Early warning system for disease trend changes
Through the early warning system for the change of the disease, using physiological indicators and drug use information collection and prediction models, the problem of difficult to provide real-time and accurate disease prediction in the existing technology is solved, real-time monitoring of the health status of the guardian and early warning of the disease, and improving medical care efficiency.
Patent Information
- Application Number
- CN202411135195.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The prior art is difficult to provide guardians with real-time and accurate predictions and warnings on the trend of disease change, especially in real-time physiological indicators of different guardians.
Design a warning system for the change of disease, collecting multiple physiological indicator data and medication information of the guardian, using historical data and prediction models to predict, generate disease status scores and provide warning information, and ensure real-time monitoring of the health status of the guardian and Abnormal situations will be discovered in time.
Real-time and accurate prediction and warning of changes in the disease are achieved, and intervention measures or adjustment of treatment plans are taken in a timely manner to reduce the possibility of worsening of the disease, improve medical efficiency, and improve prediction accuracy by optimizing the prediction model of physiological index data.
Smart Images

Figure CN118658644B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to an early warning system for disease condition change trends. Background Art
[0002] In the related art, CN110970132A discloses a disease warning system based on mobile nursing, which includes: a medical data acquisition module, a cloud storage, a disease analysis module and a mobile nursing platform; the medical data acquisition module obtains patient information; the cloud storage receives patient information and caches it; the disease analysis module obtains patient information from the cloud storage, analyzes the patient's vital signs, determines the patient's physical condition, and sends the corresponding results to the mobile nursing platform; the mobile nursing platform receives the results and displays them; and is also used to alarm when the results show that the patient's physical condition is abnormal. Compared with directly diagnosing the patient's condition through the doctor's medical knowledge, this system is based on the acquired patient's physiological signs information and the pre-stored standard symptom information of various diseases, and then diagnoses the patient's physical condition, avoiding the influence of the doctor's subjective cognition brought by the doctor's medical experience, and can more objectively and accurately reflect the patient's true physical condition.
[0003] CN113362941A discloses a method and system for disease monitoring based on big data, which is applied to a disease monitoring system, wherein the method includes: obtaining monitoring data of a first professional monitoring point of the disease monitoring system; constructing a first disease retrieval database of first infectious disease information; for a first infectious disease type; inputting it into the first disease retrieval database for retrieval, obtaining a second infectious disease type with a high correlation with the first infectious disease type in the first disease retrieval database; then obtaining a first disease development trend from the first retrieval result; and obtaining a first disease prevention and control plan based on the first disease development trend. The method solves the technical problems of insufficient accuracy of monitoring data in the disease monitoring system in the prior art and untimely response to sudden disease.
[0004] Based on the above-mentioned related technologies, the condition can be monitored. However, when monitoring the condition of different guardians, it is difficult to predict the condition based on the real-time physiological indicators of the guardians, making it difficult to provide guardians with real-time and accurate predictions and warnings of the trend of condition changes.
[0005] The information disclosed in the background technology section of this application is only intended to deepen the understanding of the general background technology of this application, and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art. Summary of the invention
[0006] The present invention provides a disease condition change trend early warning system, which can solve the technical problem of difficulty in providing real-time and accurate disease condition change trend prediction and warning to guardians.
[0007] According to the present invention, a disease condition change trend early warning system is provided, comprising:
[0008] The collection module is used to collect multiple physiological index data of the guardian at multiple moments in the i-th monitoring cycle, and obtain the medication information of the guardian in the i-th monitoring cycle, wherein the medication information includes the type of medication used by the guardian and the dosage information of each type of medication, and i is a positive integer;
[0009] An acquisition module is used to acquire historical physiological index data of multiple monitoring cycles before the i-th monitoring cycle, as well as historical medication information;
[0010] The first prediction module is used to predict the historical physiological indicator data of the historical monitoring period according to the historical physiological indicator data and historical medication information of the historical monitoring period, the medication information of the i-th monitoring period, and the physiological indicator data prediction model of the i-1-th monitoring period, to obtain the historical physiological indicator training data, and to predict the physiological indicator data in the i-th monitoring period to obtain the physiological indicator training data;
[0011] A judgment module, used for judging whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained according to the physiological indicator training data and the physiological indicator data;
[0012] A training module, used for training the physiological indicator data prediction model of the i-1th monitoring cycle according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data, to obtain the physiological indicator data prediction model of the i-th monitoring cycle when the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained;
[0013] A second prediction module is used to obtain predicted physiological indicator data at multiple moments of a future monitoring cycle based on the physiological indicator data, the historical physiological indicator data, the medication information, the historical medication information, the planned medication information of the future monitoring cycle, and the physiological indicator data prediction model of the i-th monitoring cycle;
[0014] A disease condition prediction module, used to determine the disease condition score of a future monitoring period based on the predicted physiological index data;
[0015] The early warning module is used to generate early warning information according to the disease condition score.
