Automatic input vital sign detection method and system

Abstract

The invention discloses an automatic input vital sign detection method and system, and relates to the technical field of medical information processing, and the method comprises the steps: collecting vital sign data of a target individual, and carrying out the synchronous processing; a vital sign data set is formed through cleaning and multi-channel data fusion; uploading the encrypted data to a medical information system to finish classified storage and updating of the data; performing feature extraction and analysis to generate a health risk prediction result; health management suggestions are generated, and data display is completed through a dynamic feedback mechanism. According to the method, the privacy and the integrity of the vital sign data are guaranteed through a dynamic updating and verification mechanism in a mode of combining symmetric encryption and asymmetric encryption, and the risks of data leakage, data tampering and version management disorder in a traditional system are overcome. By introducing an intelligent analysis method combining a classification model and a regression model, the limitation of traditional single-dimensional static assessment is solved, and a comprehensive health risk report is generated by integrating logic association among multi-modal feature data.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information processing, and in particular to a method and system for automatically inputting vital signs detection. Background Art

[0002] At present, the application of medical information technology in the field of health management is becoming more and more extensive, especially the monitoring and analysis based on vital signs data has become an important means of early warning of diseases and health management. However, the traditional methods mainly have the following technical defects.

[0003] The collection of vital signs data usually relies on a single device or a single feature, such as monitoring only heart rate or blood pressure. It lacks the ability to synchronously collect and integrate multimodal data, resulting in one-sidedness and lag in the comprehensive assessment of health status. Secondly, the collected data often lacks a unified encryption and dynamic update mechanism during transmission and storage, which makes the user's sensitive health data face the risk of privacy leakage. At the same time, untimely data version updates will affect the accuracy and practicality of subsequent analysis. In addition, traditional health risk assessment methods usually use static threshold judgments of single-dimensional features, which makes it difficult to combine time series data for trend prediction, and it is impossible to explore the complex correlations between multiple features, resulting in insufficient pertinence of health management recommendations. Finally, the interaction function between the health management system and the user is relatively simple, lacking a dynamic feedback mechanism. Users cannot obtain personalized health management recommendations based on real-time data updates, which seriously restricts the system's intelligence and the improvement of user experience. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to realize multimodal collection and fusion of vital signs data in the process of medical information processing, ensure the privacy and integrity of data transmission and storage, improve the accuracy of health risk assessment through intelligent analysis methods, generate personalized health management recommendations, and realize dynamic feedback and visual display at the same time, so as to meet the real-time and intelligent requirements of medical information technology in health management.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for automatically inputting vital signs detection, which comprises the following steps: The vital sign data of the target individual is collected through sensors and processed synchronously; Perform data cleaning and multi-channel data fusion on the collected data to form a complete vital signs data set; Encrypt the vital signs data set and upload it to the medical information system to complete data classification storage and update; Extract and analyze features of classified and stored data, and use classification models and regression models to analyze feature sets to generate health risk prediction results; Generate health management recommendations based on prediction results, and display and transmit data through a dynamic feedback mechanism.

[0007] As a preferred embodiment of the method for automatically recording vital signs detection of the present invention, the vital signs data of the target individual collected by the sensor includes: Perform multi-channel sampling of heart rate, blood pressure, body temperature and blood oxygen data, add timestamps to the sampled data in each sampling cycle, and align the time bases of different sensors; After sampling is completed, the mean and standard deviation of the data segment are calculated for each channel data to determine whether the data fluctuation is within the set range; If the fluctuation exceeds the range, the corresponding data is resampled, and the resampled data is concatenated with the original data in chronological order to form a multi-channel original data set.

[0008] As a preferred solution of the method for automatically inputting vital signs detection described in the present invention, the data cleaning and multi-channel data fusion of the collected data to form a complete vital signs data set includes: Arrange the heart rate, blood pressure, body temperature and blood oxygen data collected from multiple sensors by timestamp, and check whether the four data of heart rate, blood pressure, body temperature and blood oxygen are fully recorded at each time point. If any data is missing at a certain time point, all data at that time point will be eliminated; Calculate the difference between the heart rate value at each time point and the heart rate value at the next time point. By calculating all time points in sequence, a sequence of heart rate changes is obtained. For any time point, the absolute value of the heart rate change difference between the previous several time points is selected and averaged to form the heart rate short-term change rate. The difference between the systolic and diastolic blood pressures of the blood pressure data at each time point was calculated to obtain the pulse pressure. The mean arterial pressure at each time point was obtained by calculating all time points in sequence. The temperature data was smoothed using the sliding window method to extract the temperature trend; Calculate the deviation range of the blood oxygen value at each time point from the local mean to form a blood oxygen fluctuation range sequence; Select any two data from heart rate, blood pressure, body temperature and blood oxygen, compare their change directions within a set time period, and calculate the proportion of time points with consistent change directions; when the proportion reaches the preset threshold, it is determined to be synchronized, and a correlation label is generated. The synchronization of all two-way combined channels is determined to generate a synchronization matrix; When the ratio is lower than the preset threshold, it is determined to be out of sync, and the data in the current time period is marked for removal or resampling; The short-term variability of heart rate, pulse pressure, mean arterial pressure, body temperature trend, blood oxygen fluctuation range sequence and synchronization labels between channels are integrated into a multimodal feature matrix, and each row of the matrix represents all the eigenvalues ​​at a time point.

