A rapid recommendation method for medical emergency plans based on artificial intelligence
By obtaining and clustering the information matching values and disposal score values of medical events, low-dimensional data points are formed, and the optimal hyperplane is obtained using the SVM algorithm, which solves the problem of low classification accuracy of SVM under high-dimensional data, and improves the Recommended accuracy of medical emergency plans.
Patent Information
- Application Number
- CN202411053981.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-02
AI Technical Summary
The existing support vector machine (SVM) algorithm has low classification accuracy on high-dimensional data, resulting in a low accuracy of recommendations for medical emergency solutions for emergencies.
By obtaining the information matching values of the current medical event and historical medical event, clustering to obtain different cluster clusters, and combining the disposal score values to form low-dimensional data points, the optimal hyperplane is obtained using the SVM algorithm to recommend medical emergency solutions.
It improves the accuracy of medical emergency plan recommendations, reduces the classification error rate, and makes the recommended plan more valuable.
Smart Images

Figure CN118606877B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical plan recommendation, and in particular to a method for quickly recommending medical emergency plans based on artificial intelligence. Background Art
[0002] Artificial intelligence technologies such as machine learning and knowledge graphs are increasingly being used in the medical field. These technologies can use big data analysis and pattern recognition to extract useful information from massive amounts of medical data and provide support for medical decision-making. In the face of public emergencies or natural disasters, quickly and accurately recommending medical emergency plans is crucial to protecting people's health and responding to emergencies.
[0003] The influencing factors of medical events include: the patient's health status, the geographical location of the accident, medical resources and other factors. Usually, the various factors of medical events constitute data points, and the support vector machine (SVM) algorithm is used to classify the data points, and then recommend medical event plans. Due to the large number of influencing factors of medical events, the dimensions of the data points corresponding to medical events are high, and the classification accuracy of the SVM algorithm on low-dimensional data is usually higher than that on high-dimensional data, which increases the error rate of classifying the data points corresponding to medical events, resulting in a low accuracy rate in recommending medical emergency plans for sudden medical events. Summary of the invention
[0004] In order to solve the technical problem that the classification accuracy of SVM algorithm in high-dimensional data is lower than that in low-dimensional data, resulting in low accuracy in recommending medical emergency plans for sudden medical events, the purpose of the present invention is to provide a method for quickly recommending medical emergency plans based on artificial intelligence. The technical solutions adopted are as follows:
[0005] The present invention proposes a method for quickly recommending medical emergency plans based on artificial intelligence, the method comprising:
[0006] Obtaining medical details of a current medical event and different historical medical events, as well as a medical emergency plan for each historical medical event, wherein the medical details include different types of medical data;
[0007] According to the similarity between the current medical event and the same type of medical data of each historical medical event, an information matching value between the current medical event and each historical medical event is obtained;
[0008] Clustering the information matching values to obtain different first clustering clusters; obtaining a treatment score value of the historical medical event corresponding to each information matching value in each first clustering cluster according to each information matching value in each first clustering cluster and the difference between each first clustering cluster and the information matching values in other first clustering clusters;
[0009] The information matching value and the treatment score value of the current medical event and each historical medical event constitute the data point of each historical medical event; based on the position distribution of the data points, the optimal hyperplane in the binary classification algorithm is obtained;
[0010] Based on the distance between the data points of all historical medical events and the optimal hyperplane, a medical emergency plan for the current medical event is recommended.
[0011] Furthermore, the method for obtaining the information matching value between the current medical event and each historical medical event according to the similarity between the current medical event and the same type of medical data of each historical medical event includes:
[0012] Any historical medical event is selected as the target medical event. If the same type of medical data of the current medical event and the medical details of the target medical event have an intersection, the corresponding medical data is recorded as the data to be tested. The number of types of the data to be tested in all types of medical data of the medical details of the target medical event is counted and recorded as the similarity between the current medical event and the target medical event.
[0013] The valid identifier of each type of the data to be tested of the medical details of the target medical event is set to a first preset value, and the valid identifier of each type of medical data other than the data to be tested is set to a second preset value; the average of the valid identifiers of all types of medical data of the target medical event is calculated as the data similarity of the target medical event; the first preset value is greater than the second preset value;
[0014] The information matching value of the current medical event and the target medical event is obtained according to the type similarity and the data similarity between the current medical event and the target medical event; the type similarity and the data similarity are both positively correlated with the information matching value.
