Intelligent data labeling method based on artificial intelligence

Abstract

The invention discloses an intelligent data labeling method based on artificial intelligence, and relates to the technical field of intelligent medical treatment, and the method comprises the following steps: obtaining physiological parameter data, historical health parameter data and motion state data of a user; outputting motion state label data through a pre-constructed motion recognition model according to the motion state data; and performing sliding window analysis on the historical health parameter data, and dynamically establishing and updating long-term change data of the user health parameters. According to the scheme, whether the physiological parameter fluctuation is caused by the movement or not can be distinguished through associated retrieval of time window division and the movement state label, for example, when it is detected that the heart rate fluctuation of the user during running exceeds the threshold value, normal physiological response instead of abnormity can be judged in combination with the movement label; the probability that normal physiological parameter fluctuation caused by movement is mistakenly marked as abnormal can be effectively reduced, and meanwhile, the pertinence of anomaly detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and in particular to an artificial intelligence-based data intelligent labeling method. Background Art

[0002] With the rapid development of technologies such as artificial intelligence, big data analysis and the Internet of Things, the medical and health field is gradually evolving towards intelligence and automation. Smart medical devices have been widely used in scenarios such as daily health management, chronic disease monitoring and telemedicine. These devices can collect a variety of users' physiological parameter data in real time, providing a rich data foundation for medical services and health management. At the same time, with the advancement of medical informatization, data analysis and discrimination capabilities based on artificial intelligence have become an important trend and research hotspot in data processing in the medical and health field.

[0003] However, in the process of implementing the technical solutions of the invention in the embodiments of this application, it was found that the above technology has at least the following technical problems:

[0004] Existing solutions often perform anomaly detection based only on a single type of parameter data, lacking the ability to collaboratively analyze multiple types of physiological parameters and motion status data, which easily leads to insufficient accuracy and robustness in anomaly discrimination. They fail to fully consider the long-term changing trends and individual differences of user health parameters, resulting in limited sensitivity and specificity of anomaly detection. There is a lack of deep integration of motion status recognition models and physiological parameter analysis, resulting in normal parameter fluctuations caused by exercise being misjudged as anomalies, reducing the accuracy of abnormal data labeling. Summary of the Invention

[0005] The purpose of the present invention is to provide an artificial intelligence-based data intelligent labeling method to solve the problems raised in the above background technology.

[0006] In order to achieve the above object, the technical solution of the present invention is as follows:

[0007] The present invention discloses an artificial intelligence-based data intelligent labeling method, which is applied to the intelligent labeling of abnormal physiological parameters of users in intelligent medical equipment, comprising the following steps:

[0008] Obtain the user's physiological parameter data, historical health parameter data and exercise status data;

[0009] Outputting motion state label data through a pre-built motion recognition model according to the motion state data;

[0010] Performing sliding window analysis on the historical health parameter data to dynamically establish and update long-term change data of the user's health parameters;

[0011] Calculating a physiological parameter fluctuation value based on the physiological parameter data, determining whether the physiological parameter fluctuation value is greater than a preset abnormal threshold, and if so, retrieving motion state tag data, and adjusting the preset abnormal threshold based on the retrieval result;

[0012] Wherein, the preset abnormal threshold is set according to the long-term change data of the user's health parameters;

[0013] Then determine whether the physiological parameter fluctuation value is greater than the adjusted preset abnormal threshold, and if so, determine it as potential abnormal physiological parameter data;

[0014] Inputting the physiological parameter fluctuation values ​​and the long-term change data of the user's health parameters into a pre-built artificial intelligence discrimination model, and outputting the confidence level of the potential abnormal physiological parameter data;

[0015] Determine whether the confidence level is greater than a set value, and if so, determine that the data is truly abnormal physiological parameter data;

[0016] The actual abnormal physiological parameter data is marked with a structured data structure.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. This solution uses time window division and associated retrieval of motion status tags to distinguish whether physiological parameter fluctuations are caused by exercise. For example, when a user's heart rate fluctuations exceed a threshold during running, combined with the motion tag, it can be determined to be a normal physiological response rather than an abnormality. This can effectively reduce the probability of normal physiological parameter fluctuations caused by exercise being mislabeled as abnormalities, while also improving the targeted nature of anomaly detection.

