value for female users is adjusted to 5% lower than that of men. The adjusted health standard threshold is used to re-evaluate the abnormal state of the indicator. If the original abnormal mode label is generated based on the unadjusted threshold, the label is corrected by comparing the new threshold range. For example, when the original diastolic blood pressure threshold of 140mmHg is adjusted to 145mmHg, the abnormal hypertension state in the original label will be cancelled.

[0072] Among them, user age information refers to the date of birth or age value extracted from structured data, which can be implemented by date parsing algorithm or numerical field matching method to determine the physiological stage of the user. The preset age range refers to the time range corresponding to different physiological characteristics divided in advance, such as infancy, adolescence, adulthood, etc., and each interval corresponds to different baseline health parameters. Gender information refers to the user's physiological gender identification, which can be directly obtained through the data field or identified through text parsing, and is used to distinguish the differences in physiological indicators between different genders. The adjusted health standard threshold refers to the numerical range that is dynamically corrected according to age and gender. For example, the standard value of blood pressure increases with age, or the hemoglobin reference value is differentiated by gender. Dynamic adjustment is achieved through table lookup or linear interpolation algorithm.

[0073] S4: performing knowledge matching and knowledge adjustment based on the abnormal model labels through the associated medical knowledge base to generate target health description information and target medical advice information.

[0074] Specifically, through embedded or associated medical knowledge bases (such as health management knowledge, chronic disease guidelines, WHO recommended standards, etc.), the potential health impacts and lifestyle-related factors corresponding to these patterns are determined, thereby generating target health description information and target medical advice information.

[0075] See also Figure 7 , Figure 7 A specific implementation of step S4 is shown, which is described in detail as follows:

[0076] S41: performing knowledge matching based on the abnormal model label through the associated medical knowledge base to generate initial health description information.

[0077] S42: Re-search the medical knowledge base based on the initial health description information to generate initial medical advice information.

[0078] S43: If there is an association feature between the user personalized information in the structured time-series feature text and the initial health description information, knowledge adjustment is performed on the initial health description information and the initial medical advice information based on the association feature to generate the target health description information and the target medical advice information.

[0079] Specifically, when the abnormal model label identifies that the user's blood pressure indicator has a continuous fluctuation feature, it first matches the hypertension-related pathological description in the medical knowledge base to generate initial health description information. Then, based on the description, the associated medical advice information is retrieved to generate initial medical advice information. When the structured temporal feature text extracts that the user has a history of abnormal renal function, the system identifies the association characteristics between the medical history and the metabolic pathway of antihypertensive drugs, triggers the knowledge adjustment mechanism to replace the specific drugs in the initial recommendation with alternative drugs that have less burden on the kidneys, and at the same time adjusts the warning content about drug contraindications in the health description information. This application realizes the personalized generation of medical advice and solves the problem of disconnection between health descriptions and medical advice and individual characteristics of users in the prior art. Through the dynamic association analysis mechanism, the system can adjust the type of drug recommendation according to the user's specific medical history, adjust the intensity of exercise recommendations according to age characteristics, and adjust the details of the diet plan according to living habits, thereby improving the individual matching and clinical applicability of health advice.

[0080] Among them, the medical knowledge base refers to a knowledge graph database that stores entity relationships and diagnostic rules in the medical field. It can be implemented using a graph database or a relational database to provide storage and retrieval functions for standard medical knowledge. User personalized information refers to unstructured text data containing age, gender, medical history and living habits. It can be extracted from the user's health record using natural language processing technology to establish the association between individual characteristics and medical knowledge. The association feature refers to the semantic matching relationship between user personalized information and medical knowledge entities. It can be identified using a semantic similarity calculation model to dynamically adjust the applicability conditions of medical advice.

[0081] S5: Generate a health advice report according to the abnormal pattern label, the target health description information and the target medical advice information.

[0082] Specifically, a personalized health advice report is generated based on abnormal pattern labels, target health description information, and target medical advice information, so as to provide users with accurate health advice.

[0083] See also Figure 8 , Figure 8 A specific implementation of step S5 is shown, which is described in detail as follows:

[0084] S51: Generate overall evaluation information according to the number of abnormal pattern labels.

[0085] S52: Associating and splicing the target health description information with the target medical advice information to generate spliced ​​information.

[0086] S53: Generate the health advice report based on the overall evaluation information and the spliced ​​information using a preset rule engine.

[0087] Specifically, in the process of generating a health advice report, the number of identified abnormal pattern labels is first counted. For example, when the number of abnormal labels exceeds the set threshold, an overall evaluation of "moderate health risk" is generated. Subsequently, the health description information output by the medical knowledge base is semantically associated with the advice information. For example, the description of "persistently high blood pressure" is automatically matched with "low-sodium diet advice" to form an associated paragraph. Finally, the information is integrated through the logical template in the rule engine. This application effectively solves the problem of insufficient systematic information integration in health advice reports. It forms a global assessment through the statistics of the number of abnormal labels to avoid one-sided judgments. The splicing operation based on semantic association ensures the logical correspondence between medical advice and health descriptions. The application of the rule engine enables report generation to have an explainable standardized process, which significantly improves the operability and professional credibility of health advice.

