Health monitoring method and system based on deep learning
Through a deep learning-based health monitoring method, the difference in health monitoring indicators and similarity between the beds is used to solve the problem that existing systems are difficult to utilize associated patients' recovery status in patient health monitoring, and more accurate health recovery judgment and data processing are achieved.
Patent Information
- Application Number
- CN202510052340.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-14
- Publication Date
- 2025-05-02
AI Technical Summary
The existing bed management system is difficult to make full use of the recovery status of associated patients for accurate monitoring and health prediction during the patient's health monitoring process, resulting in the inability to efficiently ensure the recovery status of the postoperative disease.
The health monitoring method based on deep learning is adopted, and the health monitoring indicators on each bed are collected, the health monitoring indicator alarm level value of each patient is calculated, the risk bed is filtered out, and the disease similarity and health recovery status are determined by focusing on the matching of the risk list of the health monitoring indicators of the patients and the extended patients, and finally the health recovery conditions are determined through the deep learning model.
It improves the accuracy of patient health recovery judgment and the accuracy of health monitoring data processing, and can more effectively monitor and predict the patient's health status.
Smart Images
Figure CN119920470A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a health monitoring method and system based on deep learning. Background Art
[0002] In the modern medical environment, with the continuous improvement of the application of intelligent medical technologies and systems, bed management systems are becoming an important tool for hospital management and nursing optimization. These systems use data analysis and intelligent algorithms to help medical teams allocate and manage resource allocation between beds and nurses. However, in the process of monitoring patient health in beds, there are still challenges in adaptability and personalized monitoring.
[0003] Generally speaking, for patients placed in the same ward or adjacent wards, there is a certain correlation in their condition or care, which facilitates follow-up care. However, in the actual supervision process, the health status of each patient is monitored too independently, and it is impossible to fully utilize the recovery status of related patients to accurately monitor and predict the health of other patients, and efficiently ensure the patient's postoperative recovery status. Summary of the invention
[0004] To achieve the above objectives, this application provides the following technical solutions: According to a first aspect of the present invention, the present invention claims a health monitoring method based on deep learning, the method comprising: Collect the health monitoring indicators of each patient in each bed in each ward; Based on the difference between the number of health monitoring indicators obtained for each bed and each patient in the adjacent beds, the health monitoring indicator alarm level value of each bed of each patient is obtained; a random patient is regarded as a concerned patient, and based on the difference between the number of health monitoring indicators obtained for each bed of the concerned patient and the adjacent beds, the risk beds are filtered out from all the beds in each ward; a random bed in each ward is regarded as a candidate bed, and the risk bed closest to the candidate bed for each patient is regarded as a concerned risk bed, and the health monitoring indicator numbers of all beds between the concerned risk bed and the candidate bed are formed into a linked list, which is used as a candidate health monitoring indicator risk linked list for each patient's candidate bed; patients near the concerned patient are regarded as extended patients, and the similarity of the condition between each extended patient and the concerned patient in the candidate bed is obtained through the matching status between the concerned patient and the candidate health monitoring indicator risk linked list of the candidate bed of each extended patient; Based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, the health recovery status of the concerned patient in the candidate bed is obtained; Based on the health recovery status of the patient of interest in the candidate bed, the health recovery condition of the patient of interest in the candidate bed is determined through deep learning.
[0005] Furthermore, the health monitoring indicator alarm level value of each patient in each bed is obtained based on the difference between the health monitoring indicator number obtained for each patient in each bed and the adjacent beds before and after, including: The health monitoring index numbers obtained for each patient in all beds in each ward are sorted according to spatial distribution to form a set of linked lists, which are used as the health monitoring index linked lists for each patient; The average of the differences between each indicator in the health monitoring indicator chain list of each patient and all indicators in the previous and next neighborhoods is taken as the health monitoring indicator alarm level value of each patient and each bed.