[0019] 2. This solution uses a dynamic transition interval and a gradual adjustment mechanism to ensure that the abnormal threshold changes synchronously with the user's actual motion state, solving the problem of misjudgment caused by threshold lag or mutation. It can effectively distinguish normal physiological fluctuations during motion state switching from real abnormal data, reduce misjudgment caused by sudden changes in motion state, and improve the accuracy and reliability of abnormal data annotation.

[0020] 3. This solution effectively identifies such correlation fluctuations through a multivariate joint distribution model, avoiding misjudging normal physiological reactions as abnormalities. It can establish adaptive anomaly detection standards based on the dynamic changes in the user's individual health status, effectively solving the false alarm problem caused by traditional methods that ignore individual differences. At the same time, through multi-parameter joint analysis, it can accurately distinguish normal fluctuations caused by exercise from pathological abnormal fluctuations, thereby improving the accuracy of anomaly labeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them:

[0022] Figure 1 This is a flowchart of the steps of an artificial intelligence-based data intelligent labeling method of the present invention;

[0023] Figure 2 A schematic diagram of a process for adjusting a preset abnormality threshold provided by the present invention;

[0024] Figure 3 A schematic diagram of a process for setting a preset abnormality threshold provided by the present invention;

[0025] Figure 4 This is a flow chart of the data annotation module provided by the present invention. DETAILED DESCRIPTION

[0026] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0027] In existing technologies, smart medical devices provide support for user health management by collecting physiological parameter data in real time. However, existing anomaly detection methods usually rely on a single type of parameter and lack the ability to collaboratively analyze multi-dimensional data. For example, when changes in the user's exercise state cause heart rate fluctuations, traditional methods may misjudge it as an abnormality due to failure to identify the exercise state. At the same time, existing technologies fail to effectively track the long-term trends of user health parameters, resulting in individual differences not being fully considered and the abnormal threshold setting lacking dynamic adaptability.

[0028] In order to solve the above problems, it was found that the existing technology has defects such as fragmented analysis of multi-source data, insufficient long-term trend modeling, and failure to eliminate motion interference factors. Through analysis, it was found that the accuracy of anomaly detection is limited by the single data dimension and poor adaptability to individual differences. Based on this, it is proposed to integrate motion state recognition and physiological parameter analysis, establish a dynamic threshold adjustment mechanism, and introduce long-term health trend modeling. Further consideration is given to introducing artificial intelligence models for secondary verification of potential anomalies to form a multi-stage collaborative anomaly labeling system.

[0029] After introducing the basic concept of the present invention, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0030] Example:

[0031] See also Figure 1-Figure 4 , an artificial intelligence-based data intelligent labeling method, which is applied to the intelligent labeling of abnormal physiological parameters of users in intelligent medical devices, includes the following steps:

[0032] Obtain the user's physiological parameter data, historical health parameter data and exercise status data;

[0033] According to the motion state data, the motion state label data is output through the pre-built motion recognition model;

[0034] Perform sliding window analysis on historical health parameter data to dynamically establish and update long-term change data of user health parameters;

[0035] Calculate the physiological parameter fluctuation value based on the physiological parameter data, determine whether the physiological parameter fluctuation value is greater than the preset abnormal threshold, and if so, retrieve the motion state tag data and adjust the preset abnormal threshold according to the retrieval result;

[0036] Among them, the preset abnormal threshold is set according to the long-term change data of the user's health parameters;

[0037] Then determine whether the physiological parameter fluctuation value is greater than the adjusted preset abnormal threshold, and if so, determine it as potential abnormal physiological parameter data;

[0038] Input the physiological parameter fluctuation values ​​and the long-term change data of the user's health parameters into the pre-built artificial intelligence discrimination model, and output the confidence level of the potential abnormal physiological parameter data;

[0039] Determine whether the confidence level is greater than a set value, if so, it is determined to be true abnormal physiological parameter data;

[0040] The real abnormal physiological parameter data is marked with a structured data structure.

[0041] Physiological parameter data refers to physical signs collected in real time by medical equipment. Specifically, this can be achieved by using smart watches to monitor heart rate and blood oxygen saturation, or by using a sphygmomanometer to obtain blood pressure data, providing a basic data source for abnormality detection.

[0042] Motion state label data refers to the classification results of the user's activity state. Specifically, it can be achieved by using an accelerometer to collect motion data and then using a neural network model to identify states such as resting and walking. It is used to distinguish the normal causes of fluctuations in physiological parameters.