[0015] Furthermore, the method of obtaining the treatment score value of the historical medical event corresponding to each information matching value in each first cluster according to each information matching value in each first cluster and the difference between each first cluster and the information matching value in other first clusters includes:
[0016] For each historical medical event, if the patient of the historical medical event survives, the treatment identifier of the historical medical event is set to a third preset value; if the patient of the historical medical event does not survive, the treatment identifier of the historical medical event is set to a fourth preset value; the cumulative sum of the treatment identifiers of the historical medical events corresponding to all information matching values in each first cluster is recorded as the treatment success value of each first cluster; the third preset value is greater than the fourth preset value;
[0017] The treatment score value of the historical medical event corresponding to each information matching value in each first cluster is obtained by combining the difference between the information matching value in each first cluster and the other first clusters, and each information matching value in each first cluster and the treatment success value.
[0018] Furthermore, the calculation formula of the treatment score value of the historical medical event corresponding to each information matching value in each first cluster is as follows:
[0019] ; In the formula, is the treatment score value of the historical medical event corresponding to the j-th information matching value in the r-th first cluster; is the jth information matching value in the rth first cluster; is the maximum value of all the information matching values within the first cluster; is the treatment success value of the rth first cluster; is the mean of all information matching values in the rth first cluster; is the mean of all information matching values in the tth first cluster except the rth first cluster; T is the total number of all first clusters; exp is an exponential function with the natural constant e as the base; Norm is the normalization function.
[0020] Furthermore, the method for obtaining the optimal hyperplane in the SVM algorithm based on the position distribution of the data points includes:
[0021] A two-dimensional coordinate system is established with the information matching value as the horizontal axis and the treatment score value as the vertical axis; all data points of historical medical events are marked in the two-dimensional coordinate system to obtain analysis coordinate points of corresponding data points; the analysis coordinate points in the two-dimensional coordinate system are clustered to obtain different second clustering clusters;
[0022] Obtain the centroid of each second cluster, calculate the Euclidean distance between the centroid of each second cluster and the coordinate origin of the two-dimensional coordinate system, and record it as the judgment distance of each second cluster; select the second cluster corresponding to the smallest judgment distance as the first target cluster;
[0023] Selecting a cluster to be determined from all the second clustering clusters, the abscissa of the centroid of the cluster to be determined is greater than the abscissa of the first target cluster, and the ordinate of the centroid is greater than the ordinate of the first target cluster;
[0024] For each undetermined cluster, the mean of the horizontal coordinates of all the analysis coordinate points in the undetermined cluster is calculated and recorded as the comprehensive matching value, and the mean of the vertical coordinates is recorded as the comprehensive score value; the mean of the comprehensive matching value and the comprehensive score value of the undetermined cluster is obtained and recorded as the judgment value of the undetermined cluster; among all the undetermined clusters, the undetermined cluster corresponding to the largest judgment value is taken as the second target cluster;
[0025] An optimal hyperplane in a binary classification algorithm is obtained according to the position distribution of the centroids of the first target cluster and the second target cluster.
[0026] Furthermore, the expression of the optimal hyperplane is as follows:
[0027] , , ; In the formula, y is the ordinate of the point on the optimal hyperplane; z is the abscissa of the point on the optimal hyperplane; k is the slope of the optimal hyperplane; b is the ordinate intercept of the optimal hyperplane; is the ordinate of the centroid of the second target cluster; is the ordinate of the centroid of the first target cluster; is the abscissa of the centroid of the second target cluster; is the abscissa of the centroid of the first target cluster.
[0028] Furthermore, the method for recommending a medical emergency plan for a current medical event based on the distance between the data points of all historical medical events and the optimal hyperplane includes:
[0029] Obtain the analysis coordinate point on the right side of the optimal hyperplane and record it as the target point;
[0030] The Euclidean distance between each target point in the two-dimensional coordinate system and the optimal hyperplane is obtained, and the medical emergency plan of the historical medical event corresponding to the target point with the smallest Euclidean distance is used as the medical emergency plan for the current medical event.
[0031] Furthermore, the clustering method for clustering the information matching values to obtain different first cluster clusters is a K-means clustering algorithm.
[0032] Furthermore, the clustering method for clustering the analysis coordinate points in the two-dimensional coordinate system to obtain different second cluster clusters is a DBSCAN clustering algorithm.
[0033] Furthermore, the binary classification algorithm is a SVM algorithm.