[0016] Technical effect: According to the present invention, by collecting the physiological indicator data and medication information of the monitoring personnel, the health status of the personnel can be monitored in real time and abnormal conditions can be discovered in time. Through the physiological indicator data prediction model, real-time and accurate prediction and warning of the trend of disease changes are provided. Early warning information is generated according to the disease condition score, so that intervention measures or adjustment of the treatment plan can be taken in time to reduce the possibility of disease deterioration and improve medical efficiency. When determining whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained, the relative deviation score of the physiological indicator data can be obtained based on the weighted average of the nkth relative deviation, and when determining the weight, the data volume of the referenced historical physiological indicators and the time difference between the monitoring cycle corresponding to the referenced historical physiological indicators and the i-th monitoring cycle are considered. Various factors that may affect the prediction accuracy, such as the time difference between the monitoring cycle corresponding to the referenced historical physiological indicators and the i-th monitoring cycle, are considered to improve the accuracy and objectivity of the relative deviation score, and then it can be determined whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained, thereby continuously optimizing the physiological indicator data prediction model and improving the accuracy of the physiological indicator data prediction model of the i-1th monitoring cycle. When training the physiological indicator data prediction model of the i-1 monitoring cycle, the relative difference between the physiological indicator data, the historical physiological indicator data and the physiological indicator training data and the historical physiological indicator training data output by the physiological indicator data prediction model of the i-1 monitoring cycle can be determined, and the weights can be set based on the characteristics that the larger the time difference, the lower the prediction accuracy, and the closer to the required monitoring cycle, the greater the reference value and the more reference data, the more accurate the prediction result, so as to weight the relative errors of the physiological indicator training data and the historical physiological indicator training data output by the physiological indicator data prediction model of the i-1 monitoring cycle to obtain the loss function, so as to improve the design accuracy and objectivity of the loss function, thereby improving the training efficiency during the training process. When determining the disease condition score, the disease condition score can be determined by predicting the relative deviation of the physiological indicator data, the physiological indicator data and the standard indicator data, and the weights set based on the time difference between the future monitoring cycle and the current monitoring, so as to predict the possible changes in the condition of the guardian in the future period in advance, which is helpful to take timely treatment measures.
[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only and do not limit the present invention. Other features and aspects of the present invention will become more apparent from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other embodiments can be obtained based on these drawings without creative work.
[0019] Figure 1 A block diagram of a disease condition change trend early warning system according to an embodiment of the present invention is exemplarily shown. DETAILED DESCRIPTION
[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0021] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
[0022] Figure 1 A block diagram of a disease condition change trend early warning system according to an embodiment of the present invention is exemplarily shown, and the system includes:
[0023] The collection module is used to collect multiple physiological index data of the guardian at multiple moments in the i-th monitoring cycle, and obtain the medication information of the guardian in the i-th monitoring cycle, wherein the medication information includes the type of medication used by the guardian and the dosage information of each type of medication, and i is a positive integer;
[0024] An acquisition module is used to acquire historical physiological index data of multiple monitoring cycles before the i-th monitoring cycle, as well as historical medication information;
[0025] The first prediction module is used to predict the historical physiological indicator data of the historical monitoring period according to the historical physiological indicator data and historical medication information of the historical monitoring period, the medication information of the i-th monitoring period, and the physiological indicator data prediction model of the i-1-th monitoring period, to obtain the historical physiological indicator training data, and to predict the physiological indicator data in the i-th monitoring period to obtain the physiological indicator training data;
[0026] A judgment module, used for judging whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained according to the physiological indicator training data and the physiological indicator data;
[0027] A training module, used for training the physiological indicator data prediction model of the i-1th monitoring cycle according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data, to obtain the physiological indicator data prediction model of the i-th monitoring cycle when the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained;
[0028] A second prediction module is used to obtain predicted physiological indicator data at multiple moments of a future monitoring cycle based on the physiological indicator data, the historical physiological indicator data, the medication information, the historical medication information, the planned medication information of the future monitoring cycle, and the physiological indicator data prediction model of the i-th monitoring cycle;
[0029] A disease condition prediction module, used to determine the disease condition score of a future monitoring period based on the predicted physiological index data;
[0030] The early warning module is used to generate early warning information according to the disease condition score.