[0043] The long-term change data of user health parameters refers to the statistical characteristics that reflect individual health trends. Specifically, this can be achieved by using a sliding window to calculate statistics such as the mean and standard deviation, and by fitting the trend curve through linear regression, providing a basis for personalized threshold setting.

[0044] Adjusting the preset abnormal threshold refers to dynamically modifying the judgment criteria according to the motion state. Specifically, this can be achieved by establishing a threshold mapping table under different motion states, and automatically switching the corresponding threshold when a state switch is detected, to avoid misjudgment caused by motion interference;

[0045] The artificial intelligence discrimination model refers to an intelligent algorithm used for anomaly verification. Specifically, it can be achieved by using a deep neural network to jointly analyze fluctuation values ​​and long-term trends, and improve the reliability of anomaly judgment through confidence assessment.

[0046] The specific implementation process is as follows: after the physiological parameter data and motion status data are collected synchronously, the motion recognition model classifies and labels the motion status data; historical health data continuously updates the individual health baseline through sliding window statistics to form long-term change data of health parameters; when it is detected that the physiological parameter fluctuation value exceeds the preset abnormal threshold, the current motion status label data is retrieved, and the preset abnormal threshold is adjusted according to the retrieval result; for potential abnormal physiological parameter data that exceeds the adjusted preset abnormal threshold, it is further combined with the long-term change data of the user's health parameters to input the artificial intelligence discrimination model for comprehensive judgment, and finally only high-confidence anomalies are structuredly labeled; this process eliminates interference factors through motion status recognition, combines long-term trends to realize threshold personalization, and uses artificial intelligence discrimination models to improve judgment accuracy.

[0047] Compared with existing technologies, existing solutions usually use fixed thresholds and do not distinguish between exercise states. For example, an increase in heart rate during running is directly judged as an abnormality. This solution dynamically adjusts the threshold through exercise state labels to effectively distinguish physiological fluctuations from pathological abnormalities. At the same time, existing technologies lack long-term trend tracking, while this solution establishes an individualized health baseline through sliding window analysis, making anomaly detection more in line with the user's actual condition. In addition, existing methods mostly use a single threshold judgment, while this solution introduces an artificial intelligence discrimination model for multi-dimensional joint verification, significantly reducing the false alarm rate.

[0048] Through the above technical solutions, this application effectively solves the problem of misjudgment caused by motion interference, improves the specificity of anomaly detection through a dynamic threshold adjustment mechanism, and improves detection sensitivity by making anomaly determinations more consistent with individual health changes. The multi-stage verification process combined with motion state recognition ensures accuracy while enabling automated labeling of abnormal data, providing a reliable data foundation for subsequent medical diagnosis.

[0049] This application further proposes that the user's physiological parameter data, historical health parameter data, and exercise status data be obtained, specifically including:

[0050] Collect the user's physiological parameter data through smart medical devices (such as smart watches, electrocardiogram monitors, oximeters, etc.), including heart rate, blood oxygen saturation, body temperature, and blood pressure;

[0051] Retrieve the user's historical health parameter data from the cloud health file over a period of time. The historical health parameter data includes past illnesses, family history, and medication history.

[0052] The user's motion status data is collected through sensors integrated into smart medical devices. The motion status data includes the number of steps, energy consumption, and speed.

[0053] Timestamp tags are added to physiological parameter data, historical health parameter data, and exercise status data.

[0054] Among them, historical health parameter data refers to the user's past health records, which can be retrieved through electronic health records stored in the cloud to provide benchmark data for individualized analysis;

[0055] Motion status data refers to the user's current or historical activity status information, which can be collected through embedded sensors such as accelerometers and gyroscopes to identify the impact of exercise on physiological parameters;

[0056] Timestamp tagging refers to adding a precise time mark to each piece of data. This can be achieved by synchronizing the system clock to ensure the timing alignment of multi-source data.

[0057] Specifically, physiological parameter data are collected in real time through smart medical devices. For example, when a user wears a smart watch, his or her heart rate, blood oxygen saturation and other data are continuously monitored; historical health parameter data are retrieved from the cloud health file, such as the user's disease records and medication history in the past three months, which are integrated into the current analysis; motion status data are obtained through the built-in sensors of smart medical devices, such as the number of steps, speed and other parameters recorded by the accelerometer, which are used to judge the user's current activity intensity; all data are timestamp-tagged, such as automatically recording the collection time when the data is generated, so that data from different sources can be matched and associated according to the timeline during subsequent analysis.