[0006] Furthermore, the method of filtering out risky beds from all beds in each ward based on the difference between the health monitoring index numbers obtained for each bed of the patient and the adjacent beds includes: Collect the linked list of health monitoring indicators of the patients of concern; encode all beds according to spatial distribution, establish a candidate coordinate system with all beds as the horizontal axis and the number of health monitoring indicators of the patients of concern for each bed as the vertical axis, and map the indicators in the linked list of health monitoring indicators of the patients of concern to the candidate coordinate system; Based on the difference between the tangent value of each indicator feature point and the adjacent indicator feature points, the risk value of each indicator feature point is obtained; The risk value is not less than the preset second threshold All indicator characteristic points are regarded as risk points; the beds corresponding to the risk points are regarded as risk beds.
[0007] Furthermore, the risk value of each indicator feature point is obtained based on the difference in tangent value between each indicator feature point and the adjacent indicator feature point, and the specific calculation formula included is as follows:
[0008] In the formula, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To preset the first variable, is the absolute value symbol, For the The risk value of each indicator feature point.
[0009] Furthermore, the similarity of the condition of each extended patient and the concerned patient in the candidate bed is obtained by matching the candidate health monitoring indicator risk list of the concerned patient and each extended patient in the candidate bed, including: Based on the distance between each extended patient and the concerned patient, and the matching coefficient between each extended patient and the candidate health monitoring indicator risk list of the candidate bed of the concerned patient, the similarity of the condition between each extended patient and the concerned patient is obtained; Among them, the distance between each extended patient and the focused patient is in a negative matching relationship with the similarity of the condition between each extended patient and the focused patient.
[0010] Further, the health recovery status of the concerned patient in the candidate bed is obtained based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, including: Based on the health monitoring indicator alarm level value of each candidate bed of the extended patient, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, the importance of the concerned patient in the candidate bed to the health monitoring indicator of the extended patient is obtained; The health recovery status of the patient in the candidate bed is obtained by distributing the importance of the same hospital bed of the recovered patient to the extended patient health monitoring index.
[0011] Further, the importance of the concerned patient in the candidate bed to the health monitoring indicator of the extended patient is obtained based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, including: The product of the health monitoring indicator alarm level value of each candidate bed of the extended patient and the similarity of the condition between each extended patient and the concerned patient in the candidate bed is taken as the variable variable of each extended patient; the mean of the variable variables of all extended patients is taken as the overall variable variable of the extended patients; Based on the overall change variables of the extended patients and the number of health monitoring indicators obtained by the concerned patients in the candidate beds, the importance of the concerned patients in the candidate beds to the health monitoring indicators of the extended patients is obtained; Among them, the number of health monitoring indicators obtained by the concerned patient in the candidate bed and the importance of the concerned patient in the candidate bed to the extended patient health monitoring indicators are positively matched.
[0012] Furthermore, the health recovery status of the patient of interest in the candidate bed is obtained by distributing the importance of the same bed of the patient of interest who has recovered to the extended patient health monitoring indicator, including: Based on the information entropy and mean of the importance of the same candidate beds of the concerned patients in the recovered health to the extended patient health monitoring indicators, the health recovery status of the concerned patients in the candidate beds is obtained; Among them, the information entropy of the importance of the candidate bed of the patient who has recovered to the extended patient health monitoring indicators is negatively matched with the health recovery status of the patient in the candidate bed.
[0013] Furthermore, the determining of the health recovery condition of the patient of interest in the candidate bed through deep learning based on the health recovery status of the patient of interest in the candidate bed includes: The health recovery status of the patient of concern in all beds that have recovered is input into the LSTM deep learning model, and the LSTM deep learning model is trained to obtain the trained LSTM deep learning model; wherein, the applicable loss function in the LSTM deep learning model is the cross entropy loss function; the health recovery status of the patient of concern in the candidate bed is input into the trained LSTM deep learning model, and when the output is 1, it is determined that the patient of concern has the health recovery conditions in the candidate bed, and maintenance personnel need to be assigned to guide the patient's health recovery; when the output is 0, it is determined that the patient of concern does not have the health recovery conditions in the candidate bed.