[0034] The present invention has the following beneficial effects:
[0035] In an embodiment of the present invention, the information dimension of the medical event is reduced by converting multiple factors affecting the medical event into the similarity between the medical data information of the historical medical event and the current medical event, that is, the information matching value, for subsequent analysis; the information matching value is obtained based on the matching degree of the current medical event and the medical details of the historical medical event. However, due to the particularity and individuality of emergencies, the accuracy of recommending medical emergency plans based only on the information matching degree between the historical medical event and the current medical event is low. The first cluster is obtained based on the clustering of the information matching value. The information matching values within the same cluster correspond to the particularities of the historical medical events, so that according to each first cluster, The difference with the other information matching values in the first cluster can accurately reflect the differences in the particularity of medical events, and is analyzed in combination with each information matching value in the first cluster, so that the obtained disposal score value of the historical medical event presents the effectiveness of recommending the medical emergency plan for the current medical event; the information matching value and the disposal score value constitute the data points of the historical medical event, and subsequent analysis is performed based on low-dimensional data points to reduce the error rate of classifying the corresponding data points of the medical event; the accuracy of recommending the medical emergency plan for the current medical event based on the distance between the data points of the historical medical event and the optimal hyperplane is improved, so that the medical emergency plan recommended for the current medical event has more reference value. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0037] Figure 1 A method flow chart of a method for quickly recommending medical emergency plans based on artificial intelligence provided by one embodiment of the present invention;
[0038] Figure 2 A local schematic diagram of the position distribution of coordinate points and the optimal hyperplane in a two-dimensional coordinate system provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0039] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of a medical emergency plan rapid recommendation method based on artificial intelligence proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0040] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0041] The following is a detailed description of a method for quickly recommending medical emergency plans based on artificial intelligence provided by the present invention in conjunction with the accompanying drawings.
[0042] See also Figure 1 , which shows a method flow chart of a method for quickly recommending medical emergency plans based on artificial intelligence provided by an embodiment of the present invention, the method comprising:
[0043] Step S1: Obtain medical details of the current medical event and different historical medical events, as well as a medical emergency plan for each historical medical event, wherein the medical details include different types of medical data.
[0044] A medical emergency event refers to an event caused by natural disasters, accidental disasters, public health events, social security events, etc., which poses a serious threat to individuals, groups or public health, and requires immediate medical rescue to reduce deaths and disabilities and control the spread of diseases. This type of event usually triggers a medical emergency response mechanism, which immediately provides medical care, psychological support and basic life support by assessing the severity of the event, the scope of impact and resource requirements, including: medical personnel, medicines, equipment and other necessary supplies. Sudden medical events are usually judged by medical personnel based on experience and on-site conditions, but due to the urgency of the emergency, medical personnel may delay treatment due to incomplete emergency equipment or emergency medicines when rushing to the rescue. Therefore, in order to respond to sudden medical events and treat the wounded as much as possible, this application recommends emergency plans for current sudden medical events by analyzing the specific circumstances of historical medical events and current sudden medical events, so as to better handle sudden medical events.
[0045] Medical events that occurred in the historical time period are recorded as historical medical events, and current sudden medical events are recorded as current medical events; the medical details of each historical medical event in the historical time period are obtained through the medical information log of the emergency center or the hospital's electronic medical record system, and the medical emergency plan for each historical medical event is obtained. Each historical medical event corresponds to a medical emergency plan, and the medical emergency plan includes: emergency treatment methods, treatment duration, and patient health status information. To obtain the medical details of the current medical room event, it is mainly noted that the specific circumstances of the current medical event have not yet been entered into the medical record system. The medical details of the current medical event need to be obtained by medical staff based on the specific accident situation. Medical details include multiple types of medical data, such as: disease type, danger level, distance between the accident site and the emergency center, etc.
[0046] Among them, disease types include: car accidents, sudden illnesses and fires, etc.; in this embodiment, there are 5 levels of danger levels, namely 5, 4, 3, 2, and 1. Patients with danger level 5 are the most seriously injured, and patients with danger level 1 are the least injured. The danger level of historical medical events is set by medical staff according to the patient's injury.
[0047] It should be noted that the embodiment of the present invention sets the medical data involving time, distance and level as numerical data, and sets the medical data involving disease type and medical drugs used as text data; specifically: the disease type in the medical data is text data, the danger level is numerical data, and the distance between the occurrence site and the emergency center is numerical data. The historical time period is the time period before the current medical event. In this embodiment, the length of the historical time period is 30 days, and the implementer can set it according to the specific situation.
[0048] Step S2: Obtain information matching values between the current medical event and each historical medical event based on the similarity between the current medical event and the same type of medical data of each historical medical event.
[0049] Specifically, when recommending medical emergency plans for current sudden medical events based on medical emergency plans corresponding to historical medical events, it is necessary to analyze the common characteristics and specific circumstances between different medical events, and select medical emergency plans of historical medical plans similar to the sudden current medical event to provide guidance for the sudden current medical event.