[0088] Among them, the overall evaluation information refers to the comprehensive assessment of health status formed by counting the number of abnormal pattern labels. Specifically, it can be implemented by using a label counting weighted algorithm or a graded threshold judgment method to filter out minor fluctuations and form a global health conclusion. Among them, association splicing refers to the operation of combining health descriptions and medical recommendations according to semantic associations. Specifically, it can be implemented by using entity linking technology based on knowledge graphs or a two-way attention mechanism to ensure that symptom descriptions and intervention measures form a corresponding relationship. Among them, the rule engine refers to a preset structured logical judgment template. Specifically, it can be implemented by using a decision tree rule set or an event-condition-action model to achieve dynamic arrangement and standardized output of report content through conditional branches.

[0089] In one specific embodiment, the generated personalized health advice report reads: Dear user, based on an analysis of your health data over the past seven days, your overall health is below average. Main issues: Your heart rate is elevated and fluctuates widely (78-95 bpm). You are advised to focus on your mood and stress levels. Your blood pressure is continuously rising, with an average systolic blood pressure of 145 mmHg over the past three days, exceeding the normal range. You are advised to limit your high-salt diet and increase your exercise. Your sleep quality is low, with a score of 58. High nighttime temperatures may affect your sleep. You are advised to improve your sleeping environment and lower the air conditioning temperature. Your average daily step count is low (3,200 steps), far below the recommended level (6,000-8,000 steps). You are advised to gradually increase your daily walking time. Lifestyle recommendations: Maintain a daily walking volume of at least 6,000 steps, reduce your intake of high-salt foods, increase your intake of fruits, vegetables, and whole grains, maintain a room temperature between 22-24°C, improve your nighttime sleep, perform 20 minutes of aerobic exercise daily, such as brisk walking or jogging, maintain a stable mood, and reduce stress appropriately. Follow-up recommendations: Continuously monitor blood pressure changes. If systolic blood pressure remains above 140 mmHg, seek medical attention. Regularly monitor sleep quality and improve sleep schedules.

[0090] It should be noted that the lightweight language model in the embodiment of the present application does not directly output the probability of occurrence or risk score of a specific disease, but generates a personalized health advice report in natural language based on the user's temporal health characteristics and trend changes. The report content includes a summary of the current health status, existing abnormal patterns, potential health risks, and lifestyle, diet, exercise, and environmental adjustment suggestions inferred based on the medical knowledge base, with good readability and operability. The lightweight language model dynamically matches medical standards and health management suggestions by combining with an embedded or linked medical knowledge base to ensure that the content of the suggestions is scientific and reliable. At the same time, the lightweight language model dynamically adjusts the abnormal judgment threshold according to the user's individual health trends, avoiding relying solely on fixed standards, thereby improving the personalization, interpretability and practicality of health analysis and early warning.

[0091] Furthermore, the lightweight language model of the embodiment of the present application continuously updates its medical cognition by dynamically accessing medical knowledge bases (such as UpToDate, PubMed or clinical guideline databases) to provide the latest and most reliable health analysis. For example, if recent medical research finds that a certain lifestyle has a new impact on cardiovascular health, the lightweight language model can automatically adjust the weight and content of health recommendations. The lightweight language model can also continuously optimize the health analysis results based on user feedback. For example, if a user marks a certain type of warning as a false alarm multiple times, the system will automatically adjust the warning rules to improve accuracy.

[0092] In an embodiment of the present application, the user's multi-dimensional time-series health data is collected; the multi-dimensional time-series health data is preprocessed to generate a structured time-series feature text; an abnormal pattern label is identified based on the structured time-series feature text through a lightweight language model; knowledge matching and knowledge adjustment are performed based on the abnormal model label through an associated medical knowledge base to generate target health description information and target medical advice information; a health advice report is generated based on the abnormal pattern label, the target health description information and the target medical advice information. The embodiment of the present invention achieves in-depth analysis of complex health data and accurate advice generation by integrating multi-dimensional time-series health data, dynamically adjusting health standard thresholds, and combining medical knowledge bases for personalized knowledge matching. It has the advantages of improving the accuracy of health assessments, enhancing the ability to generate personalized advice, and realizing dynamic threshold adjustment.