[0031] According to the disease condition change trend warning system of the embodiment of the present invention, by collecting the physiological index data and medication information of the monitored personnel, the health status of the monitored personnel can be monitored in real time and abnormal conditions can be discovered in time. Through the physiological index data prediction model, real-time and accurate disease condition change trend prediction and warning are provided. Early warning information is generated according to the disease condition score, so that intervention measures or treatment plans can be taken in time to reduce the possibility of disease deterioration and improve medical efficiency.
[0032] According to one embodiment of the present invention, each monitoring cycle can be set to one day, and the interval between adjacent moments can be set to 3 hours, 6 hours, etc., which is not limited by the present invention. The acquisition module can collect multiple physiological index data of the guardian, such as blood pressure, heart rate, body temperature and other physiological parameters. Obtain the medication information of the guardian in the i-th monitoring cycle, including recording the type of drugs (for example, antihypertensive drugs) being used by the guardian, and the dosage information of each drug (including dosage, frequency of administration, etc.). The medication information of the guardian will affect multiple physiological index data.
[0033] According to an embodiment of the present invention, the acquisition module can acquire historical physiological indicator data of multiple monitoring cycles before the i-th monitoring cycle, as well as historical medication information, to provide important references for subsequent analysis and comparison.
[0034] According to one embodiment of the present invention, the physiological indicator data prediction model may be a neural network model. The first prediction module may predict the historical physiological indicator data of the historical monitoring period by inputting the historical physiological indicator data and historical medication information of the historical monitoring period and the medication information of the i-th monitoring period into the physiological indicator data prediction model of the i-1-th monitoring period, thereby obtaining the historical physiological indicator training data. Similarly, the physiological indicator data within the i-th monitoring period may be predicted to obtain the physiological indicator training data.
[0035] According to one embodiment of the present invention, based on the historical physiological indicator data and historical medication information of the historical monitoring period, the medication information of the i-th monitoring period, and the physiological indicator data prediction model of the i-1-th monitoring period, the historical physiological indicator data of the historical monitoring period is predicted to obtain historical physiological indicator training data, and the physiological indicator data within the i-th monitoring period is predicted to obtain physiological indicator training data, including: the historical physiological indicator data and historical medication information from the i-th monitoring period to the i-n+k-th monitoring period, the i-n+k+1-th monitoring period to the i-1-th monitoring period The historical medication information of the monitoring cycle and the medication information of the i-th monitoring cycle are input into the physiological indicator data prediction model of the i-1-th monitoring cycle to obtain the historical physiological indicator training data from the i-n+k+1-th monitoring cycle to the i-1-th monitoring cycle based on the historical physiological indicator data from the in-th monitoring cycle to the i-n+k-th monitoring cycle, and the physiological indicator training data of the i-th monitoring cycle based on the historical physiological indicator data from the in-th monitoring cycle to the i-n+k-th monitoring cycle, wherein 0≤k≤n-1, k is an integer, n is a positive integer, and n<i.
[0036] According to one embodiment of the present invention, the training data obtained is different based on different monitoring cycles. For example, when i=20, n=10, and k=5, the historical physiological indicator data and historical medication information from the 10th monitoring cycle to the 15th monitoring cycle, the historical medication information from the 16th monitoring cycle to the 19th monitoring cycle, and the medication information of the 20th monitoring cycle are input into the physiological indicator data prediction model of the 19th monitoring cycle to obtain the historical physiological indicator training data from the 16th monitoring cycle to the 19th monitoring cycle based on the historical physiological indicator data from the 10th monitoring cycle to the 15th monitoring cycle, and the physiological indicator training data of the 20th monitoring cycle based on the historical physiological indicator data from the 10th monitoring cycle to the 15th monitoring cycle. When i=20, n=10, k=9, the historical physiological indicator data and historical medication information from the 10th monitoring cycle to the 19th monitoring cycle, as well as the medication information of the 20th monitoring cycle, are input into the physiological indicator data prediction model of the 19th monitoring cycle to obtain the physiological indicator training data of the 20th monitoring cycle based on the historical physiological indicator data from the 10th monitoring cycle to the 19th monitoring cycle.
[0037] According to one embodiment of the present invention, the judgment module can compare the physiological indicator training data and the physiological indicator data, that is, analyze the error between the predicted result of the physiological indicator data prediction model of the i-1th monitoring cycle for the physiological indicator data of the i-1th detection cycle and the actual data, so as to judge whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained.