[0009] As a preferred solution of the method for automatically inputting vital signs detection described in the present invention, wherein: the vital signs data set is encrypted and uploaded to the medical information system, and the data classification storage and update are completed, including: The generated feature set is divided into several data blocks according to the preset time period, and each data block contains all the feature values ​​and corresponding metadata in the current time period; The metadata includes a timestamp, a source device number, and a data type of the data block; For each data block, a unique symmetric encryption key is generated, and the data block is encrypted using the generated key to form a preliminary encrypted data set, forming a first encrypted feature data set, and the symmetric encryption key is encrypted using the public key of the recipient to form a second encrypted feature data set, and the metadata is encrypted using the public key of the recipient to form a third encrypted feature data set; Generate a checksum for each data block in the first encrypted feature data set using a hash algorithm, and upload the checksum to the medical information system; After the upload is completed, the receiver recalculates the verification code of the received first encrypted feature data set and compares it with the verification code before uploading. If the verification code is inconsistent, the data retransmission mechanism is automatically triggered, and only the damaged data block is re-uploaded; The receiver uses the private key to decrypt the second encrypted feature data set, recovers the symmetric encryption key, uses the private key to decrypt the third encrypted feature data set to recover the metadata, and uses the decrypted symmetric encryption key to decrypt the first encrypted feature data set to obtain the original feature value and metadata. The original feature values ​​and metadata are obtained through decryption and classified and stored in the corresponding database partitions; Each storage record is appended with a timestamp, device number, and synchronization tag to form an index; In the database, dynamic update rules are set. When the newly uploaded data contains feature values ​​of the same time period, the old data is automatically replaced and stored as a historical version. When the uploaded new data reaches the set threshold size or covers a new time period, the update mechanism is automatically triggered to synchronize the latest data to the database.

[0010] As a preferred solution of the method for automatically inputting vital signs detection described in the present invention, wherein: the feature extraction and analysis of the classified stored data, and the use of the classification model and the regression model to analyze the feature set to generate the health risk prediction result include: Extract all feature records within the time period specified by the target user from the database classified and stored in the medical information system, and align the matrix data of different time periods by timestamps to form a continuous time series feature set; Based on the extracted time series feature set, the optimization process is performed, the correlation between features is calculated using the Pearson correlation coefficient method, and the correlation judgment criteria are set, including significant correlation, weak correlation, and no significant correlation; Combine the correlation with the synchronization label generated after decryption to analyze the physiological logic consistency between the features; For the significantly correlated feature pairs, they are marked as having good logical consistency and retained. For the weakly correlated feature pairs, their logical rationality is further verified by combining the physiological significance and synchronization labels of the features, and they are selectively retained or reduced in dimension according to specific analysis requirements. For the feature pairs without significant correlation, redundant features are removed; the feature matrix is ​​optimized; The optimized feature matrix is ​​input into the health risk prediction model.

[0011] As a preferred solution of the method for automatically inputting vital signs detection described in the present invention, wherein: the feature extraction and analysis of the classified stored data, and the use of the classification model and the regression model to analyze the feature set to generate the health risk prediction result also include: The health risk prediction model includes a classification model and a regression model; The classification model performs multi-level classification on the feature data at each time point according to the input of the training rules and feature matrix to generate a health status category label; The training rules include whether the feature value exceeds the standard health threshold, the joint abnormality between the features, and the comparison between the features and the historical health data, and the classification accuracy of the classification results is optimized through the logistic regression model; The regression model uses an autoregressive moving average algorithm to predict the changing trend of target characteristics in the future time period based on time series characteristics; The input of the regression model includes the feature matrix and historical health data in the current time period, and the output is the future change value of the target feature; The health risk prediction model combines the classification and regression results to generate a risk score for each time point of the feature matrix; The risk score includes: the risk level of abnormal short-term heart rate change rate, the risk level of abnormal blood oxygen fluctuation and body temperature trend, and the trend prediction score of mean arterial pressure deviation from normal value; Generate a health risk report based on the risk scores and trend prediction results output by the prediction model; The health risk report includes the risk level of the current health status, the health change trend in a specific time period in the future, and personalized intervention suggestions for different risk levels.

[0012] As a preferred solution of the method for automatically inputting vital signs detection described in the present invention, the method of generating health management suggestions according to the prediction results and completing the display and transmission of data through a dynamic feedback mechanism includes: Extract the risk level of each feature from the risk score and trend prediction results output by the health risk prediction model; combine the user's historical health data, match the risk level with the preset health management rules, and generate personalized health management suggestions; The generated health management suggestions are displayed in a visual way, with the real-time values ​​and change trends of heart rate, blood pressure, body temperature and blood oxygen displayed on the dashboard; the current health status is highlighted through risk level color identification, and the time series trend of health characteristics and future change predictions are displayed through line charts and bar charts; The health risk prediction model is called based on the newly entered data to dynamically adjust the risk score and trend prediction results; new health management recommendations are generated based on the adjusted prediction results to prompt users of their updated health status.

[0013] Another object of the present invention is to provide an automatically input vital signs detection system, which can organically combine multimodal data acquisition, encryption and storage, risk assessment and feedback mechanism, and can dynamically analyze health risks and generate targeted health management suggestions by collecting vital signs data in real time, thereby effectively solving the problems of lag, one-sidedness and lack of adaptability of existing technologies in the field of intelligent health management.