[0050] When matching information between historical medical events and current medical events, the matching status of historical medical events and current medical events in the same type of medical data is determined by analyzing the differences in the same type of medical data. The matching status of all types of medical data is combined to obtain the similarity between the medical information of the historical medical event and the current medical event, that is, the information matching value.
[0051] Preferably, the specific method for obtaining the information matching value is: randomly select any historical medical event and record it as the target medical event. If there is an intersection between the same type of medical data of the medical details of the current medical event and the target medical event, the corresponding medical data will be recorded as the data to be tested; the number of types of data to be tested in all types of medical data of the medical details of the target medical event is counted, and recorded as the type similarity between the current medical event and the target medical event; the valid identifier of each type of data to be tested in the medical details of the target medical event is set to a first preset value, and the valid identifier of each type of medical data except the data to be tested is set to a second preset value; the average of the valid identifiers of all types of medical data of the target medical event is calculated as the data similarity of the target medical event; the first preset value is greater than the second preset value; according to the type similarity and data similarity between the current medical event and the target medical event, the information matching value of the current medical event and the target medical event is obtained; the type similarity and data similarity are both positively correlated with the information matching value.
[0052] It should be noted that in the embodiment of the present invention, since the disease type is text data, the intersection of this type of medical data is equivalent to the existence of the same words, phrases or sentences in the two text data; the danger level and the distance between the occurrence location and the emergency center are average numerical data, so the intersection of each type of medical data is equivalent to the numerical value of the medical data being equal. The data to be tested is the medical data with relatively matching information among the multiple types of medical data of the current medical event and the historical medical event, and the information of other types of medical data other than the data to be tested is relatively different. Therefore, the data to be tested should be given more consideration than other types of medical data other than the data to be tested in analyzing the information matching value, and the first preset value needs to be set to be greater than the second preset value. Since other types of medical data other than the data to be tested in this embodiment have basically no reference value when calculating the information matching value, the first preset value takes the empirical value 1, and the second preset value takes the empirical value 0, which can be set by the implementer according to the specific situation.
[0053] In other embodiments of the present invention, different numerical values may be set for the valid identifiers of each type of data to be tested for obtaining current medical events and historical medical events according to the importance of each type of medical data to the medical event, so as to provide different degrees of contribution to the calculation of the information matching value based on the importance of the data to be tested.
[0054] In the embodiment of the present invention, a specific example of a calculation formula for the information matching value between the current medical event and the target medical event is as follows:
[0055]
[0056] In the formula, is the information matching value between the current medical event and the target medical event; d is the current medical event; m is the target medical event; N is the similarity between the types of the current medical event and the target medical event, that is, the number of types of medical data whose valid identifier is the first preset value in the medical details of the target medical event; I is the number of types of medical data contained in the medical details of the historical medical event; is a valid identifier of the i-th category of medical data in the medical details of the target medical event; is the data similarity of the target medical event; Norm is the normalization function.
[0057] In another embodiment of the present invention, the result of normalizing the sum of the category similarity and the data similarity can be used as the information matching value, so that both the category similarity and the data similarity are positively correlated with the information matching value. In other embodiments of the present invention, other basic mathematical operations can be selected to form a positive correlation, which is not limited here.
[0058] It should be noted that when the category similarity N is larger, the more types of medical data with matching information in the medical details of the current medical event and the target medical event are, which means that the specific circumstances of the current medical event and the target medical event are closer, the greater the information matching degree of the two medical events, and the higher the information matching value. The bigger. Reflects the average matching degree between the current medical event and the medical data corresponding to the target medical event. The value range is between 0 and 1. The closer it is to 1, the more it indicates a historical medical event. The closer the data information is to the current medical event, the higher the information matching value. The bigger.
[0059] According to the calculation method of the information matching value between the current medical event and the target medical event, the information matching value between the current medical event and each historical medical event is obtained.
[0060] Step S3: clustering the information matching values to obtain different first clustering clusters; obtaining the disposal score value of the historical medical event corresponding to each information matching value in each first clustering cluster based on each information matching value in each first clustering cluster and the difference between the information matching values in each first clustering cluster and the other first clustering clusters.
[0061] Specifically, the information matching value is obtained based on the matching degree of similar medical data in the medical details of the current medical event and the historical medical event. However, due to the particularity and individuality of emergencies, the accuracy of recommending medical emergency plans based only on the information matching degree between historical medical events and current medical events is low. The present invention analyzes different historical medical events, obtains the disposal score values of the historical medical events, and then recommends medical emergency plans based on the disposal score values.