[0038] According to one embodiment of the present invention, based on the physiological indicator training data and the physiological indicator data, it is determined whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained, including: obtaining the nkth relative deviation between the physiological indicator training data of the i-th monitoring cycle based on the historical physiological indicator data from the in-th monitoring cycle to the i-n+k-th monitoring cycle and the physiological indicator data; based on the nkth relative deviation, obtaining a relative deviation score of the physiological indicator data; if the relative deviation score is greater than or equal to a preset deviation threshold, it is determined that the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained.
[0039] According to one embodiment of the present invention, by comparing the physiological indicator data of the ith monitoring cycle with the physiological indicator data prediction model of the i-1th monitoring cycle, the physiological indicator training data of the ith monitoring cycle based on the historical physiological indicator data from the inth monitoring cycle to the i-n+kth monitoring cycle is calculated, and the value interval of k is [0, n-1]. Therefore, the first relative deviation, the second relative deviation...the nth relative deviation, a total of n relative deviations, can be obtained. According to the comparison result of the relative deviation score and the preset deviation threshold, it is determined whether it is necessary to train the physiological indicator data prediction model of the i-1th monitoring cycle. If the relative deviation score is greater than or equal to the preset deviation threshold, it is determined that the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained.
[0040] According to an embodiment of the present invention, obtaining the relative deviation score of the physiological indicator data according to the nkth relative deviation includes: obtaining the relative deviation score E of the physiological indicator data according to formula (1),
[0041] (1)
[0042] in, is the sth physiological index data at the jth moment of the ith monitoring cycle, is the sth type of physiological indicator training data at the jth moment of the ith monitoring cycle based on the historical physiological indicator data from the inth monitoring cycle to the i-n+kth monitoring cycle, M is the number of types of physiological indicator data, N is the number of moments in each monitoring cycle, is the n-th relative deviation of the physiological indicator data, s≤M, j≤N, and s, j, M and N are all positive integers.
[0043] According to one embodiment of the present invention, in formula (1), It is the average value of the nkth relative deviations of various physiological index data. is the total number of moments in n monitoring cycles before the current cycle, is the number of moments between the last moment of the i-n+kth monitoring cycle and the jth moment of the i-th monitoring cycle. The smaller the number of moments, the shorter the time interval between the last moment of the i-n+kth monitoring cycle and the jth moment of the i-th monitoring cycle, the more reliable the reference data, and the more accurate the prediction result. Therefore, the larger the weight, The relative difference between the total number of moments in the n monitoring cycles before the current cycle and the number of moments between the last moment of the i-n+kth monitoring cycle and the jth moment of the i-th monitoring cycle can be used as the weight of the average value of the nkth relative deviations of multiple physiological index data. The larger the relative difference, the greater the weight. is the weighted average of the nkth relative deviations of the multiple physiological index data at N moments in the ith monitoring cycle. Furthermore, the larger the k, the closer the serial number of the reference monitoring cycle (i.e., the inth monitoring cycle to the i-n+kth monitoring cycle) is to the ith monitoring cycle during prediction. Therefore, the accuracy of the prediction result of the physiological index data in the ith monitoring cycle is higher. At the same time, the larger the k, the more reference data (i.e., the historical physiological index data of k monitoring cycles between the inth monitoring cycle to the i-n+kth monitoring cycle can be referenced), and the more accurate the prediction result is. Therefore, the larger the weight, the greater the prediction result. As a weight, A weighted average is performed to obtain a relative deviation score of the physiological indicator data. The higher the relative deviation score, the worse the prediction accuracy of the physiological indicator data prediction model of the i-1th monitoring cycle for the physiological indicator data of the i-th monitoring cycle. If the relative deviation score is greater than or equal to the preset deviation threshold, the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained.
[0044] In this way, the relative deviation score of the physiological indicator data can be obtained based on the weighted average of the nkth relative deviations. When determining the weights, various factors that may affect the prediction accuracy, such as the amount of data of the reference historical physiological indicators and the time difference between the monitoring period corresponding to the reference historical physiological indicators and the i-th monitoring period, are considered to improve the accuracy and objectivity of the relative deviation score. Furthermore, it can be determined whether the physiological indicator data prediction model for the i-1th monitoring period needs to be trained, thereby continuously optimizing the physiological indicator data prediction model and improving the accuracy of the physiological indicator data prediction model for the i-1th monitoring period.