[0093] Please refer to Figure 9 , as a response to the above Figure 2 The present application provides an embodiment of a device for generating a health advice report. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0094] like Figure 9 As shown, the health advice report generation device of this embodiment includes: a data acquisition module 61, a data preprocessing module 62, an abnormal pattern label recognition module 63, a knowledge matching module 64 and a health advice report generation module 65, wherein:

[0095] The data collection module 61 is used to collect the user's multi-dimensional time series health data;

[0096] A data preprocessing module 62 is used to preprocess the multi-dimensional time series health data to generate structured time series feature text;

[0097] An abnormal pattern label recognition module 63 is used to recognize abnormal pattern labels based on the structured temporal feature text using a lightweight language model;

[0098] a knowledge matching module 64 for performing knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information;

[0099] The health advice report generating module 65 is configured to generate a health advice report according to the abnormal pattern label, the target health description information and the target medical advice information.

[0100] Furthermore, the abnormal pattern label identification module 63 includes:

[0101] A feature extraction unit, configured to input the structured temporal feature text into the lightweight language model, and extract health indicator features and numerical features corresponding to the health indicator features from the structured temporal feature text using natural semantic processing technology;

[0102] A feature recognition unit, configured to recognize a temporal trend feature in the structured temporal feature text;

[0103] A weight calculation unit is used to calculate abnormal weight information based on the time series trend feature, and identify the abnormal pattern label based on the abnormal weight information, the health indicator feature and the numerical feature.

[0104] Furthermore, the weight calculation unit includes:

[0105] A first comparison unit is configured to compare the time series trend feature with a preset parameter range to calculate the abnormal weight information;

[0106] a second comparing unit, configured to compare the numerical feature with a preset health standard threshold to determine an abnormal health indicator feature;

[0107] The correlation determination unit is used to determine the correlation between the abnormal health indicator features and generate the abnormal pattern label based on the correlation and the abnormal weight information.

[0108] Furthermore, the relevance determination unit further includes:

[0109] a third comparing unit, configured to compare the user age information in the structured time-series feature text with a preset age range to determine a target age range for the user;

[0110] a threshold adjustment unit, configured to adjust the preset health standard threshold according to the target age range and the gender information in the structured time series feature text to obtain an adjusted health standard threshold;

[0111] A label adjustment unit is used to adjust the abnormal pattern label according to the adjusted health standard threshold.

[0112] Furthermore, the knowledge matching module 64 includes:

[0113] a knowledge matching unit, configured to perform knowledge matching based on the abnormal model label through the associated medical knowledge base to generate initial health description information;

[0114] A knowledge retrieval unit, configured to perform a knowledge search in the medical knowledge base again based on the initial health description information to generate initial medical advice information;

[0115] A knowledge adjustment unit is used to perform knowledge adjustment on the initial health description information and the initial medical advice information based on the association features if there are association features between the user personalized information in the structured time-series feature text and the initial health description information, so as to generate the target health description information and the target medical advice information.

[0116] Furthermore, the health advice report generating module 65 includes:

[0117] an evaluation information generating unit, configured to generate overall evaluation information according to the number of abnormal pattern labels;

[0118] an associating and splicing unit, configured to associate and splice the target health description information with the target medical advice information to generate spliced ​​information;

[0119] A report generating unit is used to generate the health advice report based on the overall evaluation information and the spliced ​​information using a preset rule engine.

[0120] Furthermore, the data preprocessing module 62 includes:

[0121] an outlier detection unit, configured to detect outliers in the multi-dimensional time series health data using an isolation forest algorithm or a local anomaly factor, and process the outliers to obtain basic time series health data;

[0122] A feature data extraction unit, configured to extract time series feature data from the basic time series health data, wherein the time series feature data includes time domain feature data, frequency domain feature data, and statistical feature data;

[0123] The data conversion unit is used to convert the time series feature data into natural language description information to generate the structured time series feature text.

[0124] To solve the above technical problems, the present application also provides a computer device. Figure 10 , Figure 10 This is a basic structural block diagram of the computer device in this embodiment.

[0125] The computer device 7 includes a memory 71, a processor 72, and a network interface 73 that are interconnected through a system bus. It should be noted that Figure 10Only a computer device 7 having three components, memory 71, processor 72, and network interface 73, is shown. However, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead. It should be understood by those skilled in the art that a computer device herein is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0126] Computer devices can be desktop computers, laptops, PDAs, cloud servers, etc. Computer devices can interact with users through keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0127] The memory 71 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, magnetic disk, optical disk, etc. In some embodiments, the memory 71 may be an internal storage unit of the computer device 7, such as the hard disk or memory of the computer device 7. In other embodiments, the memory 71 may also be an external storage device of the computer device 7, such as a plug-in hard disk equipped on the computer device 7, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, etc. Of course, the memory 71 may also include both the internal storage unit of the computer device 7 and its external storage devices. In this embodiment, the memory 71 is generally used to store the operating system and various application software installed on the computer device 7, such as the program code of the method for generating a health advice report. In addition, the memory 71 may also be used to temporarily store various types of data that have been output or are about to be output.