[0062] First, in the embodiment of the present invention, a K-means clustering algorithm is used, where K represents the number of clusters, and K takes an empirical value of 10, which can be set by the implementer according to the specific situation; the information matching values of the current medical event and all historical medical events are clustered to obtain at least two first clustering clusters; the information matching degree of the historical medical events corresponding to the information matching values in the same first clustering cluster is relatively close to that of the current medical event. Among them, the K-means clustering algorithm is a well-known technology for those skilled in the art and will not be described in detail here.
[0063] Secondly, the information matching values within the first clusters are analyzed. The information matching values present the degree of information similarity between the corresponding historical medical events and the current medical events. The difference between the information matching values within each first cluster and the remaining first clusters reflects the information similarity between the historical medical events corresponding to each first cluster and the remaining historical medical events and the current medical events. Comprehensive analysis of the above situations enables the obtained disposal score value to present the accuracy of the recommended medical emergency plan for the current medical event.
[0064] Preferably, the method for obtaining the treatment score value is: for each historical medical event, if the patient of the historical medical event survives, the treatment identifier of the historical medical event is set to a third preset value; if the patient of the historical medical event does not survive, the treatment identifier of the historical medical event is set to a fourth preset value; the cumulative sum of the treatment identifiers of the historical medical events corresponding to all information matching values in each first cluster is recorded as the treatment success value of each first cluster; the third preset value is greater than the fourth preset value; combined with the difference between the information matching values in each first cluster and the remaining first clusters, as well as each information matching value and the treatment success value in each first cluster, the treatment score value of the historical medical event corresponding to each information matching value in each first cluster is obtained.
[0065] It should be noted that for each historical medical event, compared with the case where the patient did not survive, the medical emergency plan adopted in the medical event when the patient survived is more effective and more worthy of recommendation, and the third preset value is set to be greater than the fourth preset value. Since the medical emergency plan for the patient who did not survive in this embodiment has basically no reference value, the third preset value takes the empirical value 1 and the fourth preset value takes the empirical value 0, and the implementer can set it according to the specific situation.
[0066] The calculation formula for the treatment score of the historical medical event corresponding to each information matching value in each first cluster is as follows:
[0067]
[0068] In the formula, is the treatment score value of the historical medical event corresponding to the j-th information matching value in the r-th first cluster; is the jth information matching value in the rth first cluster; is the maximum value of all the information matching values within the first cluster; is the treatment success value of the rth first cluster; is the mean of all information matching values in the rth first cluster; is the mean of all information matching values in the tth first cluster except the rth first cluster; T is the total number of all first clusters; exp is an exponential function with the natural constant e as the base; Norm is the normalization function.
[0069] It should be noted that the embodiment of the present invention utilizes right Normalization is performed, when When the value is larger, the historical medical event corresponding to the jth information matching value in the rth cluster is more similar to the information of the current medical event. When recommending an emergency plan for the current medical event, the possibility of considering the medical emergency plan of the historical medical event is greater, and the treatment score value is The purpose of recommending medical emergency plans is to select a more appropriate rescue plan in an emergency situation based on the patient's own situation. Therefore, the rescue success rate is an important indicator for emergency plan recommendation; when the treatment success value is When the value is larger, the survival rate of the patient in the historical medical event corresponding to the information matching value in the first cluster is higher, which means that the medical emergency plan for the historical medical event corresponding to the information matching value in the first cluster is more worthy of recommendation, and the treatment score value is The larger the The larger the value, the better the medical data of the historical medical events corresponding to the information matching values in the rth first cluster match the medical data of the current medical events compared with the other first clusters. Therefore, it is necessary to select the corresponding medical emergency plan from the historical medical events corresponding to the information matching values in the first clusters with higher information matching degrees for recommendation. The treatment score value is The bigger; May be less than 0, to avoid Influence and The correlation between The value range of is adjusted to be greater than 0.
[0070] Step S4: The information matching value and treatment score value of the current medical event and each historical medical event constitute the data point of each historical medical event; based on the position distribution of the data points, the optimal hyperplane in the binary classification algorithm is obtained.
[0071] Specifically, this embodiment uses the SVM algorithm to classify the data points of historical medical events, because in low-dimensional space, data points are more easily separated, and the SVM model is more likely to capture the linear or nonlinear relationship between data, and may face some challenges when processing high-dimensional data points, and the accuracy will be affected. Therefore, the SVM algorithm may have a higher accuracy when processing low-dimensional data points. Among them, the SVM algorithm is a well-known technology for those skilled in the art and will not be described in detail here.