[0014] According to a second aspect of the present invention, the present invention claims a deep learning-based health monitoring system, comprising one or more processors; A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the deep learning-based health monitoring method.
[0015] The present invention relates to the field of data processing technology, and in particular to a health monitoring method and system based on deep learning, including: based on the difference between the health monitoring indicator numbers of adjacent beds, obtaining the health monitoring indicator alarm level value of each bed of each patient and determining the risky bed; obtaining the similarity of the condition between each extended patient and the concerned patient in the candidate bed through the matching state between the concerned patient and the candidate health monitoring indicator risk chain list of each extended patient's candidate bed; obtaining the health recovery state of the concerned patient in the candidate bed based on the change variable of each extended patient's candidate bed and the health monitoring indicator number obtained by the concerned patient in the candidate bed; and determining the health recovery condition of the concerned patient in the candidate bed. The present invention improves the accuracy of patient health recovery judgment and also improves the accuracy of patient health monitoring data processing. . BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 This is a flowchart of the steps of the deep learning-based health monitoring method claimed in the embodiments of the present application. DETAILED DESCRIPTION
[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0018] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined as "first", "second" and "third" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications in the embodiments of this application (such as up, down, front and back, right, front, back...) are only used to explain the relative position relationship, movement conditions, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.
[0019] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0020] The specific scheme of the health monitoring method and system based on deep learning provided by the present invention is described in detail below with reference to the accompanying drawings.
[0021] See also Figure 1 , which shows a flowchart of the steps of a health monitoring method based on deep learning provided by an embodiment of the present invention, the method comprising: Step S001: Collect the health monitoring indicators of each patient in each bed in each ward.
[0022] It should be noted that in order to analyze whether the patient has recovered his health, it is necessary to collect the health monitoring indicator risks obtained by the patient over a period of time, and analyze the patient's health recovery conditions through the health monitoring indicator risks obtained by the patient over a period of time.
[0023] Specifically, a patient monitor is installed above each patient, and each ward is equally divided into 25 beds at intervals of 10 minutes. The health monitoring indicators of each bed are collected through the patient monitor.
[0024] At this point, the number of health monitoring indicators obtained for each patient in each bed in each ward is obtained.
[0025] Step S002: Based on the difference between the number of health monitoring indicators obtained for each bed and each patient on the adjacent beds, obtain the health monitoring indicator alarm level value for each bed of each patient; based on the difference between the number of health monitoring indicators obtained for each bed of the concerned patient and the adjacent beds, filter out risky beds from all beds in each ward; regard the risky bed closest to the candidate bed of each patient as the concerned risk bed, and form a linked list with the health monitoring indicator numbers of all beds between the concerned risk bed and the candidate beds, which is used as the candidate health monitoring indicator risk linked list for the candidate beds of each patient; obtain the similarity of the condition between each extended patient and the concerned patient in the candidate bed through the matching status between the concerned patient and the candidate health monitoring indicator risk linked list of the candidate beds of each extended patient.
[0026] It should be noted that when a patient shows conditions for health recovery, other patients will generally be chosen for monitoring. Therefore, when the health monitoring indicators of a patient reach the maximum load, the doctor will choose to monitor nearby patients, which will cause a sudden increase in the health monitoring indicators of other patients. Therefore, the health monitoring indicators of each patient and the change conditions of the health monitoring indicators of several nearby patients are used to comprehensively analyze the conditions for the patient's health recovery, so as to avoid the possibility of a patient being mistakenly believed to have recovered due to false indicators.
[0027] It should be further explained that when the patient recovers, it is generally in the morning, noon, afternoon and evening. When the patient recovers, the patient's health monitoring indicators will show a sudden increase during the recovery period. Therefore, the possibility of deterioration of the patient can be analyzed by the changes in health monitoring indicators over a period of time.
[0028] Specifically, a first variable is preset , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. It can be determined based on specific implementation conditions. The number of health monitoring indicators obtained for each patient in all beds in each ward is sorted according to spatial distribution to form a set of linked lists, which are used as the health monitoring indicator linked lists for each patient.