[0014] In order to solve the above technical problems, the present invention provides the following technical solutions: an automatic input vital sign detection system, comprising: a data acquisition and synchronization processing module, a data processing and storage module, a health risk analysis and prediction module, and a dynamic feedback and visualization display module; The data acquisition and synchronization processing module collects the vital sign data of the target individual through sensors, completes multi-channel sampling, adds timestamps during the sampling process, and performs time base alignment on the data of different sensors; The data processing and storage module cleans abnormal data points, removes incomplete records, completes the synchronization determination of multi-channel data, generates a complete vital sign data set and extracts key features, uploads the encrypted data to the medical information system through a combination of symmetric encryption and asymmetric encryption, completes classified storage, and adds a dynamic update mechanism; The health risk analysis and prediction module uses a classification model to evaluate the health status at each time point, generates a health status label, uses a regression model to predict the change trend of target characteristics in the future time period, generates a personalized risk score based on the classification and regression results, and generates a health risk report by combining the risk score and trend prediction; The dynamic feedback and visual display module is a dynamic closed loop constructed by data display and interaction, so that health management suggestions are updated in real time and presented to users.

[0015] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method for automatically recording vital signs detection when executing the computer program.

[0016] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for automatically recording vital signs detection.

[0017] Beneficial effects of the present invention: The present invention protects the privacy and integrity of vital signs data through a combination of symmetric encryption and asymmetric encryption, as well as a dynamic update and verification mechanism, and overcomes the risks of data leakage, tampering and chaotic version management in traditional systems. The present invention solves the limitations of traditional single-dimensional static evaluation by introducing an intelligent analysis method that combines classification models and regression models. It can comprehensively analyze the logical associations between multimodal feature data, generate comprehensive health risk reports including risk scores and trend forecasts, and provide higher foresight and accuracy for disease warning and health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them: Figure 1 An overall flow chart of an automatic vital sign detection method provided in Example 1 of the present invention; Figure 2 A comparison chart of load balancing rates of multiple algorithms under the same task request in an automatic vital sign detection method provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0020] Example 1, reference Figure 1 According to an embodiment of the present invention, there is provided a method for detecting vital signs with automatic recording, comprising: The vital sign data of the target individual is collected through sensors and processed synchronously; Perform data cleaning and multi-channel data fusion on the collected data to form a complete vital signs data set; Encrypt the vital signs data set and upload it to the medical information system to complete data classification storage and update; Extract and analyze features of classified and stored data, and use classification models and regression models to analyze feature sets to generate health risk prediction results; Generate health management recommendations based on prediction results, and display and transmit data through a dynamic feedback mechanism.

[0021] Perform multi-channel sampling of heart rate, blood pressure, body temperature and blood oxygen data, add timestamps to the sampled data in each sampling cycle, and align the time bases of different sensors; After sampling is completed, the mean and standard deviation of the data segment are calculated for each channel data to determine whether the data fluctuation is within the set range; If the fluctuation exceeds the range, the corresponding data is resampled, and the resampled data is concatenated with the original data in chronological order to form a multi-channel original data set.

[0022] It is further explained that: The heart rate, blood pressure, body temperature and blood oxygen of a target individual are collected synchronously through sensors, with a sampling period of every 5 seconds. During the collection process, the data are as follows: Heart rate (bpm): 72, 74, 95, 75, 73; Blood pressure (mmHg, systolic): 120, 122, 140, 119, 121; Body temperature (℃): 36.5, 36.6, 37.9, 36.4, 36.6; Blood oxygen (%): 98, 97, 92, 98, 99; Fluctuation range setting: Heart rate fluctuation range: ±5bpm; Blood pressure fluctuation range: ±10mmHg; Body temperature fluctuation range: ±0.5℃; Blood oxygen fluctuation range: ±2%; Determine whether the fluctuation is out of range: Heart rate: The third sample value is 95, which is outside the fluctuation range (74±5=69~79) and is marked as abnormal.

[0023] Blood pressure: The third sample value is 140, which is outside the fluctuation range (122±10=112~132) and is marked as abnormal.

[0024] Body temperature: The third sampling value was 37.9, which was beyond the fluctuation range (36.6±0.5=36.1~37.1) and was marked as abnormal.

[0025] Blood oxygen: The third sampling value is 92, which is beyond the fluctuation range (97±2=95~99) and is marked as abnormal.

[0026] Perform resampling: Resample each outlier point and generate the following data: Heart rate: resampled value is 76; Blood pressure: resampled value is 125; Body temperature: resampled value is 36.8; Blood oxygen: resampling value is 96; Splice in chronological order, insert the resampled values ​​into the corresponding time points, and generate the final multi-channel original data set: Heart rate (bpm): 72, 74, 76, 75, 73; Blood pressure (mmHg): 120, 122, 125, 119, 121; Body temperature (℃): 36.5, 36.6, 36.8, 36.4, 36.6; Blood oxygen (%): 98, 97, 96, 98, 99; The heart rate, blood pressure, body temperature and blood oxygen data collected from multiple sensors are arranged by timestamp, and it is checked whether the four data of heart rate, blood pressure, body temperature and blood oxygen are fully recorded at each time point. If any one of the data is missing at a certain time point, all the data at that time point will be discarded.

[0027] Calculate the difference between the heart rate value at each time point and the heart rate value at the next time point. By calculating all time points in sequence, a sequence of heart rate changes is obtained. For any time point, the absolute value of the heart rate change difference between the previous several time points is selected and averaged to form the short-term change rate of heart rate.