[0058] Beneficial effects:

[0059] This application solves the problem of misjudgment caused by a single data source in the existing technology, and improves the accuracy of anomaly detection through multi-dimensional data collection and time synchronization mechanism; at the same time, the combination of historical health parameters and real-time motion status enables the system to dynamically adjust the anomaly judgment threshold to adapt to the individual differences of different users; the addition of timestamp tags further ensures the consistency of analysis across data sources, providing a reliable data foundation for subsequent intelligent discrimination models.

[0060] This application further proposes that, based on the motion state data, the motion state label data can be outputted through a pre-built motion recognition model, specifically including:

[0061] Preprocess the motion state data, including noise filtering and feature extraction (such as acceleration modulus, angular velocity change rate, etc.);

[0062] and inputting the pre-processed motion state data into a pre-built motion recognition model;

[0063] The motion recognition model is built based on neural networks, including multi-layer perceptron, recurrent neural network, and long short-term memory network structures;

[0064] The motion state data and corresponding motion state label data for training are obtained through big data, and used as supervision signals to train the motion recognition model using supervised learning.

[0065] Among them, the motion status label data includes resting, walking, running and strenuous exercise;

[0066] The motion state label data is output through the pre-built motion recognition model, and the motion state label data is associated with the physiological parameter data at the corresponding time point.

[0067] Noise filtering refers to eliminating high-frequency interference signals in sensor data through digital filtering algorithms. Specifically, it can be achieved by using a Butterworth low-pass filter or a moving average filter to improve the signal-to-noise ratio of motion state data.

[0068] Feature extraction refers to extracting discriminative physical quantities from raw motion data. Specifically, the acceleration modulus can be used to calculate the amplitude of the three-axis acceleration vector, or the angular velocity change rate can be calculated through differential operation to characterize the dynamic characteristics of different motion modes.

[0069] Neural network construction refers to the use of a network structure with time series processing capabilities. Specifically, a long short-term memory network can be used to capture the temporal correlation of motion data, or a multi-layer perceptron can be used to implement nonlinear feature mapping to improve the accuracy of motion state classification.

[0070] Supervised learning refers to the use of labeled samples to train a classification model. Specifically, the cross-entropy loss function can be used to optimize the model parameters, or the back-propagation algorithm can be used to update the network weights to establish a mapping relationship between motion data and state labels.

[0071] The specific implementation process is as follows: the motion state data is first passed through a Butterworth low-pass filter to eliminate high-frequency noise, and then the modulus mean of the acceleration vector and the standard deviation of the angular velocity difference are extracted through a sliding window as feature vectors; the feature vector is input into a neural network model containing a bidirectional long-short-term memory layer, and the model is trained on a data set containing labeled samples such as resting, walking, and running through supervised learning; the trained model classifies the motion data collected in real time, outputs the corresponding motion state label data, and establishes a timestamp association between the motion state label data and the synchronously collected physiological parameter data; for example, when an intense exercise label is detected, the label will be associated with the heart rate and blood oxygen data in the corresponding time period and stored.

[0072] Beneficial effects:

[0073] This application can accurately identify the dynamic characteristics under different motion states, avoiding the fluctuations of physiological parameters that are incorrectly labeled as abnormalities due to misjudgment of motion patterns; the precise association between motion state labels and physiological parameters provides a reliable basis for subsequent dynamic adjustment of thresholds, effectively reducing the false alarm rate and improving the accuracy of abnormal data labeling.

[0074] This application further proposes to perform sliding window analysis on historical health parameter data, and dynamically establish and update long-term change data of user health parameters, specifically including:

[0075] Taking a preset time window (e.g., 7 days, 30 days, etc.) as a unit, extract relevant data sequences within the time window from the collected historical health parameter data;

[0076] In each sliding window, for each historical health parameter data, the mean, standard deviation, maximum value, minimum value, and quantile statistical characteristics are calculated respectively;

[0077] Based on the statistical results of each sliding window, the linear regression method is used to fit the long-term change trend of the user's health parameters to obtain the long-term change data of the user's health parameters;

[0078] As historical health parameter data is continuously collected, the sliding window moves forward in real time, old data is automatically removed, and new data is added, dynamically updating the long-term change data of user health parameters.

[0079] The time window refers to a fixed time interval used for segmented analysis of historical health parameter data. Specifically, it can be used as a unit, for example, 7 days or 30 days. By periodically dividing the data sequence, the phased feature extraction of historical health parameters can be achieved.