[0128] In some embodiments, the processor 72 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 72 is generally used to control the overall operation of the computer device 7. In this embodiment, the processor 72 is used to execute program code stored in the memory 71 or process data, such as executing the program code of the above-mentioned method for generating a health advice report, to implement various embodiments of the method for generating a health advice report.

[0129] The network interface 73 may include a wireless network interface or a wired network interface. The network interface 73 is generally used to establish a communication connection between the computer device 7 and other electronic devices.

[0130] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores a computer program. The computer program can be executed by at least one processor to enable the at least one processor to perform the steps of the method for generating a health advice report as described above.

[0131] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of each embodiment of the present application.

[0132] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. A method for generating a health advice report, characterized in that: include: Collect users' multi-dimensional time-series health data; Preprocessing the multi-dimensional time series health data to generate structured time series feature text; Identifying abnormal pattern labels based on the structured temporal feature text using a lightweight language model; Performing knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information; A health advice report is generated according to the abnormal pattern label, the target health description information and the target medical advice information.

2. The method for generating a health advice report according to claim 1, wherein: The identifying of abnormal pattern labels based on the structured temporal feature text using a lightweight language model includes: Inputting the structured time series feature text into the lightweight language model, and using natural semantic processing technology to extract health indicator features and numerical features corresponding to the health indicator features in the structured time series feature text; Identifying temporal trend features in the structured temporal feature text; Abnormal weight information is calculated based on the time series trend feature, and the abnormal pattern label is identified based on the abnormal weight information, the health indicator feature and the numerical feature.

3. The method for generating a health advice report according to claim 2, wherein: The calculating of abnormal weight information based on the time series trend feature, and identifying the abnormal pattern label based on the abnormal weight information, the health indicator feature, and the numerical feature, includes: Comparing the time series trend characteristics with a preset parameter range to calculate the abnormal weight information; Comparing the numerical feature with a preset health standard threshold to determine an abnormal health indicator feature; The correlation between the abnormal health indicator features is determined, and the abnormal pattern label is generated based on the correlation and the abnormal weight information.

4. The method for generating a health advice report according to claim 3, wherein: After determining the correlation between the abnormal health indicator features and generating the abnormal pattern label based on the correlation and the abnormal weight information, the method further includes: Comparing the user age information in the structured time-series feature text with a preset age range to determine the target age range of the user; Adjusting the preset health standard threshold according to the target age range and the gender information in the structured time series feature text to obtain an adjusted health standard threshold; The abnormal pattern label is adjusted according to the adjusted health standard threshold.

5. The method for generating a health advice report according to claim 1, wherein: The method of performing knowledge matching and knowledge adjustment based on the abnormal model label through the associated medical knowledge base to generate target health description information and target medical advice information includes: Performing knowledge matching based on the abnormal model labels through the associated medical knowledge base to generate initial health description information; Re-searching the medical knowledge base based on the initial health description information to generate initial medical advice information; If there is an association feature between the user personalized information in the structured time-series feature text and the initial health description information, knowledge adjustment is performed on the initial health description information and the initial medical advice information based on the association feature to generate the target health description information and the target medical advice information.

6. The method for generating a health advice report according to claim 1, wherein: Generating a health advice report according to the abnormal pattern label, the target health description information, and the target medical advice information includes: generating overall evaluation information according to the number of abnormal pattern labels; Associating and splicing the target health description information with the target medical advice information to generate spliced ​​information; A preset rule engine is used to generate the health advice report based on the overall evaluation information and the spliced ​​information.

7. The method for generating a health advice report according to any one of claims 1 to 6, characterized in that: The preprocessing of the multi-dimensional time series health data to generate structured time series feature text includes: Using an isolation forest algorithm or a local anomaly factor to detect outliers in the multi-dimensional time series health data, and processing the outliers to obtain basic time series health data; Extracting time series feature data from the basic time series health data, wherein the time series feature data includes time domain feature data, frequency domain feature data, and statistical feature data; The time series feature data is converted into natural language description information to generate the structured time series feature text.

8. A device for generating a health advice report, characterized in that: include: Data collection module, used to collect users' multi-dimensional time-series health data; A data preprocessing module, configured to preprocess the multi-dimensional time series health data to generate structured time series feature text; An abnormal pattern label recognition module, configured to recognize abnormal pattern labels based on the structured temporal feature text using a lightweight language model; A knowledge matching module, configured to perform knowledge matching and knowledge adjustment based on the abnormal model labels through an associated medical knowledge base to generate target health description information and target medical advice information; A health advice report generating module is used to generate a health advice report based on the abnormal pattern label, the target health description information and the target medical advice information.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method for generating a health advice report according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for generating a health advice report according to any one of claims 1 to 7.

Images