[0028] The pulse pressure was obtained by calculating the difference between the systolic pressure and the diastolic pressure of the blood pressure data at each time point. The mean arterial pressure at each time point was obtained by calculating all time points in sequence.

[0029] The sliding window method is used to smooth the body temperature data and extract the body temperature trend.

[0030] The deviation range of the blood oxygen value at each time point from the local mean is calculated to form a blood oxygen fluctuation range sequence.

[0031] Select any two data of heart rate, blood pressure, body temperature and blood oxygen, compare their change directions within the set time period, and calculate the proportion of time points with consistent change directions; when the proportion reaches the preset threshold, it is determined to be synchronized, and a correlation label is generated. The synchronization of all two-way combinations of channels is determined to generate a synchronization matrix.

[0032] When the ratio is lower than a preset threshold, it is determined to be out of sync, and the data in the current time period is marked for removal or resampling.

[0033] The short-term variability of heart rate, pulse pressure, mean arterial pressure, body temperature trend, blood oxygen fluctuation range sequence and synchronization labels between channels are integrated into a multimodal feature matrix, and each row of the matrix represents all the eigenvalues ​​at a time point.

[0034] It is further explained that: A high pulse pressure may indicate arteriosclerosis and decreased vascular elasticity, increasing the risk of cardiovascular events.

[0035] A low pulse pressure may indicate inadequate cardiac output, indicating a potential risk of heart failure.

[0036] Mean arterial pressure (MAP) is a key indicator for measuring systemic organ perfusion pressure: A mean arterial pressure (MAP) below 60 mmHg may result in insufficient blood supply to vital organs such as the brain, heart, and kidneys.

[0037] An increase in mean arterial pressure (MAP) indicates excessive vascular resistance, which may be related to hypertension.

[0038] Pulse pressure reflects arterial elasticity and pulse wave propagation characteristics, and more reflects the health of local blood vessels. Mean arterial pressure reflects the blood supply pressure and resistance of the systemic circulatory system, and is more used to evaluate organ perfusion levels.

[0039] It may be difficult to determine whether a patient has abnormalities in vascular resistance or cardiac output based on pulse pressure alone.

[0040] MAP provides information about the overall circulatory status and, combined with pulse pressure, can provide a more complete description of hemodynamic characteristics. For example: An increase in MAP and a normal pulse pressure indicate an increase in global vascular resistance with no significant changes in local elasticity.

[0041] The increased pulse pressure and normal MAP indicate a decrease in local vascular elasticity but no significant increase in overall vascular resistance.

[0042] By placing pulse pressure and MAP into a matrix, it is possible to observe their dynamic changes in the time dimension and identify trends and fluctuations.

[0043] Correlation analysis of pulse pressure, MAP, heart rate, body temperature and other characteristics can more accurately determine the physiological logic between vital signs. For example: An increased heart rate and pulse pressure suggest an increase in cardiac output, which may be related to the stress state.

[0044] An elevated MAP with a normal pulse pressure suggests elevated global vascular resistance, which may be associated with hypertension.

[0045] The generated feature set is divided into several data blocks according to the preset time period, and each data block contains all the feature values ​​and corresponding metadata in the current time period.

[0046] The metadata includes the data block's timestamp, source device number, and data type.

[0047] For each data block, a unique symmetric encryption key is generated, and the data block is encrypted using the generated key to form a preliminary encrypted data set, forming a first encrypted feature data set, and the symmetric encryption key is encrypted using the public key of the recipient to form a second encrypted feature data set. The metadata is encrypted using the public key of the recipient to form a third encrypted feature data set.

[0048] A check code is generated for each data block in the first encrypted feature data set using a hash algorithm and uploaded to the medical information system.

[0049] After the upload is completed, the receiver recalculates the verification code for the received first encrypted feature data set and compares it with the verification code before uploading. If the verification code is inconsistent, the data retransmission mechanism is automatically triggered and only the damaged data block is re-uploaded.

[0050] The receiver uses the private key to decrypt the second encrypted feature data set, recover the symmetric encryption key, uses the private key to decrypt the third encrypted feature data set to recover the metadata, and uses the decrypted symmetric encryption key to decrypt the first encrypted feature data set to obtain the original feature value and metadata.

[0051] The original feature values ​​and metadata are obtained through decryption and classified and stored in the corresponding database partitions; A timestamp, device number, and synchronization tag are attached to each storage record to form an index.

[0052] In the database, dynamic update rules are set. When the newly uploaded data contains feature values ​​of the same time period, the old data is automatically replaced and stored as a historical version. When the uploaded new data reaches the set threshold size or covers a new time period, the update mechanism is automatically triggered to synchronize the latest data to the database.

[0053] It is further explained that: Symmetric encryption uses a single key for encryption and decryption, and the algorithm complexity is low, which is suitable for processing large quantities of data, such as the multimodal feature data blocks generated in the present invention. The encryption and decryption time is relatively short, which is suitable for vital sign data processing scenarios with high real-time requirements. However, symmetric encryption requires both parties to share a key, and the key can be easily stolen during transmission, resulting in the risk of data leakage. Symmetric encryption itself does not solve the problem of data tampering, and needs to be used in conjunction with other mechanisms (such as hash verification).