[0080] Statistical characteristics refer to quantitative descriptive indicators of the data distribution within a window. Specifically, the mean can reflect the overall level, the standard deviation can measure the degree of fluctuation, and the quantile can identify the range of extreme values. Through multi-dimensional calculations, a comprehensive representation of the dynamic changes of health parameters can be achieved.

[0081] The linear regression method refers to a mathematical model that fits the trend of parameter changes through the least squares method. Specifically, time can be used as the independent variable and health parameters as the dependent variable. The regression coefficient reflects the rate and direction of parameter change over time.

[0082] Dynamic update means automatically adjusting the analysis window as new data is added. Specifically, a sliding window mechanism can be used to remove old data and include new data, and the timeliness of long-term changing data can be guaranteed through real-time iterative calculation.

[0083] Specifically, historical health parameter data is divided into continuous time windows of fixed length, for example, every 30 days is an analysis cycle. In each window, statistics such as mean and standard deviation are calculated for parameters such as heart rate and blood pressure to quantify the data distribution characteristics within the cycle. Based on the statistical results of multiple windows, a linear regression model is used to fit the trend line of health parameters over time, such as the average annual growth rate of heart rate or the average monthly fluctuation range of blood pressure. As smart devices continue to collect new data, the window automatically slides forward. For example, the data of the earliest day is eliminated and the data of the latest day is added every day. The statistical characteristics are recalculated and the long-term change data of the user's health parameters are updated, so that the long-term change data can reflect the evolution of the user's health status in real time.

[0084] Through the above technical solution, this application can dynamically track the long-term evolution of user health parameters, provide a personalized baseline reference for abnormality detection, effectively distinguish normal physiological fluctuations from potential abnormal signals, reduce misjudgments due to individual differences or the progression of chronic diseases, and improve the sensitivity and specificity of abnormality labeling.

[0085] The present application further proposes to determine whether the fluctuation value of the physiological parameter is greater than a preset abnormal threshold, and if so, to retrieve the motion state tag data, specifically including:

[0086] For the physiological parameter data, according to the set time window, the physiological parameter fluctuation value within the window is calculated. The physiological parameter fluctuation value is the difference between the maximum and minimum values ​​of the physiological parameter data within the time window;

[0087] The physiological parameter fluctuation value is compared with the preset abnormal threshold. If the physiological parameter fluctuation value is greater than the preset abnormal threshold, the motion state label data corresponding to the time window is automatically retrieved within the time window in which the physiological parameter fluctuation value is determined to be abnormal.

[0088] Among them, the physiological parameter fluctuation value refers to the dynamic change of the parameter calculated by the difference between the maximum and minimum values ​​within the time window. It can be implemented by using a sliding window statistical method. For example, with a time window length of 5 minutes or 30 minutes, the range of the physiological parameters within the window is calculated in real time. This feature is used to dynamically reflect the immediate fluctuation status of the user's physiological parameters.

[0089] Among them, the preset abnormal threshold refers to the critical value of the fluctuation range dynamically set according to the long-term change data of the user's health parameters. It can be implemented by using the quantile or standard deviation multiple of the statistical distribution of historical data. For example, the threshold is set to the 95% quantile of the historical fluctuation value. This feature is used to personalize the judgment of whether the fluctuation of physiological parameters exceeds the normal range.

[0090] Specifically, physiological parameter data is divided into continuous time windows, such as a window every five minutes. Within each window, the difference between the maximum and minimum values ​​of parameters such as heart rate and blood pressure is calculated in real time as the physiological parameter fluctuation value. When the physiological parameter fluctuation value exceeds the preset abnormal threshold, such as heart rate fluctuation exceeding 20 beats / minute, the system automatically triggers the retrieval of the motion state tag data corresponding to the window time, such as whether the user was running or resting during this period. By associating and matching the time window of abnormal fluctuation with the motion state tag, it is possible to identify whether the physiological parameter abnormality is caused by exercise, thereby avoiding misjudgment.

[0091] Beneficial effects:

[0092] The present application can effectively reduce the probability of normal physiological parameter fluctuations caused by exercise being mislabeled as abnormal. For example, when a user performs strenuous exercise, the system can dynamically adjust the judgment logic by retrieving the exercise status label to avoid mislabeling the increased heart rate caused by exercise as arrhythmia. At the same time, the method improves the specificity of anomaly detection, for example, only focusing on abnormal fluctuations in non-exercise states, reducing the burden of redundant data processing.