[0029] As an embodiment, the specific calculation method of the health monitoring indicator alarm level value of each patient and each bed is: the average of the difference between each indicator in the health monitoring indicator chain list of each patient and all indicators in the front and back neighborhood is used as the health monitoring indicator alarm level value of each patient and each bed. Among them, the number of indicators in the front and back neighborhood of each indicator is indivual.
[0030] In one embodiment of the present invention, the specific formula is:
[0031] In the formula, The first in the health monitoring indicator chain list for each patient indicators, The first in the health monitoring indicator chain list for each patient indicators, To preset the first variable, For each patient The health monitoring indicator alarm level value of each bed, is an absolute value symbol. When the first five indicators in the health monitoring indicator chain list of each patient are used to calculate the health monitoring indicator alarm level value, only the difference between the indicators with the distance is calculated.
[0032] Among them, the smaller the difference between the number of health monitoring indicators obtained for each bed of each patient and the number of health monitoring indicators obtained for the first few adjacent beds, the lower the possibility of deterioration in the patient in the bed; the larger the difference between the number of health monitoring indicators obtained for each bed of each patient and the number of health monitoring indicators obtained for the first few adjacent beds, the higher the possibility of deterioration in the patient in the bed.
[0033] At this point, the health monitoring indicator alarm level value of each patient and each bed is obtained.
[0034] It should be noted that when a patient's health monitoring index suddenly increases, causing the patient to recover to a certain state of health, a diagnostic effect will appear, that is, the health monitoring index of patients nearby the patient will increase. Therefore, the changes in the health monitoring index of the patient and the nearby patients have a certain positive matching relationship.
[0035] It should be further explained that when a patient recovers, there is generally a sudden increase from a relatively stable health monitoring indicator, and then the health monitoring indicator continues to increase until the health monitoring indicator on the patient reaches the patient's maximum load. At this time, the health monitoring indicator of the patient is relatively stable; at this time, the conditions for diagnosis will appear, that is, the health monitoring indicators of patients near the patient will also show a process from stability to sudden increase, and finally to stability; and the change process of the health monitoring indicator will become weaker as the distance between nearby patients increases, and the matching relationship of the corresponding health monitoring indicator change process becomes weaker.
[0036] Specifically, a random patient is taken as the patient of interest. A second variable is preset , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. It can be determined based on specific implementation conditions. Treat each patient as an extension of the patient of interest.
[0037] Collect a list of health monitoring indicators for patients of concern; encode all beds according to their spatial distribution, establish a candidate coordinate system with all beds as the horizontal axis and the number of health monitoring indicators of patients of concern for each bed as the vertical axis, map the indicators in the list of health monitoring indicators for patients of concern in the candidate coordinate system, calculate the tangent value between two adjacent indicator feature points, and obtain the risk value of each indicator feature point based on the difference in tangent value between each indicator feature point and its adjacent indicator feature point.
[0038] In one embodiment of the present invention, the specific formula is:
[0039] In the formula, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To preset the first variable, For the The risk value of each indicator feature point.
[0040] Among them, when the difference between the tangent values of adjacent indicator feature points of each indicator feature point is greater, it means that the risk value of the indicator feature point is greater; when the difference between the tangent values of adjacent indicator feature points of each indicator feature point is smaller, it means that the risk value of the indicator feature point is smaller. It is the tangent value between adjacent indicator feature points after each indicator feature point. The larger the tangent value is, the greater the possibility that the bed corresponding to this indicator feature point will have an increase in health monitoring indicators. It is the tangent value between each indicator feature point and the adjacent indicator feature point. The smaller the tangent value is, the greater the possibility that the bed corresponding to this indicator feature point will have an increase in the health monitoring indicator.
[0041] Preset a first threshold , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. The risk value is not less than the preset second threshold value. All indicator characteristic points are regarded as risk points; the beds corresponding to the risk points are regarded as risk beds.