[0045] According to one embodiment of the present invention, when the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained, the training module may train the physiological indicator data prediction model of the i-1th monitoring cycle according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data to obtain the physiological indicator data prediction model of the i-th monitoring cycle.
[0046] According to one embodiment of the present invention, when the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained, the physiological indicator data prediction model of the i-1th monitoring cycle is trained according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data to obtain the physiological indicator data prediction model of the i-1th monitoring cycle, including: determining the loss function of the physiological indicator data prediction model of the i-1th monitoring cycle according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data; training the physiological indicator data prediction model of the i-1th monitoring cycle according to the loss function to obtain the physiological indicator data prediction model of the i-1th monitoring cycle.
[0047] According to one embodiment of the present invention, the loss function can be obtained based on the difference between the model predicted value and the actual value, so that the loss function can be used to optimize the model, and the loss function can be reduced by the gradient descent method, thereby reducing the prediction error of the model. After multiple trainings, the physiological indicator data prediction model of the i-th monitoring cycle can be obtained.
[0048] According to one embodiment of the present invention, determining the loss function of the physiological indicator data prediction model of the i-1th monitoring period according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data includes: determining the loss function L of the physiological indicator data prediction model of the i-1th monitoring period according to formula (2),
[0049] (2)
[0050] Among them, at t=i, is the sth physiological index data at the jth moment of the ith monitoring cycle, is the sth physiological index training data at the jth moment of the ith monitoring cycle based on the historical physiological index data from the inth monitoring cycle to the th monitoring cycle. When t≠i, is the s-th historical physiological index data at the j-th moment of the t-th monitoring cycle, is the sth historical physiological indicator training data at the jth moment of the tth monitoring cycle based on the historical physiological indicator data from the inth monitoring cycle to the th monitoring cycle, M is the number of types of physiological indicator data, N is the number of moments in each monitoring cycle, s≤M, j≤N, 1≤h≤t-i+n, i-n+1≤t≤i, and s, j, t, h, M and N are all positive integers.
[0051] According to one embodiment of the present invention, in formula (2), is the relative deviation between the sth measured physiological index data at the jth moment of the tth monitoring cycle and the sth historical physiological index training data at the jth moment of the tth monitoring cycle based on the historical physiological index data from the inth monitoring cycle to the th monitoring cycle, It is the average value of the relative deviation of multiple physiological index data. is the total number of moments in all monitoring cycles before the tth monitoring cycle, is the number of moments between the last moment of the th monitoring cycle and the jth moment of the tth monitoring cycle. The smaller the number of moments, the shorter the time interval between the inth monitoring cycle to the last moment of the th monitoring cycle and the jth moment of the tth detection cycle referenced in the prediction, the more reliable the reference data, and the more accurate the prediction result. The relative difference between the total number of moments in the monitoring cycle before the t-th monitoring cycle and the number of moments between the last moment of the th monitoring cycle and the j-th moment of the t-th monitoring cycle can be used as the weight of the average value of the relative deviations of multiple physiological index data. The larger the relative difference, the greater the weight. is the weighted average of the relative deviations of multiple physiological index data at N moments in the t-th monitoring cycle. h is the difference between the t-th monitoring cycle and the th-th monitoring cycle. The larger the difference, the longer the time interval between the reference monitoring cycle (the in-th monitoring cycle to the th-th monitoring cycle) and the t-th cycle, and the lower the prediction accuracy. At the same time, the larger h is, the fewer the number of cycles between the in-th monitoring cycle and the th-th monitoring cycle, and the smaller the amount of historical physiological index data that can be referenced. Therefore, the larger h is, the smaller the weight is, and the prediction accuracy is lower. As The weights of are weighted averaged, and the value range of h is 1≤h≤t-i+n. Therefore, after weighted average, we can get Furthermore, the larger t is, the closer the number of the monitoring period for prediction is to the i-th monitoring period, and the greater the reference value of the prediction result. Therefore, t can be used as a weight to The weighted average is performed to obtain the loss function of the physiological indicator data prediction model of the i-1th monitoring cycle. In the process of training the physiological indicator data prediction model of the i-1th monitoring cycle, the loss function is back-propagated and some parameters within the model are adjusted to reduce the value of the loss function of the physiological indicator data prediction model of the i-1th monitoring cycle, thereby improving the accuracy of the physiological indicator data prediction model and obtaining the physiological indicator data prediction model of the i-th monitoring cycle.