[0072] Since there are many influencing factors of medical events, there are also many types of medical data, and the dimensions of the data points composed of various types of medical data of medical events are relatively high. In order to reduce the error of using the SVM algorithm to recommend medical emergency plans for sudden current medical events for data points with higher dimensions, the data point of each historical medical event is composed of the information matching value and disposal score value between the current medical event and each historical medical event.
[0073] The method for obtaining the optimal hyperplane based on the position distribution of the data points of historical medical events is as follows: establish a two-dimensional coordinate system with the information matching value as the horizontal axis and the treatment score value as the vertical axis; mark all the data points of historical medical events in the two-dimensional coordinate system to obtain the analysis coordinate points of the corresponding data points; cluster the analysis coordinate points in the two-dimensional coordinate system to obtain different second clustering clusters; obtain the centroid of each second clustering cluster, calculate the Euclidean distance between the centroid of each second clustering cluster and the coordinate origin of the two-dimensional coordinate system, and record it as the judgment distance of each second clustering cluster; select the second clustering cluster corresponding to the smallest judgment distance as the first target cluster ; Select a cluster to be determined from all the second clustering clusters, the horizontal coordinate of the centroid of the cluster to be determined is greater than the horizontal coordinate of the first target cluster, and the vertical coordinate of the centroid is greater than the vertical coordinate of the first target cluster; for each cluster to be determined, calculate the mean of the horizontal coordinates of all analysis coordinate points in the cluster to be determined as the comprehensive matching value, and the mean of the vertical coordinates as the comprehensive score value; obtain the mean of the comprehensive matching value and the comprehensive score value of the cluster to be determined and record it as the judgment value of the cluster to be determined; among all the clusters to be determined, take the cluster to be determined corresponding to the largest judgment value as the second target cluster; according to the position distribution of the centroids of the first target cluster and the second target cluster, obtain the optimal hyperplane in the binary classification algorithm.
[0074] It should be noted that the optimal hyperplane in the SVM algorithm can separate the analysis coordinate point in the two-dimensional coordinate system into two parts, and the interval between the two parts is the largest. Because the horizontal axis and the vertical axis of the two-dimensional coordinate system are the information matching value and the treatment score value respectively, when the above two indicators of a historical medical event are larger, the information difference between the historical medical event and the current medical event is more similar and the recommendation degree is the highest. Therefore, the information difference between the historical medical event corresponding to the analysis coordinate point in the upper right position of the two-dimensional coordinate system and the current medical event is more similar and is worth recommending.
[0075] The reasoning process of the expression of the optimal hyperplane in the embodiment of the present invention is based on the two-point equation of a straight line, and using the mathematical knowledge that the product of the slopes of two perpendicular intersecting straight lines is -1, if the straight line passing through the centroid of the first target cluster and the centroid of the second target cluster is taken as the analysis line, the optimal hyperplane is equivalent to the straight line perpendicular to the analysis line and passing through the midpoint of the line connecting the centroid of the first target cluster and the centroid of the second target cluster. The horizontal coordinate of the centroid of the second target cluster is greater than the horizontal coordinate of the centroid of the first target cluster, and the vertical coordinate of the centroid of the second target cluster is greater than the vertical coordinate of the centroid of the first target cluster, then the slope of the analysis line in this embodiment is a positive number, and because the optimal hyperplane is perpendicular to the analysis line, the slope k of the optimal hyperplane is less than 0.
[0076] It should be noted that in the embodiment of the present invention, a density-based spatial clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) is used to cluster the corresponding coordinate points of all data points of historical medical events in a two-dimensional coordinate system. The neighborhood radius Eps takes an empirical value of 3, and the minimum number of points MinPts takes an empirical value of 4. The implementer can set them according to the specific situation; among them, the DBSCAN clustering algorithm and the method of obtaining the centroid of the cluster are well-known technologies and will not be repeated here.
[0077] The expression of the optimal hyperplane is as follows:
[0078]
[0079]
[0080]
[0081] In the formula, y is the ordinate of the point on the optimal hyperplane; z is the abscissa of the point on the optimal hyperplane; k is the slope of the optimal hyperplane; b is the ordinate intercept of the optimal hyperplane; is the ordinate of the centroid of the second target cluster; is the ordinate of the centroid of the first target cluster; is the abscissa of the centroid of the second target cluster; is the horizontal coordinate of the centroid of the first target cluster.
[0082] Step S5: Based on the distance between the data points of all historical medical events and the optimal hyperplane, a medical emergency plan for the current medical event is recommended.