[0042] A random bed in each ward is taken as a candidate bed; the risk beds closest to the candidate beds are collected as risk beds of concern; the number of health monitoring indicators of all beds between the risk bed of concern and the candidate bed that pass the patients of concern are combined into a set of linked lists according to the spatial distribution, which is used as the risk linked list of candidate health monitoring indicators for the patient of concern on the candidate bed; then, the number of health monitoring indicators of all beds between the risk bed of concern and the candidate bed that pass each extended patient is combined into a set of linked lists according to the spatial distribution, which is used as the risk linked list of candidate health monitoring indicators for each extended patient on the candidate bed.
[0043] Therefore, based on the matching status between the candidate health monitoring indicator risk list of the candidate bed of the concerned patient and the candidate health monitoring indicator risk list of each extended patient on the candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed can be obtained.
[0044] As an embodiment, a specific calculation method for the similarity of the condition between each extended patient and the patient of concern is as follows: based on the distance between each extended patient and the patient of concern, and the matching coefficient between the risk list of candidate health monitoring indicators of each extended patient and the candidate bed of the patient of concern, the similarity of the condition between each extended patient and the patient of concern is obtained; wherein, the distance between each extended patient and the patient of concern and the similarity of the condition between each extended patient and the patient of concern are in a negative matching relationship.
[0045] In one embodiment of the present invention, the specific formula is:
[0046] In the formula, For the The matching coefficient between the extended patients and the candidate health monitoring indicator risk list of the candidate bed of the concerned patient, For the The distance between the expanded patient and the concerned patient, for Function, used for linear normalization; For the The similarity of the condition between the extended patient and the focused patient. The distance between the two patients is the minimum distance between the two patients. The matching coefficient is obtained by the Pearson matching coefficient, and the collection process of the Pearson matching coefficient is a well-known technology and will not be described in detail here.
[0047] in, middle As The weight of , that is, when the distance between each extended patient and the concerned patient is closer, the extended patient is more affected by the concerned patient's diagnosis; when the distance between each extended patient and the concerned patient is farther, the extended patient is less affected by the concerned patient's diagnosis. When the matching coefficient between each extended patient and the candidate health monitoring indicator risk chain list of the concerned patient's candidate bed is larger, it means that the similarity of the condition between each extended patient and the concerned patient is greater; when the matching coefficient between each extended patient and the candidate health monitoring indicator risk chain list of the concerned patient's candidate bed is smaller, it means that the similarity of the condition between each extended patient and the concerned patient is smaller.
[0048] At this point, the similarity of the condition between each extended patient and the focus patient in the candidate bed is obtained.
[0049] Step S003: Based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, the health recovery status of the concerned patient in the candidate bed is obtained.
[0050] It should be noted that when the number of health monitoring indicators obtained on a certain bed of the concerned patient is more, the importance of the concerned patient to the health monitoring indicators of the extended patient is greater at this time, and vice versa, the importance of the concerned patient to the health monitoring indicators of the extended patient is smaller. When the similarity of the condition between each extended patient and the concerned patient is greater, the extended patient is more affected by the concerned patient; when the similarity of the condition between each extended patient and the concerned patient is smaller, the extended patient is less affected by the concerned patient. When the alarm level value of the patient health monitoring indicator of each bed of each extended patient is larger, it means that the possibility of the extended patient being affected by the diagnosis of the concerned patient is greater; vice versa, the possibility of the extended patient being affected by the diagnosis of the concerned patient is smaller. Therefore, based on the alarm level value of the health monitoring indicator of each candidate bed of the extended patient, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, the importance of the concerned patient in the candidate bed to the health monitoring indicator of the extended patient is obtained.
[0051] As an embodiment, a specific calculation method for the importance of the patient of concern in the candidate bed to the health monitoring indicators of the extended patients is as follows: the product of the alarm level value of the health monitoring indicator of each extended patient candidate bed and the similarity of the condition between each extended patient and the patient of concern in the candidate bed is regarded as the variable variable of each extended patient; the mean of the variable variables of all extended patients is regarded as the overall variable variable of the extended patients; based on the overall variable variable of the extended patients and the number of health monitoring indicators obtained by the patient of concern in the candidate bed, the importance of the patient of concern in the candidate bed to the health monitoring indicators of the extended patients is obtained; wherein, the number of health monitoring indicators obtained by the patient of concern in the candidate bed and the importance of the patient of concern in the candidate bed to the health monitoring indicators of the extended patients are in a positive matching relationship.