[0052] In this way, the relative differences between the physiological indicator data, historical physiological indicator data and the physiological indicator training data and historical physiological indicator training data output by the physiological indicator data prediction model of the i-1th monitoring cycle can be determined, and weights can be set based on the characteristics that the larger the time difference, the lower the prediction accuracy, and the closer to the required monitoring cycle, the greater the reference value and the more reference data, the more accurate the prediction result. The weights are set, thereby performing weighted summation on the relative errors of the physiological indicator training data and historical physiological indicator training data output by the physiological indicator data prediction model of the i-1th monitoring cycle to obtain a loss function, so as to improve the design accuracy and objectivity of the loss function, thereby improving the training efficiency during the training process.
[0053] According to one embodiment of the present invention, the second prediction module can use the physiological indicator data, historical physiological indicator data, medication information, historical medication information, and planned medication information of the future monitoring cycle as input data of the physiological indicator data prediction model of the i-th monitoring cycle. The physiological indicator data prediction model of the i-th monitoring cycle can predict the predicted physiological indicator data at multiple moments of the future monitoring cycle based on the physiological indicator data and medication information.
[0054] According to an embodiment of the present invention, the condition prediction module can determine a condition score, which indicates the severity of the condition of the guardian in the current monitoring cycle and future monitoring cycles, and can reflect the trend of condition changes.
[0055] According to one embodiment of the present invention, determining a disease condition score based on the predicted physiological indicator data includes: acquiring standard indicator data of multiple physiological indicator data; and determining a disease condition score based on the standard indicator data of the multiple physiological indicator data and the predicted physiological indicator data.
[0056] According to one embodiment of the present invention, the standard indicator data of each physiological indicator data can be a standard parameter range that has been confirmed in the medical field, which is used to evaluate the physiological state and health status of the guardian. For example, the normal range of high blood pressure is about 90-120 mmHg, and the normal range of low blood pressure is about 60-80 mmHg. The preset standard indicator can be selected within the above normal range. According to the standard indicator data of multiple physiological indicator data and the predicted physiological indicator data, the disease condition score is determined.
[0057] According to an embodiment of the present invention, based on the standard indicator data of the multiple physiological indicator data and the predicted physiological indicator data, determining the disease condition score of the future monitoring period includes: determining the disease condition score S according to formula (3),
[0058] (3)
[0059] in, is the sth predicted physiological index data at the jth moment of the future i+xth monitoring cycle, is the standard indicator data of the sth type of physiological indicator data, is the sth type of physiological indicator data at the jth moment of the ith monitoring cycle, M is the number of types of physiological indicator data, N is the number of moments in each monitoring cycle, s≤M, j≤N, x≤n, and s, j, x, M and N are all positive integers, and The preset weights.
[0060] According to one embodiment of the present invention, in formula (3), It is the average value of the relative deviations between various predicted physiological index data and standard index data at the jth moment of the future i+xth monitoring cycle. The larger the average value is, the worse the disease condition in the future monitoring cycle will be. is the number of moments between the last moment of the ith monitoring cycle and the jth moment of the i+xth monitoring cycle. The smaller the number of moments, the more reliable the predicted data and the more accurate the prediction results. The relative difference between the total number of moments of all monitoring periods referenced in the prediction and the number of moments from the last moment of the i-th monitoring period to the j-th moment of the i+x-th monitoring period can be used as The greater the relative difference, the greater the weight. is the weight of the cycle difference between the i+xth monitoring cycle and the i-th monitoring cycle. The larger x is, the less accurate the prediction result is and the smaller its weight is. Using the above-mentioned multiple weights, the average values of the above-mentioned relative deviations are weighted averaged to obtain the first item of the medical condition score, which can be used to represent the expected medical condition in future monitoring cycles. On the other hand, the average value of the relative deviations between the various predicted physiological indicator data and the standard indicator data at each moment in the i-th monitoring cycle can be used as the second item of the medical condition score, which can be used to represent the medical condition in the current monitoring cycle. The weighted sum of the above two items can be used to obtain the medical condition score. The larger the medical condition score is, the worse the trend of the guardian's condition change is, that is, the more serious the condition is.
[0061] In this way, the disease condition score can be determined by predicting physiological indicator data, the relative deviation between physiological indicator data and standard indicator data, and the weight set based on the time difference between the future monitoring cycle and the current monitoring period, so as to predict in advance the possible changes in the condition of the guardian in the future period, which helps to take timely treatment measures.
[0062] According to one embodiment of the present invention, the early warning module can be used to generate corresponding early warning information when the medical condition score meets the early warning conditions (for example, when the medical condition score is higher than a preset threshold), including the basic information of the guardian, the medical condition score, the reason for triggering the early warning and treatment measures.