[0054] Asymmetric encryption uses a public key and a private key pair. The key does not need to be shared. Even if it is intercepted during transmission, the attacker cannot decrypt it. It is particularly suitable for encrypting and protecting important information such as keys and metadata to prevent information from being tampered with or leaked during transmission. However, the algorithm is relatively complex, and the speed of asymmetric encryption is much lower than that of symmetric encryption, making it unsuitable for encrypting large amounts of data.

[0055] In the present invention, the characteristics of vital sign data are large data volume and high sensitivity. Using symmetric encryption or asymmetric encryption alone cannot meet the requirements.

[0056] If only symmetric encryption is used, although it is highly efficient, there are security risks in the transmission of keys.

[0057] If only asymmetric encryption is used, although security is high, the high computational overhead of the encryption and decryption process will lead to a decrease in system efficiency.

[0058] The multimodal feature data block in the present invention contains a large number of feature values, and the use of symmetric encryption processing can complete encryption with low computing overhead, ensuring high efficiency of data encryption. The generated symmetric encryption key is encrypted by the public key of the recipient and transmitted with the data. Even if it is intercepted during transmission, the attacker cannot decrypt the key, thereby ensuring the security of the symmetric encrypted data block. Metadata (such as timestamps, device numbers) are also protected by asymmetric encryption to prevent tampering during transmission, ensuring the accuracy of classified storage and traceability analysis. The encryption and decryption of data blocks are separated from the encryption and decryption of keys. Even if an attacker obtains part of the data blocks, he cannot decrypt other data blocks or restore the symmetric key, which can enhance the system's anti-attack capability.

[0059] All feature records within the time period specified by the target user are extracted from the database classified and stored in the medical information system, and the matrix data of different time periods are aligned according to timestamps to form a continuous time series feature set.

[0060] Based on the extracted time series feature set, optimization processing is performed, the correlation between features is calculated using the Pearson correlation coefficient method, and the correlation judgment criteria are set.

[0061] When the absolute value of the correlation coefficient is greater than 0.5, it is judged to be a significant correlation; when the absolute value of the correlation coefficient is between 0.2 and 0.5, it is judged to be a weak correlation; when the absolute value of the correlation coefficient does not exceed 0.2, it is judged to be no significant correlation.

[0062] The correlation is combined with the synchronization labels generated after decryption to perform physiological logic consistency analysis between features.

[0063] For significantly correlated feature pairs, they are marked as having good logical consistency and retained. For weakly correlated feature pairs, their logical rationality is further verified by combining the physiological significance and synchronization labels of the features, and they are selectively retained or reduced in dimension according to specific analysis needs. For feature pairs without significant correlation, redundant features are eliminated and the feature matrix is ​​optimized.

[0064] The optimized feature matrix is ​​input into the health risk prediction model.

[0065] It is further explained that: Although the weak correlation feature pairs show a low correlation between the features numerically, in some specific cases, this weak correlation may have potential physiological significance or diagnostic value. For example: Short-term fluctuations between heart rate and blood pressure after high-intensity exercise may show a weak correlation.

[0066] Blood oxygen and body temperature may have a low correlation in the early stages of certain diseases, but they have certain diagnostic value.

[0067] Health risk prediction models include classification models and regression models.

[0068] The classification model performs multi-level classification on the feature data at each time point according to the input of the training rules and the feature matrix to generate a health status category label.

[0069] The training rules include whether the feature value exceeds the standard health threshold, the joint abnormalities between features, and the comparison between features and historical health data. The classification accuracy is optimized through the logistic regression model.

[0070] The regression model uses the autoregressive moving average algorithm to predict the changing trend of target characteristics in the future time period based on the characteristics of the time series.

[0071] The input of the regression model includes the feature matrix and historical health data in the current time period, and the output is the future change value of the target feature.

[0072] The health risk prediction model combines the classification and regression results to generate a risk score for each time point of the feature matrix.

[0073] It is further explained that: The classification model outputs risk category weights. For each time point, the classification model assigns weights based on the health status category label, for example, normal status is 0 points, mild abnormality is 1 point, moderate abnormality is 2 points, and severe abnormality is 3 points.

[0074] The regression model outputs a trend deviation score. The regression model calculates the trend deviation risk based on the future feature change trend. For example, by evaluating the short-term rate of change of heart rate or the rate of change of blood pressure fluctuations, the degree of deviation is converted into a risk score according to the set rules.

[0075] For each target feature (such as heart rate, blood pressure, body temperature, and blood oxygen), set the normal range and classification thresholds for mild, moderate, and severe deviations based on medical standards and the user's historical health data.

[0076] For example, the normal range for heart rate is set at 60–100 bpm, with mild deviations being ±10%, moderate deviations being ±20%, and severe deviations being more than ±30%.

[0077] The regression model predicts the future feature value sequence based on time series data, for example, predicting the trend of heart rate changes in the next 10 minutes, 30 minutes or 1 hour.

[0078] The output prediction result is a set of continuous feature value sequences, and the predicted heart rate values ​​in the next 10 minutes are [75, 78, 81, 85, 88] bpm.

[0079] Compare the deviation of the predicted feature values ​​from the normal range: Deviation = predicted value - center value of normal range Deviation Rate

[0080] For example, if the predicted value is 88 bpm and the central value is 80 bpm, then the deviation rate is .

[0081] Map the deviation rate to a risk score according to the set rules: The deviation rate is If the score is less than 0, it will be assigned a score of 0, indicating no risk.