[0052] In one embodiment of the present invention, the specific formula is:
[0053] In the formula, For the The health monitoring indicator alarm level value of the expanded patient candidate bed, For the The similarity of the conditions of the extended patients and the focus patients in the candidate beds, To focus on the number of health monitoring indicators obtained by patients in candidate beds, is the second preset variable, which is the total number of all extended patients of the patient of interest, To focus on the importance of patients in candidate beds for expanding patient health monitoring indicators.
[0054] Among them, when the number of health monitoring indicators obtained on a certain bed of the concerned patient is more, the importance of the concerned patient to the health monitoring indicators of the extended patient is greater at this time, and vice versa, the importance of the concerned patient to the health monitoring indicators of the extended patient is smaller. The greater the similarity between the condition of each extended patient and the concerned patient, the greater the influence of the concerned patient on the extended patient; vice versa, the less influence the concerned patient has on the extended patient. The greater the alarm level value of the patient health monitoring indicator of each bed of each extended patient, the greater the possibility that the extended patient is affected by the diagnosis of the concerned patient; vice versa, the less likely the extended patient is affected by the diagnosis of the concerned patient.
[0055] So far, we have obtained the importance of focusing on the patient's candidate bed to expanding the patient's health monitoring indicators.
[0056] It should be noted that since the patient's condition generally deteriorates in the morning, noon, afternoon and evening, there will be conditions for health recovery; and when the patient has unexpected conditions such as patient collision, the patient will also recover; and the health recovery caused by the morning, noon, afternoon and evening will have the same conditions in each ward, and the health recovery caused by the monitoring of the patient's accident is only an occasional health recovery, so it is also necessary to comprehensively analyze the patient's health recovery conditions based on the importance of the patient in the same bed to the extended patient health monitoring indicators in the historical indicators. Therefore, based on the distribution of the importance of the patient who has recovered health in the same candidate bed to the extended patient health monitoring indicators, the health recovery status of the patient in the candidate bed is obtained.
[0057] As an embodiment, a specific calculation method for the health recovery status of the patient of concern in the candidate bed is as follows: based on the information entropy and the mean of the importance of the same candidate bed as the patient of concern in the recovered health to the extended patient health monitoring indicators, the health recovery status of the patient of concern in the candidate bed is obtained; wherein, the information entropy of the importance of the same candidate bed as the patient of concern in the recovered health to the extended patient health monitoring indicators is in a negative matching relationship with the health recovery status of the patient of concern in the candidate bed.
[0058] In one embodiment of the present invention, the specific formula is:
[0059] In the formula, To focus on the information entropy of the importance of the candidate beds of the patients who have recovered to the extended patient health monitoring indicators, To focus on the mean importance of candidate beds for patients who have recovered to the extended patient health monitoring indicators, is an exponential function with a natural constant as base, To focus on the patient's health recovery status in the candidate bed.
[0060] Among them, when the information entropy of the importance of the candidate bed of the patient who has recovered to the extended patient health monitoring index is smaller, it means that the importance of the extended patient health monitoring index on the same bed on all days is relatively close, and the distribution of the importance is relatively concentrated, that is, the greater the possibility caused by the morning, noon, afternoon and evening, the greater the health recovery state of the patient in the candidate bed; otherwise, the smaller the health recovery state of the patient in the candidate bed. When the mean of the importance of the candidate bed of the patient who has recovered to the extended patient health monitoring index is larger, the greater the health recovery state of the patient in the candidate bed; otherwise, the smaller the health recovery state of the patient in the candidate bed is.
[0061] At this point, attention is paid to the patient's health recovery status in the candidate bed.