[0063] According to the disease change trend warning system of the embodiment of the present invention, by collecting the physiological index data and medication information of the monitoring personnel, the health status of the monitoring personnel can be monitored in real time and abnormal conditions can be discovered in time. Through the physiological index data prediction model, real-time and accurate disease change trend prediction and warning are provided. Early warning information is generated according to the disease condition score, so that intervention measures or adjustment of the treatment plan are taken in time to reduce the possibility of disease deterioration and improve medical efficiency. When determining whether the physiological index data prediction model of the i-1th monitoring cycle needs to be trained, the relative deviation score of the physiological index data can be obtained based on the weighted average of the nkth relative deviation, and when determining the weight, the data amount of the referenced historical physiological index and the time difference between the monitoring cycle corresponding to the referenced historical physiological index and the i-th monitoring cycle are considered. A variety of factors that may affect the prediction accuracy, such as the time difference between the monitoring cycle corresponding to the referenced historical physiological index and the i-th monitoring cycle, are considered to improve the accuracy and objectivity of the relative deviation score, and then it can be determined whether the physiological index data prediction model of the i-1th monitoring cycle needs to be trained, thereby continuously optimizing the physiological index data prediction model and improving the accuracy of the physiological index data prediction model of the i-1th monitoring cycle. When training the physiological indicator data prediction model of the i-1 monitoring cycle, the relative difference between the physiological indicator data, the historical physiological indicator data and the physiological indicator training data and the historical physiological indicator training data output by the physiological indicator data prediction model of the i-1 monitoring cycle can be determined, and the weights can be set based on the characteristics that the larger the time difference, the lower the prediction accuracy, and the closer to the required monitoring cycle, the greater the reference value and the more reference data, the more accurate the prediction result, so as to weight the relative errors of the physiological indicator training data and the historical physiological indicator training data output by the physiological indicator data prediction model of the i-1 monitoring cycle to obtain the loss function, so as to improve the design accuracy and objectivity of the loss function, thereby improving the training efficiency during the training process. When determining the disease condition score, the disease condition score can be determined by predicting the relative deviation of the physiological indicator data, the physiological indicator data and the standard indicator data, and the weights set based on the time difference between the future monitoring cycle and the current monitoring, so as to predict the possible changes in the condition of the guardian in the future period in advance, which is helpful to take timely treatment measures.
[0064] It should be understood by those skilled in the art that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.
[0065] Finally, 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A disease condition change trend early warning system, characterized in that: include: The collection module is used to collect multiple physiological index data of the guardian at multiple moments in the i-th monitoring cycle, and obtain the medication information of the guardian in the i-th monitoring cycle, wherein the medication information includes the type of medication used by the guardian and the dosage information of each type of medication, and i is a positive integer; An acquisition module is used to acquire historical physiological index data of multiple monitoring cycles before the i-th monitoring cycle, as well as historical medication information; The first prediction module is used to predict the historical physiological indicator data of the historical monitoring period according to the historical physiological indicator data and historical medication information of the historical monitoring period, the medication information of the i-th monitoring period, and the physiological indicator data prediction model of the i-1-th monitoring period, to obtain the historical physiological indicator training data, and to predict the physiological indicator data in the i-th monitoring period to obtain the physiological indicator training data; A judgment module, used for judging whether the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained according to the physiological indicator training data and the physiological indicator data; A training module, used for training the physiological indicator data prediction model of the i-1th monitoring cycle according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data, to obtain the physiological indicator data prediction model of the i-th monitoring cycle when the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained; A second prediction module is used to obtain predicted physiological indicator data at multiple moments of a future monitoring cycle based on the physiological indicator data, the historical physiological indicator data, the medication information, the historical medication information, the planned medication information of the future monitoring cycle, and the physiological indicator data prediction model of the i-th monitoring cycle; A disease condition prediction module, used to determine a disease condition score according to the predicted physiological index data; An early warning module, used to generate early warning information according to the disease condition score; According to the historical physiological indicator data and historical medication information of the historical monitoring period, the medication information of the i-th monitoring period, and the physiological indicator data prediction model of the i-1-th monitoring period, the historical