[0083] Specifically, this solution needs to find the historical medical event that has the greatest matching degree with the current medical event; since the slope k of the optimal hyperplane in the embodiment of the present invention is less than 0, the optimal hyperplane divides the analysis coordinate point in the two-dimensional coordinate system into a lower left part and an upper right part, and the historical medical event corresponding to the analysis coordinate point in the upper right position of the two-dimensional coordinate system is similar to the information difference of the current medical event and is worthy of recommendation, and the historical medical event corresponding to the analysis coordinate point on the right side of the optimal hyperplane is similar to the information of the current medical event and is worthy of recommendation. Therefore, the analysis coordinate point on the right side of the optimal hyperplane is obtained and recorded as the target point. Figure 2 A local schematic diagram of the position distribution of coordinate points and the optimal hyperplane in a two-dimensional coordinate system provided by an embodiment of the present invention, Figure 2 As shown, Figure 2 The horizontal axis Q of the two-dimensional coordinate system is the information matching value, the vertical axis K is the treatment score value, the straight line L is the optimal hyperplane, the circle and square points are the data points of historical medical events and the corresponding analysis coordinate points in the two-dimensional coordinate system, among which the square points are the analysis coordinate points on the right side of the optimal hyperplane.
[0084] In the SVM algorithm, the closer the data points are to the hyperplane, the less abnormal they are, and the closer the data points are to the hyperplane, the more similar they are to the current medical event information. The Euclidean distance between each target point and the optimal hyperplane in the two-dimensional coordinate system is obtained, and the medical emergency plan for the historical medical event corresponding to the target point with the smallest Euclidean distance is used as the medical emergency plan for the current medical event.
[0085] It should be noted that the method of judging whether a point is on the left or right side of a straight line is a commonly used method for the positional relationship between a point and a line, and the method of calculating the distance between a point and a straight line is common mathematical knowledge and will not be introduced here.
[0086] So far, the present invention is completed.
[0087] In summary, in an embodiment of the present invention, the information matching value between the current medical event and each historical medical event is obtained, and the information matching value is clustered to obtain a first cluster; based on the difference between the information matching value in each first cluster and the other first clusters, the disposal score value of the historical medical event corresponding to the information matching value in the first cluster is obtained; based on the information matching value and the disposal score value, the data points of the historical medical event are formed, and the optimal hyperplane is obtained based on the position distribution of the data points; based on the distance between the data points of the historical medical event and the optimal hyperplane, the medical emergency plan for the current medical event is recommended. The present invention constitutes the low-dimensional data points of the historical medical event by the acquired information matching value and the disposal score value, and recommends the medical emergency plan for the current medical event based on the position distribution of the data points.
[0088] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0089] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.
Claims
1. A rapid recommendation method for medical emergency plans based on artificial intelligence, characterized in that: The method includes: Obtaining medical details of a current medical event and different historical medical events, as well as a medical emergency plan for each historical medical event, wherein the medical details include different types of medical data; According to the similarity between the current medical event and the same type of medical data of each historical medical event, an information matching value between the current medical event and each historical medical event is obtained; Clustering the information matching values to obtain different first clustering clusters; obtaining a treatment score value of the historical medical event corresponding to each information matching value in each first clustering cluster according to each information matching value in each first clustering cluster and the difference between each first clustering cluster and the information matching values in other first clustering clusters; The information matching value and the treatment score value of the current medical event and each historical medical event constitute the data point of each historical medical event; based on the position distribution of the data points, the optimal hyperplane in the binary classification algorithm is obtained; Recommending a medical emergency plan for the current medical event based on the distance between the data points of all historical medical events and the optimal hyperplane; The method for obtaining the treatment score value of the historical medical event corresponding to each information matching value in each first cluster according to each information matching value in each first cluster and the difference between each first cluster and the information matching value in other first clusters includes: For each historical medical event, if the patient of the historical medical event survives, the treatment identifier of the historical medical event is set to a third preset value; if the patient of the historical medical event does not survive, the treatment identifier of the historical medical event is set to a fourth preset value; the cumulative sum of the treatment identifiers of the historical medical events corresponding to all information matching values in each first cluster is recorded as the treatment success value of each first cluster; the third preset value is greater than the fourth preset value; Combining the difference between the information matching values in each first cluster and the other first clusters, and each information matching value in each first cluster and the treatment success value, obtaining the treatment score value of the historical medical event corresponding to each information matching value in each first cluster; The calculation formula of the treatment score value of the historical medical event corresponding to each information matching value in each first cluster is as follows: ; In the formula, is the treatment score value of the historical medical event corresponding to the j-th information matching value in the r-th first cluster; is the jth information matching value in the rth first cluster; is the maximum value of all the information matching values within the first cluster; is the treatment success value of the rth first cluster; is the mean of all information matching values in the rth first cluster; is the mean of all information matching values in the tth first cluster except the rth first cluster; T is the total number of all first clusters; exp is an exponential function with the natural constant e as the base; Norm is the normalization function.