[0062] Step S004: determining the health recovery condition of the patient of interest in the candidate bed through deep learning based on the health recovery status of the patient of interest in the candidate bed.
[0063] The health recovery status of the patient of concern in all beds that have recovered is input into the LSTM deep learning model, and the LSTM deep learning model is trained to obtain the trained LSTM deep learning model; wherein, the applicable loss function in the LSTM deep learning model is the cross entropy loss function. The health recovery status of the patient of concern in the candidate bed is input into the trained LSTM deep learning model. When the output is 1, it is determined that the patient of concern has the health recovery conditions in the candidate bed, and maintenance personnel need to be assigned to guide the patient's health recovery; when the output is 0, it is determined that the patient of concern does not have the health recovery conditions in the candidate bed.
[0064] Among them, the LSTM deep learning model is a well-known technology and will not be described in detail here.
[0065] It should be noted that the The model is only used for negative matching relationships and the output of the constraint model is in In the specific implementation, it can be replaced by other models with the same purpose. This embodiment is only based on The model is described as an example without any specific limitation. is the input to the model.
[0066] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
[0067] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be physically provided separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the description and drawings of this application, or directly or indirectly used in other matching technical fields, is also included in the patent protection scope of the present application.
[0068] The specific implementation methods of the invention are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.
Claims
1. A health monitoring method based on deep learning, characterized in that: The method includes: Collect the health monitoring indicators of each patient in each bed in each ward; Based on the difference between the number of health monitoring indicators obtained for each bed and each patient in the adjacent beds, the health monitoring indicator alarm level value of each bed of each patient is obtained; a random patient is taken as a concerned patient, and based on the difference between the number of health monitoring indicators obtained for each bed of the concerned patient and the adjacent beds, the risk beds are filtered out from all the beds in each ward; a random bed in each ward is taken as a candidate bed, and the risk bed closest to the candidate bed for each patient is taken as a concerned risk bed, and the health monitoring indicator numbers of all beds between the concerned risk bed and the candidate bed are formed into a linked list, which is used as a candidate health monitoring indicator risk linked list for each patient's candidate bed; patients near the concerned patient are taken as extended patients, and the similarity of the condition between each extended patient and the concerned patient in the candidate bed is obtained through the matching status between the concerned patient and the candidate health monitoring indicator risk linked list of the candidate bed of each extended patient; Based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, the health recovery status of the concerned patient in the candidate bed is obtained; Based on the health recovery status of the patient of interest in the candidate bed, the health recovery condition of the patient of interest in the candidate bed is determined through deep learning.
2. The health monitoring method based on deep learning as claimed in claim 1, characterized in that: The health monitoring indicator alarm level value of each patient in each bed is obtained based on the difference between the health monitoring indicator number obtained for each bed and the number of health monitoring indicators obtained for each patient in the adjacent beds, including: The health monitoring index numbers obtained for each patient in all beds in each ward are sorted according to spatial distribution to form a set of linked lists, which are used as the health monitoring index linked lists for each patient; The average of the differences between each indicator in the health monitoring indicator chain list of each patient and all indicators in the previous and next neighborhoods is taken as the health monitoring indicator alarm level value of each patient and each bed.
3. The health monitoring method based on deep learning as claimed in claim 2, characterized in that: Based on the difference between the health monitoring index numbers obtained for each bed of the patient and the adjacent beds, risky beds are filtered out from all beds in each ward, including: Collect the linked list of health monitoring indicators of the patients of concern; encode all beds according to spatial distribution, establish a candidate coordinate system with all beds as the horizontal axis and the number of health monitoring indicators of the patients of concern for each bed as the vertical axis, and map the indicators in the linked list of health monitoring indicators of the patients of concern to the candidate coordinate system; Based on the difference between the tangent value of each indicator feature point and the adjacent indicator feature points, the risk value of each indicator feature point is obtained; The risk value is not less than the preset second threshold All indicator characteristic points are regarded as risk points; the beds corresponding to the risk points are regarded as risk beds.