[0093] This application further proposes adjusting the preset abnormality threshold according to the search results, specifically including:

[0094] Retrieve the motion state tag data within the current moment and its adjacent time window;

[0095] Preset the normal fluctuation range and abnormal fluctuation threshold of physiological parameter data under different motion state label data;

[0096] Matching the abnormal fluctuation threshold of the corresponding physiological parameter data according to the retrieval result of the motion state label data;

[0097] When it is detected that the motion state tag data switches in a short period of time (such as from rest to intense exercise), a dynamic transition interval is set to gradually adjust the abnormal fluctuation threshold of the physiological parameter data;

[0098] The preset abnormality threshold is adjusted to the dynamically adjusted fluctuation abnormality threshold of the physiological parameter data corresponding to the motion state label data.

[0099] The dynamic transition interval refers to the buffer zone set when the motion state switches rapidly. It can be implemented by using a time-weighted or state probability model to avoid misjudgment caused by sudden changes in thresholds.

[0100] Gradual adjustment refers to gradually changing the abnormal threshold according to the duration and intensity of the motion state switching. Specifically, it can be achieved by using linear interpolation or exponential smoothing algorithm to synchronize the threshold change with the user's actual physiological changes.

[0101] Specifically, when the system detects a switch from resting to strenuous exercise, it identifies the switch and triggers a dynamic transition interval. Within this interval, the threshold for abnormal fluctuations in physiological parameters will not immediately adopt the higher threshold corresponding to strenuous exercise. Instead, it will gradually adjust based on the duration of the switch. For example, within the first 30 seconds after the switch, the threshold will increase by a fixed percentage per second to reach the target value. This gradual adjustment mechanism effectively prevents fluctuations in physiological parameters caused by sudden changes in exercise status from being misjudged as abnormal.

[0102] Through the above technical solution, the present application can effectively distinguish normal physiological fluctuations during motion state switching from real abnormal data, reduce misjudgment caused by sudden changes in motion state, and improve the accuracy and reliability of abnormal data labeling.

[0103] This application further proposes that the preset abnormal threshold value be set according to the long-term change data of the user's health parameters, specifically including:

[0104] Model the long-term change data of each user's health parameters to obtain the current user's health parameter baseline and fluctuation range;

[0105] Dynamically set personalized abnormal thresholds for each health parameter based on the current user's health parameter baseline and fluctuation range, and use them as the initial preset abnormal thresholds;

[0106] Combined with the correlation between health parameters, the initial preset abnormal threshold is determined by using multivariate joint distribution;

[0107] As the long-term change data of health parameters are updated in real time, the preset abnormal thresholds are adaptively adjusted periodically.

[0108] The health parameter baseline refers to the typical value of the user's health parameter calculated through statistical features. Specifically, the mean or median within the sliding window can be used as the baseline value to establish a benchmark reference for personalized anomaly detection.

[0109] Multivariate joint distribution refers to considering the statistical correlation between different health parameters. Specifically, the covariance matrix or principal component analysis method can be used to construct a joint probability distribution model to avoid the risk of misjudgment caused by isolated judgment of a single parameter.

[0110] Specifically, during the operation of smart medical equipment, the long-term change data of user heart rate, blood oxygen saturation and other parameters are first continuously updated through sliding window analysis. For example, every 30 days is an analysis cycle, and the mean, standard deviation and trend line slope of each parameter are calculated. Based on this data, a user-specific health parameter baseline is established. For example, the average heart rate of the past three months is used as the baseline value in the resting state. Combined with the correlation between blood pressure and heart rate, a multivariate Gaussian distribution model is used to calculate the joint fluctuation range, and the value that deviates from the distribution by more than twice the standard deviation is set as the initial abnormal threshold. When the device detects a trend change in the user's recent health parameters, such as a decrease of 5 beats / minute in the average resting heart rate for two consecutive weeks, the system automatically triggers the threshold update mechanism, recalculates the multivariate joint distribution and generates a new abnormal threshold.

[0111] Through the above technical solution, the present application can establish adaptive anomaly detection standards based on the dynamic changes in the user's individual health status, effectively solving the false alarm problem caused by traditional methods ignoring individual differences. For example, for users with sinus bradycardia, the system can automatically raise the lower limit of the abnormal heart rate threshold to avoid mislabeling normal physiological status as abnormal data. At the same time, through multi-parameter joint analysis, it can accurately distinguish between normal fluctuations caused by exercise and pathological abnormal fluctuations, thereby improving the accuracy of anomaly labeling.