[0082] The deviation rate is to In between, a score of 1 is assigned, indicating a mild risk.

[0083] The deviation rate is to In between, a score of 2 is assigned, indicating moderate risk.

[0084] Deviation rate exceeds , assigned a score of 3, indicating severe risk.

[0085] The trend deviation score is generated by integrating the time dimension, and the corresponding risk score is calculated for all predicted values ​​in the future period, for example: .

[0086] All risk scores in the time dimension are averaged or weighted to generate a trend deviation score: Average risk score: Average the scores across all time points.

[0087] Weighted risk scores: assign higher weights to points in the future that are closer, for example:

[0088] Among them, the closer the time point is, the higher the weight is, such as: , , .

[0089] Assume that the normal range of heart rate is , the predicted value for the next five time points is , according to the set rules: First time point (72 bpm): Deviation rate , score = 0 (normal) Second time point (110 bpm): Deviation rate , score = 3 (severe risk) Time point 3 (115 bpm): Deviation rate , score = 3 (severe risk) The fourth time point (125 bpm): Deviation rate , score = 3 (severe risk) The 5th time point (130bpm): Deviation rate , score = 3 (severe risk) The final trend deviation score is calculated according to the weighted rule: Trend Deviation Score .

[0090] For each time point, the risk category weights of the classification model and the trend deviation scores of the regression model are combined according to the weighted rules to form a comprehensive risk score.

[0091] The comprehensive risk score includes: the risk level of abnormal short-term heart rate change rate, the risk level of abnormal blood oxygen fluctuation and body temperature trend, and the trend prediction score of mean arterial pressure deviation from normal values.

[0092] Generate a health risk report based on the risk score and trend prediction results output by the prediction model.

[0093] The health risk report includes the risk level of the current health status, the health change trend in a specific time period in the future, and personalized intervention recommendations for different risk levels.

[0094] Extract the risk level of each feature from the risk score and trend prediction results output by the health risk prediction model; combine the user's historical health data, match the risk level with the preset health management rules, and generate personalized health management suggestions; The generated health management suggestions are displayed in a visual way, with the real-time values ​​and change trends of heart rate, blood pressure, body temperature and blood oxygen displayed on the dashboard; the current health status is highlighted through risk level color identification, and the time series trend of health characteristics and future change predictions are displayed through line charts and bar charts; The health risk prediction model is called based on the newly entered data to dynamically adjust the risk score and trend prediction results; new health management recommendations are generated based on the adjusted prediction results to prompt users of their updated health status.

[0095] Example 2, reference Figure 2 As another embodiment of the present invention, a system for automatically recording vital signs detection method is provided, comprising: a data acquisition and synchronization processing module, a data processing and storage module, a health risk analysis and prediction module, and a dynamic feedback and visualization display module; The data acquisition and synchronization processing module collects the vital signs data of the target individual through sensors, completes multi-channel sampling, adds timestamps during the sampling process, and aligns the time base of the data from different sensors; The data processing and storage module is to clean abnormal data points, remove incomplete records, complete the synchronization judgment of multi-channel data, generate a complete vital sign data set and extract key features, and upload the encrypted data to the medical information system through a combination of symmetric encryption and asymmetric encryption, complete classified storage, and add a dynamic update mechanism; The health risk analysis and prediction module uses a classification model to evaluate the health status at each time point, generates a health status label, and uses a regression model to predict the change trend of target characteristics in the future time period. Based on the classification and regression results, it generates a personalized risk score, a comprehensive risk score and trend prediction, and generates a health risk report; The dynamic feedback and visual display module builds a dynamic closed loop of data display and interaction, so that health management suggestions can be updated in real time and presented to users.

[0096] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0097] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0098] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0099] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0100] Embodiment 3: In this embodiment, in order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments. This embodiment experiments the existing traditional method and the method of this embodiment respectively.

[0101] Simulated vital signs data were extracted from the public MIMIC-III database, including 24-hour multimodal sampling data of heart rate, blood pressure, body temperature, and blood oxygen. The data volume was 50 people × 24 hours × 1 sampling point per second, totaling 4.32 million records.

[0102] Although traditional methods also process multimodal data, they process multimodal data separately and lack the ability to align and synchronize the time base, resulting in weak correlation between data. Even if there is multimodal data, the dynamic correlation between features is difficult to reflect due to the lack of a fusion mechanism. Traditional methods usually only extract simple features (such as mean, maximum, minimum, etc.) for each vital sign data independently, without considering the synergistic effect of multimodal features, and the feature description is relatively one-sided. Traditional methods lack comprehensive analysis capabilities in health risk assessment, and usually judge risks based on the static threshold of a single vital sign, and are unable to capture the deep correlation between vital signs or predict future trends. The effect comparison of the data processing stage of the present invention is shown in Table 1 Table 1 Processing effect table of data collection stage

[0103] Traditional method (Solution 1): Encrypt data blocks using only symmetric encryption.

[0104] The present invention (Solution 2): combines symmetric encryption and asymmetric encryption mechanisms, encrypts data blocks, keys and metadata separately, and uses hash verification to ensure data integrity.

[0105] The data comparison results are shown in Table 2.

[0106] Table 2 Data comparison table

[0107] The invention adds asymmetric encryption processing, which increases the encryption and decryption time by about 30ms respectively. However, considering the security requirements of medical information data transmission, this small time increase is acceptable. By introducing a hash verification mechanism, the tampering or loss of data during transmission can be effectively detected and repaired, and the data integrity is increased from 87% to 98%, ensuring the reliability of sensitive data in the transmission link.