2. According to claim 1, a method for rapid recommendation of medical emergency plans based on artificial intelligence is characterized in that: The method for obtaining the information matching value between the current medical event and each historical medical event according to the similarity between the current medical event and the same type of medical data of each historical medical event includes: Any historical medical event is selected as the target medical event. If the same type of medical data of the current medical event and the medical details of the target medical event have an intersection, the corresponding medical data is recorded as the data to be tested. The number of types of the data to be tested in all types of medical data of the medical details of the target medical event is counted and recorded as the similarity between the current medical event and the target medical event. The valid identifier of each type of the data to be tested of the medical details of the target medical event is set to a first preset value, and the valid identifier of each type of medical data other than the data to be tested is set to a second preset value; the average of the valid identifiers of all types of medical data of the target medical event is calculated as the data similarity of the target medical event; the first preset value is greater than the second preset value; The information matching value of the current medical event and the target medical event is obtained according to the type similarity and the data similarity between the current medical event and the target medical event; the type similarity and the data similarity are both positively correlated with the information matching value.
3. According to claim 1, a method for rapid recommendation of medical emergency plans based on artificial intelligence is characterized in that: The method for obtaining the optimal hyperplane in the SVM algorithm based on the position distribution of the data points includes: A two-dimensional coordinate system is established with the information matching value as the horizontal axis and the treatment score value as the vertical axis; all data points of historical medical events are marked in the two-dimensional coordinate system to obtain analysis coordinate points of corresponding data points; the analysis coordinate points in the two-dimensional coordinate system are clustered to obtain different second clustering clusters; Obtain the centroid of each second cluster, calculate the Euclidean distance between the centroid of each second cluster and the coordinate origin of the two-dimensional coordinate system, and record it as the judgment distance of each second cluster; select the second cluster corresponding to the smallest judgment distance as the first target cluster; Selecting a cluster to be determined from all the second clustering clusters, the abscissa of the centroid of the cluster to be determined is greater than the abscissa of the first target cluster, and the ordinate of the centroid is greater than the ordinate of the first target cluster; For each undetermined cluster, the mean of the horizontal coordinates of all the analysis coordinate points in the undetermined cluster is calculated and recorded as the comprehensive matching value, and the mean of the vertical coordinates is recorded as the comprehensive score value; the mean of the comprehensive matching value and the comprehensive score value of the undetermined cluster is obtained and recorded as the judgment value of the undetermined cluster; among all the undetermined clusters, the undetermined cluster corresponding to the largest judgment value is taken as the second target cluster; An optimal hyperplane in a binary classification algorithm is obtained according to the position distribution of the centroids of the first target cluster and the second target cluster.
4. The method for rapid recommendation of medical emergency plans based on artificial intelligence according to claim 3, characterized in that: The expression of the optimal hyperplane is as follows: , , ; In the formula, y is the ordinate of the point on the optimal hyperplane; z is the abscissa of the point on the optimal hyperplane; k is the slope of the optimal hyperplane; b is the ordinate intercept of the optimal hyperplane; is the ordinate of the centroid of the second target cluster; is the ordinate of the centroid of the first target cluster; is the abscissa of the centroid of the second target cluster; is the abscissa of the centroid of the first target cluster.
5. The method for rapid recommendation of medical emergency plans based on artificial intelligence according to claim 3, characterized in that: The method for recommending a medical emergency plan for a current medical event based on the distance between the data points of all historical medical events and the optimal hyperplane comprises: Obtain the analysis coordinate point on the right side of the optimal hyperplane and record it as the target point; The Euclidean distance between each target point in the two-dimensional coordinate system and the optimal hyperplane is obtained, and the medical emergency plan of the historical medical event corresponding to the target point with the smallest Euclidean distance is used as the medical emergency plan for the current medical event.
6. The method for rapid recommendation of medical emergency plans based on artificial intelligence according to claim 1, characterized in that: The clustering method for clustering the information matching values to obtain different first cluster clusters is a K-means clustering algorithm.
7. The method for rapid recommendation of medical emergency plans based on artificial intelligence according to claim 3, characterized in that: The clustering method for clustering the analysis coordinate points in the two-dimensional coordinate system to obtain different second cluster clusters is a DBSCAN clustering algorithm.
8. The method for rapid recommendation of medical emergency plans based on artificial intelligence according to claim 1, characterized in that: The binary classification algorithm is a SVM algorithm.