[0112] This application further proposes to input the physiological parameter fluctuation values ​​and the long-term change data of the user's health parameters into a pre-built artificial intelligence discrimination model, and output the confidence level of the potential abnormal physiological parameter data, specifically including:

[0113] The fluctuation values ​​of physiological parameters and the long-term change data of user health parameters are taken as input features and combined to form a comprehensive feature vector;

[0114] The comprehensive feature vector is input into a pre-built artificial intelligence discrimination model, and the artificial intelligence discrimination model is constructed through a neural network;

[0115] The confidence level of potential abnormal physiological parameter data is output through the artificial intelligence discrimination model.

[0116] The comprehensive feature vector refers to a multidimensional data set formed by normalizing the fluctuation value of physiological parameters and the long-term change data. It can be implemented by using feature fusion technology to provide multidimensional input for the model.

[0117] An artificial intelligence discriminant model refers to a classification or regression model built based on a neural network. It can be implemented using a convolutional neural network or a deep neural network structure, which is used to perform nonlinear mapping of input features and output a confidence score.

[0118] The specific implementation process is as follows: the fluctuation value of physiological parameters is calculated through a sliding window, for example, the difference between the maximum and minimum heart rate values ​​is calculated within a 30-second time window. This value reflects the immediate fluctuation of physiological parameters. The long-term change data of user health parameters are analyzed through a sliding window to analyze historical health parameter data, such as the mean and standard deviation of heart rate in the past 30 days, and linear regression is used to fit the long-term trend to form dynamically updated baseline data. The comprehensive feature vector is obtained by standardizing and merging the above two types of data, for example, splicing features such as heart rate fluctuation value, long-term mean, standard deviation into a vector, and inputting it into a pre-trained artificial intelligence discriminant model. The artificial intelligence discriminant model adopts a deep neural network structure, such as a multi-layer perceptron comprising an input layer, a hidden layer, and an output layer. The nonlinear combination of features is calculated through forward propagation, and finally a confidence score between 0 and 1 is output, indicating the credibility of potential abnormal data.

[0119] Through the above-mentioned technical solution, the present application can combine the short-term fluctuation characteristics and long-term change trends of physiological parameters, and realize accurate identification of potential abnormal data through multi-dimensional feature fusion and deep neural network modeling. For example, after the user performs strenuous exercise, even if the heart rate fluctuation value exceeds the conventional threshold, the system can still identify that the user is in the normal exercise recovery stage through long-term data to avoid misjudgment; at the same time, when abnormal deviations of long-term trends accompanied by heart rate fluctuations are detected, the model can output a high confidence score and accurately identify real abnormal events. This technical solution effectively solves the misjudgment problem caused by ignoring individual differences and dynamic changes in existing methods, and improves the accuracy of abnormal labeling.

[0120] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0121] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based data intelligent labeling method, applied to the intelligent labeling of abnormal physiological parameters of users in intelligent medical equipment, characterized in that: The following steps are involved: Obtain the user's physiological parameter data, historical health parameter data and exercise status data; Outputting motion state label data through a pre-built motion recognition model according to the motion state data; Performing sliding window analysis on the historical health parameter data to dynamically establish and update long-term change data of the user's health parameters; Calculating a physiological parameter fluctuation value based on the physiological parameter data, determining whether the physiological parameter fluctuation value is greater than a preset abnormal threshold, and if so, retrieving motion state tag data, and adjusting the preset abnormal threshold based on the retrieval result; Wherein, the preset abnormal threshold is set according to the long-term change data of the user's health parameters; Then determine whether the physiological parameter fluctuation value is greater than the adjusted preset abnormal threshold, and if so, determine it as potential abnormal physiological parameter data; Inputting the physiological parameter fluctuation values ​​and the long-term change data of the user's health parameters into a pre-built artificial intelligence discrimination model, and outputting the confidence level of the potential abnormal physiological parameter data; Determine whether the confidence level is greater than a set value, and if so, determine that the data is truly abnormal physiological parameter data; The actual abnormal physiological parameter data is marked with a structured data structure.

2. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: Acquiring the user's physiological parameter data, historical health parameter data, and exercise status data specifically includes: Collecting physiological parameter data of the user through smart medical devices, wherein the physiological parameter data includes heart rate, blood oxygen saturation, body temperature, and blood pressure; Retrieving the user's historical health parameter data from the cloud health file over a period of time, including past illnesses, family history, and medication history; Collecting user's motion status data through sensors integrated into smart medical devices, the motion status data including number of steps, activity energy consumption, and speed; Timestamp tags are added to the physiological parameter data, historical health parameter data and exercise status data.

3. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: Outputting motion state label data using a pre-built motion recognition model according to the motion state data specifically includes: Preprocessing the motion state data, wherein the preprocessing includes noise filtering and feature extraction; and inputting the pre-processed motion state data into a pre-built motion recognition model; The motion recognition model is constructed based on a neural network, including a multi-layer perceptron, a recurrent neural network, and a long short-term memory network structure; The motion state data and corresponding motion state label data for training are obtained through big data, and used as supervision signals to train the motion recognition model using supervised learning. The motion state label data includes resting, walking, running and strenuous exercise; The motion state label data is output through the pre-built motion recognition model, and the motion state label data is associated with the physiological parameter data at the corresponding time point.

4. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: Performing a sliding window analysis on the historical health parameter data to dynamically establish and update the long-term change data of the user's health parameters specifically includes: Taking the preset time window as the unit, extract the relevant data sequence within the time window from the collected historical health parameter data; In each sliding window, for each historical health parameter data, the mean, standard deviation, maximum value, minimum value, and quantile statistical characteristics are calculated respectively; Based on the statistical results of each sliding window, the linear regression method is used to fit the long-term change trend of the user's health parameters to obtain the long-term change data of the user's health parameters; As historical health parameter data is continuously collected, the sliding window moves forward in real time, old data is automatically removed, and new data is added, dynamically updating the long-term change data of user health parameters.

5. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: Determining whether the physiological parameter fluctuation value is greater than a preset abnormal threshold, and if so, retrieving the motion state tag data specifically includes: For the physiological parameter data, according to a set time window, calculate the physiological parameter fluctuation value within the window, wherein the physiological parameter fluctuation value is the difference between the maximum value and the minimum value of the physiological parameter data within the time window; The physiological parameter fluctuation value is compared with the preset abnormal threshold. If the physiological parameter fluctuation value is greater than the preset abnormal threshold, the motion state label data corresponding to the time window is automatically retrieved within the time window in which the physiological parameter fluctuation value is determined to be abnormal.

6. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: Adjusting the preset anomaly threshold based on search results specifically includes: Retrieve the motion state tag data within the current moment and its adjacent time window; Preset the normal fluctuation range and abnormal fluctuation threshold of physiological parameter data under different motion state label data; Matching the abnormal fluctuation threshold of the corresponding physiological parameter data according to the retrieval result of the motion state label data; When it is detected that the motion status tag data switches in a short period of time, a dynamic transition interval is set to gradually adjust the abnormal fluctuation threshold of the physiological parameter data; The preset abnormality threshold is adjusted to the dynamically adjusted fluctuation abnormality threshold of the physiological parameter data corresponding to the motion state label data.

7. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: The preset abnormal threshold is set according to the long-term change data of the user's health parameters, specifically including: Model the long-term change data of each user's health parameters to obtain the current user's health parameter baseline and fluctuation range; Dynamically set personalized abnormal thresholds for each health parameter based on the current user's health parameter baseline and fluctuation range, and use them as the initial preset abnormal thresholds; Combined with the correlation between health parameters, the initial preset abnormal threshold is determined by using multivariate joint distribution; As the long-term change data of health parameters are updated in real time, the preset abnormal thresholds are adaptively adjusted periodically.

8. The method for intelligent data annotation based on artificial intelligence according to claim 1, characterized in that: Inputting the physiological parameter fluctuation values ​​and the long-term change data of the user's health parameters into a pre-built artificial intelligence discrimination model, and outputting the confidence level of the potential abnormal physiological parameter data specifically includes: The physiological parameter fluctuation value and the long-term change data of the user's health parameter are used as input features and combined to form a comprehensive feature vector; Inputting the comprehensive feature vector into a pre-built artificial intelligence discrimination model, wherein the artificial intelligence discrimination model is constructed by a neural network; The confidence level of potential abnormal physiological parameter data is output through the artificial intelligence discrimination model.

Images