[0108] Traditional symmetric encryption schemes score low in man-in-the-middle attack simulations due to the inherent risks of key sharing. The present invention protects key transmission through asymmetric encryption, greatly reducing the possibility of interception and tampering.

[0109] Risk assessment is carried out through the encryption and decryption links. The data comparison is shown in Table 3.

[0110] Table 3 Core data comparison table

[0111] The present invention combines classification and regression models to analyze the logical associations between multimodal data, thereby increasing the accuracy of health risk assessment from 65% of traditional methods to 91%. This improvement is mainly due to: Dynamic correlation analysis between multimodal features (such as the synergistic relationship between the short-term change rate of heart rate and blood pressure fluctuations) introduces time series regression models to predict future trends and enhance the foresight of health assessment.

[0112] The time series regression model of the present invention significantly reduces the error of risk prediction, with the MSE reduced from 0.45 to 0.15. This shows that the present invention can more accurately predict the user's future health status, especially in capturing early disease trends. The traditional method has an error rate of up to 20% due to single feature judgment. The present invention reduces the error rate to 6% through multi-dimensional feature classification model optimization, providing users with a more reliable health risk classification.

[0113] It can be further seen that the dynamic feedback mechanism of the present invention significantly optimizes the efficiency of data processing and suggestion generation, shortening the average response time from the traditional 5 seconds to 1.3 seconds, providing a guarantee for users to understand their health status in real time and take intervention measures. Due to the lack of comprehensive analysis of historical data and prediction results, the traditional method has a personalized matching rate of only 50%. The present invention combines the user's historical health data, risk score and trend prediction results to increase the personalized suggestion matching rate to 85%, which can more accurately meet user needs.

[0114] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for automatically recording vital signs detection, characterized in that: include: The vital sign data of the target individual is collected through sensors and processed synchronously; Perform data cleaning and multi-channel data fusion on the collected data to form a complete vital signs data set; Encrypt the vital signs data set and upload it to the medical information system to complete data classification storage and update; Extract and analyze features of classified and stored data, and use classification models and regression models to analyze feature sets to generate health risk prediction results; Generate health management recommendations based on prediction results, and display and transmit data through a dynamic feedback mechanism.

2. The method for automatically recording vital signs according to claim 1, characterized in that: The collecting of vital sign data of the target individual by the sensor includes: Perform multi-channel sampling of heart rate, blood pressure, body temperature and blood oxygen data, add timestamps to the sampled data in each sampling cycle, and align the time bases of different sensors; After sampling is completed, the mean and standard deviation of the data segment are calculated for each channel data to determine whether the data fluctuation is within the set range; If the fluctuation exceeds the range, the corresponding data is resampled, and the resampled data is concatenated with the original data in chronological order to form a multi-channel original data set.

3. The method for automatically recording vital signs according to claim 2, characterized in that: The data cleaning and multi-channel data fusion of the collected data to form a complete vital sign data set includes: Arrange the heart rate, blood pressure, body temperature and blood oxygen data collected from multiple sensors by timestamp, and check whether the four data of heart rate, blood pressure, body temperature and blood oxygen are fully recorded at each time point. If any data is missing at a certain time point, all data at that time point will be eliminated; Calculate the difference between the heart rate value at each time point and the heart rate value at the next time point. By calculating all time points in sequence, a sequence of heart rate changes is obtained. For any time point, the absolute value of the heart rate change difference between the previous several time points is selected and averaged to form the heart rate short-term change rate. The difference between the systolic and diastolic blood pressures of the blood pressure data at each time point was calculated to obtain the pulse pressure. The mean arterial pressure at each time point was obtained by calculating all time points in sequence. Use the sliding window method to smooth the temperature data and extract the temperature trend; Calculate the deviation range of the blood oxygen value at each time point from the local mean to form a blood oxygen fluctuation range sequence; Select any two data from heart rate, blood pressure, body temperature and blood oxygen, compare their change directions within a set time period, and calculate the proportion of time points with consistent change directions; when the proportion reaches the preset threshold, it is determined to be synchronized, and a correlation label is generated. The synchronization of all two-way combined channels is determined to generate a synchronization matrix; When the ratio is lower than the preset threshold, it is determined to be out of sync, and the data in the current time period is marked for removal or resampling; The short-term variability of heart rate, pulse pressure, mean arterial pressure, body temperature trend, blood oxygen fluctuation range sequence and synchronization labels between channels are integrated into a multimodal feature matrix, and each row of the matrix represents all the eigenvalues ​​at a time point.