physiological indicator data of the historical monitoring period is predicted to obtain the historical physiological indicator training data, and the physiological indicator data in the i-th monitoring period is predicted to obtain the physiological indicator training data, including: Input the historical physiological index data and historical medication information from the in-th monitoring cycle to the i-n+k-th monitoring cycle, the historical medication information from the i-n+k+1-th monitoring cycle to the i-1-th monitoring cycle, and the medication information of the i-th monitoring cycle into the physiological index data prediction model of the i-1-th monitoring cycle, and obtain the historical physiological index training data from the i-n+k+1-th monitoring cycle to the i-1-th monitoring cycle based on the historical physiological index data from the in-th monitoring cycle to the i-n+k-th monitoring cycle, and the physiological index training data of the i-th monitoring cycle based on the historical physiological index data from the in-th monitoring cycle to the i-n+k-th monitoring cycle, wherein 0≤k≤n-1, k is an integer, n is a positive integer, and n<i; In the case where the physiological indicator data prediction model of the i-1th monitoring cycle needs to be trained, the physiological indicator data prediction model of the i-1th monitoring cycle is trained according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data and the historical physiological indicator data to obtain the physiological indicator data prediction model of the i-th monitoring cycle, including: Determine a loss function of a physiological indicator data prediction model for an i-1th monitoring period according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data, and the historical physiological indicator data; According to the loss function, the physiological indicator data prediction model of the i-1th monitoring cycle is trained to obtain the physiological indicator data prediction model of the i-th monitoring cycle; Determining a loss function of a physiological indicator data prediction model for an i-1th monitoring cycle according to the physiological indicator training data, the historical physiological indicator training data, the physiological indicator data, and the historical physiological indicator data, including: According to the formula Determine the loss function L of the physiological index data prediction model for the i-1th monitoring cycle, where, at t=i, is the sth physiological index data at the jth moment of the ith monitoring cycle, is the sth physiological index training data at the jth moment of the ith monitoring cycle based on the historical physiological index data from the inth monitoring cycle to the th monitoring cycle. When t≠i, is the s-th historical physiological index data at the j-th moment of the t-th monitoring cycle, is the sth historical physiological indicator training data at the jth moment of the tth monitoring cycle based on the historical physiological indicator data from the inth monitoring cycle to the th monitoring cycle, M is the number of types of physiological indicator data, N is the number of moments in each monitoring cycle, s≤M, j≤N, 1≤h≤t-i+n, i-n+1≤t≤i, and s, j, t, h, M and N are all positive integers.
2. The disease condition change trend early warning system according to claim 1, characterized in that: Judging whether a physiological indicator data prediction model of an i-1th monitoring cycle needs to be trained according to the physiological indicator training data and the physiological indicator data, including: Obtaining the nkth relative deviation between the physiological indicator training data of the ith monitoring period and the physiological indicator data based on the historical physiological indicator data from the inth monitoring period to the i-n+kth monitoring period; According to the nkth relative deviation, obtaining a relative deviation score of the physiological indicator data; If the relative deviation score is greater than or equal to the preset deviation threshold, it is determined that the physiological indicator data prediction model for the i-1th monitoring cycle needs to be trained.
3. The disease condition change trend early warning system according to claim 2, characterized in that: Obtaining a relative deviation score of the physiological indicator data according to the nkth relative deviation includes: According to the formula Obtain the relative deviation score E of the physiological indicator data, wherein, is the sth physiological index data at the jth moment of the ith monitoring cycle, is the sth type of physiological indicator training data at the jth moment of the ith monitoring cycle based on the historical physiological indicator data from the inth monitoring cycle to the i-n+kth monitoring cycle, M is the number of types of physiological indicator data, N is the number of moments in each monitoring cycle, is the n-th relative deviation of the physiological indicator data, s≤M, j≤N, and s, j, M and N are all positive integers.
4. The disease condition change trend early warning system according to claim 1, characterized in that: Determining a disease condition score according to the predicted physiological index data includes: Obtain standard indicator data for multiple physiological indicator data; The disease condition score is determined based on the standard indicator data of the multiple physiological indicator data and the predicted physiological indicator data.
5. The disease condition change trend early warning system according to claim 4, characterized in that: Determining a disease condition score according to the standard indicator data of the plurality of physiological indicator data and the predicted physiological indicator data includes: According to the formula Determine the disease status score S for the future monitoring period, where is the sth predicted physiological index data at the jth moment of the future i+xth monitoring cycle, is the standard indicator data of the sth type of physiological indicator data, is the sth type of physiological indicator data at the jth moment of the ith monitoring cycle, M is the number of types of physiological indicator data, N is the number of moments in each monitoring cycle, s≤M, j≤N, x≤n, and s, j, x, M and N are all positive integers, and The preset weights.