4. The health monitoring method based on deep learning as claimed in claim 3, characterized in that: The risk value of each indicator feature point is obtained based on the difference in tangent value between each indicator feature point and the adjacent indicator feature point, and the specific calculation formula included is as follows: , In the formula, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To focus on the patient's health monitoring indicators list The indicator and The tangent value between the indicators, To preset the first variable, is the absolute value symbol, For the The risk value of each indicator feature point.
5. The health monitoring method based on deep learning as claimed in claim 1, characterized in that: The similarity of the condition of each extended patient and the concerned patient in the candidate bed is obtained by matching the candidate health monitoring indicator risk chain list of the concerned patient and each extended patient in the candidate bed, including: Based on the distance between each extended patient and the concerned patient, and the matching coefficient between each extended patient and the candidate health monitoring indicator risk list of the candidate bed of the concerned patient, the similarity of the condition between each extended patient and the concerned patient is obtained; Among them, the distance between each extended patient and the focused patient is in a negative matching relationship with the similarity of the condition between each extended patient and the focused patient.
6. The health monitoring method based on deep learning as claimed in claim 1, characterized in that: The health recovery state of the concerned patient in the candidate bed is obtained based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, including: Based on the health monitoring indicator alarm level value of each candidate bed of the extended patient, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, the importance of the concerned patient in the candidate bed to the health monitoring indicator of the extended patient is obtained; The health recovery status of the patient in the candidate bed is obtained by distributing the importance of the same hospital bed of the recovered patient to the extended patient health monitoring index.
7. The health monitoring method based on deep learning as claimed in claim 6, characterized in that: The importance of the concerned patient in the candidate bed to the health monitoring indicator of the extended patient is obtained based on the health monitoring indicator alarm level value of each extended patient candidate bed, the similarity of the condition between each extended patient and the concerned patient in the candidate bed, and the number of health monitoring indicators obtained by the concerned patient in the candidate bed, including: The product of the health monitoring indicator alarm level value of each candidate bed of the extended patient and the similarity of the condition between each extended patient and the concerned patient in the candidate bed is taken as the variable variable of each extended patient; the mean of the variable variables of all extended patients is taken as the overall variable variable of the extended patients; Based on the overall change variables of the extended patients and the number of health monitoring indicators obtained by the concerned patients in the candidate beds, the importance of the concerned patients in the candidate beds to the health monitoring indicators of the extended patients is obtained; Among them, the number of health monitoring indicators obtained by the concerned patient in the candidate bed and the importance of the concerned patient in the candidate bed to the extended patient health monitoring indicators are positively matched.
8. The health monitoring method based on deep learning as claimed in claim 6, characterized in that: The health recovery status of the patient of interest in the candidate bed is obtained by distributing the importance of the same bed of the patient of interest who has recovered to the extended patient health monitoring indicator, including: Based on the information entropy and mean of the importance of the same candidate beds of the concerned patients in the recovered health to the extended patient health monitoring indicators, the health recovery status of the concerned patients in the candidate beds is obtained; Among them, the information entropy of the importance of the candidate bed of the patient who has recovered to the extended patient health monitoring indicators is negatively matched with the health recovery status of the patient in the candidate bed.
9. The health monitoring method based on deep learning as claimed in claim 1, characterized in that: The method of determining the health recovery condition of the patient of interest in the candidate bed through deep learning based on the health recovery status of the patient of interest in the candidate bed includes: The health recovery status of the patient of concern in all beds that have recovered is input into the LSTM deep learning model, and the LSTM deep learning model is trained to obtain the trained LSTM deep learning model; wherein, the applicable loss function in the LSTM deep learning model is the cross entropy loss function; the health recovery status of the patient of concern in the candidate bed is input into the trained LSTM deep learning model, and when the output is 1, it is determined that the patient of concern has the health recovery conditions in the candidate bed, and maintenance personnel need to be assigned to guide the patient's health recovery; when the output is 0, it is determined that the patient of concern does not have the health recovery conditions in the candidate bed.
10. A health monitoring system based on deep learning, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the deep learning-based health monitoring method as described in any one of claims 1 to 9.