4. The method for automatically recording vital signs according to claim 3, characterized in that: The encryption of the vital signs data set and uploading it to the medical information system to complete the data classification storage and update includes: The generated feature set is divided into several data blocks according to the preset time period, and each data block contains all the feature values ​​and corresponding metadata in the current time period; The metadata includes a timestamp, a source device number, and a data type of the data block; For each data block, a unique symmetric encryption key is generated, and the data block is encrypted using the generated key to form a preliminary encrypted data set, forming a first encrypted feature data set, and the symmetric encryption key is encrypted using the public key of the recipient to form a second encrypted feature data set, and the metadata is encrypted using the public key of the recipient to form a third encrypted feature data set; Generate a checksum for each data block in the first encrypted feature data set using a hash algorithm, and upload the checksum to the medical information system; After the upload is completed, the receiver recalculates the verification code of the received first encrypted feature data set and compares it with the verification code before uploading. If the verification code is inconsistent, the data retransmission mechanism is automatically triggered, and only the damaged data block is re-uploaded; The receiver uses the private key to decrypt the second encrypted feature data set, recovers the symmetric encryption key, uses the private key to decrypt the third encrypted feature data set to recover the metadata, and uses the decrypted symmetric encryption key to decrypt the first encrypted feature data set to obtain the original feature value and metadata. The original feature values ​​and metadata are obtained through decryption and classified and stored in the corresponding database partitions; Each storage record is appended with a timestamp, device number, and synchronization tag to form an index; In the database, dynamic update rules are set. When the newly uploaded data contains feature values ​​of the same time period, the old data is automatically replaced and stored as a historical version. When the uploaded new data reaches the set threshold size or covers a new time period, the update mechanism is automatically triggered to synchronize the latest data to the database.

5. The method for automatically recording vital signs according to claim 4, characterized in that: The feature extraction and analysis of the classified stored data, and the use of classification models and regression models to analyze the feature set to generate health risk prediction results include: Extract all feature records within the time period specified by the target user from the database classified and stored in the medical information system, and align the matrix data of different time periods by timestamps to form a continuous time series feature set; Based on the extracted time series feature set, the optimization process is performed, the correlation between features is calculated using the Pearson correlation coefficient method, and the correlation judgment criteria are set, including significant correlation, weak correlation, and no significant correlation; Combine the correlation with the synchronization label generated after decryption to analyze the physiological logic consistency between the features; For the significantly correlated feature pairs, they are marked as having good logical consistency and retained. For the weakly correlated feature pairs, their logical rationality is further verified by combining the physiological significance and synchronization labels of the features, and they are selectively retained or reduced in dimension according to specific analysis requirements. For the feature pairs without significant correlation, redundant features are eliminated. Optimize feature matrix; The optimized feature matrix is ​​input into the health risk prediction model.

6. A method for automatically recording vital signs detection as claimed in claim 5, characterized in that: The extracting and analyzing the features of the classified stored data, and using the classification model and regression model to analyze the feature set to generate the health risk prediction result also includes: The health risk prediction model includes a classification model and a regression model; The classification model performs multi-level classification on the feature data at each time point according to the input of the training rules and feature matrix to generate a health status category label; The training rules include whether the feature value exceeds the standard health threshold, the joint abnormality between the features, and the comparison between the features and the historical health data, and the classification accuracy of the classification results is optimized through the logistic regression model; The regression model uses an autoregressive moving average algorithm to predict the changing trend of target characteristics in the future time period based on time series characteristics; The input of the regression model includes the feature matrix and historical health data in the current time period, and the output is the future change value of the target feature; The health risk prediction model combines the classification and regression results to generate a risk score for each time point of the feature matrix; The risk score includes: the risk level of abnormal short-term heart rate change rate, the risk level of abnormal blood oxygen fluctuation and body temperature trend, and the trend prediction score of mean arterial pressure deviation from normal value; Generate a health risk report based on the risk scores and trend prediction results output by the prediction model; The health risk report includes the risk level of the current health status, the health change trend in a specific time period in the future, and personalized intervention suggestions for different risk levels.

7. A method for automatically recording vital signs detection as claimed in claim 6, characterized in that: The generating of health management suggestions based on the prediction results and the display and transmission of data through a dynamic feedback mechanism include: Extract the risk level of each feature from the risk score and trend prediction results output by the health risk prediction model; Combine the user's historical health data, match the risk level with the preset health management rules, and generate personalized health management suggestions; The generated health management suggestions are displayed in a visual way, with the real-time values ​​and change trends of heart rate, blood pressure, body temperature and blood oxygen displayed on the dashboard; the current health status is highlighted through risk level color identification, and the time series trend of health characteristics and future change predictions are displayed through line charts and bar charts; The health risk prediction model is called based on the newly entered data to dynamically adjust the risk score and trend prediction results; new health management recommendations are generated based on the adjusted prediction results to prompt users of their updated health status.

8. A system using the method for automatically recording vital signs according to any one of claims 1 to 7, characterized in that: It includes data collection and synchronization processing module, data processing and storage module, health risk analysis and prediction module, dynamic feedback and visualization display module; The data acquisition and synchronization processing module collects the vital sign data of the target individual through sensors, completes multi-channel sampling, adds timestamps during the sampling process, and performs time base alignment on the data of different sensors; The data processing and storage module cleans abnormal data points, removes incomplete records, completes the synchronization determination of multi-channel data, generates a complete vital sign data set and extracts key features, uploads the encrypted data to the medical information system through a combination of symmetric encryption and asymmetric encryption, completes classified storage, and adds a dynamic update mechanism; The health risk analysis and prediction module uses a classification model to evaluate the health status at each time point, generates a health status label, uses a regression model to predict the change trend of target characteristics in the future time period, generates a personalized risk score based on the classification and regression results, and generates a health risk report by combining the risk score and trend prediction; The dynamic feedback and visual display module is a dynamic closed loop constructed by data display and interaction, so that health management suggestions are updated in real time and presented to users.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for automatically recording vital signs detection according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for automatically recording vital signs detection as described in any one of claims 1 to 7 are implemented.

Images