Emergency pre-examination triage method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of medical triage, in particular to an emergency pre-examination triage method and device, electronic equipment and a storage medium, and the method comprises the steps: fusing the actual illness state data and historical treatment data of an emergency patient to obtain a data set of the emergency patient, inputting the data set of the emergency patient into a triage model, and obtaining a corresponding illness state emergency degree and a preliminary triage department, and in combination with the illness state emergency degree, the preliminary triage department and an actual reception department list at the current time, executing a dynamic adjustment strategy based on a dynamic adjustment rule to obtain a final triage department and consulting room queuing result. According to the method, the actual illness state data and the historical treatment data of the emergency treatment patient are fully combined to predict the illness state emergency degree and the triage department of the model, and the triage department and the treatment sequence are dynamically adjusted, so that the triage result better conforms to the actual situation, the emergency treatment patient is ensured to accurately and timely see a doctor, the waiting time of the emergency treatment patient is shortened, and the efficiency is improved. The medical efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical triage, and is an emergency pre-triage method, device, electronic device and storage medium. Background Art

[0002] Pre-triage is an important link in the treatment of emergency patients. Quickly identifying critically ill patients, accurately triaging them, and arranging medical treatment according to the severity of the condition can make full use of emergency resources, maintain the order of emergency treatment, reduce the waiting time of critically ill patients, ensure patient safety, improve the efficiency of emergency work, and provide a basis for the rational allocation of emergency medical resources, thereby improving the quality and efficiency of emergency medical treatment.

[0003] In the traditional medical model, the pre-triage method for emergency patients is mainly that the medical staff on emergency duty judge the severity of the patient's condition based on the description of the patient's condition by the emergency patient and their experience and feeling. Then, according to dynamic data such as the personal level of the triage doctor and the actual situation of the consulting room, the emergency patient is pre-triaged. After the emergency patient arrives at the hospital, the emergency doctor understands the medical record and clinical condition of the emergency patient before rescue and observes and confirms them. This method lacks sufficient data support, resulting in subjectivity in the diagnosis process. Moreover, in the emergency scenario, it is very likely that the patient cannot communicate verbally, making it difficult to clearly describe the historical condition. In addition, the levels of the medical staff on emergency duty vary, and the judgment experience of the triage process is also inconsistent, which may lead to triage errors and delay the condition of some emergency patients.

[0004] With the rapid development of Internet and artificial intelligence technologies, artificial intelligence has been introduced into the pre-triage of emergency departments. For example: Existing publicly disclosed patent document 1, with the publication number CN118761596B, discloses an intelligent triage method and system. The method includes: S1, obtaining the patient's condition information and identifying the key information in the condition information; S2, constructing a dynamic condition model based on the patient's historical data, medical knowledge, and key information; S3, reasoning about the key information according to the large model and knowledge base to generate multiple condition hypotheses of the patient and the credibility of each condition hypothesis; the large model is used to identify and predict the possibility of the development of the condition; the knowledge base is used to provide relevant medical rules and facts; S4, analyzing the urgency and triage requirements of the patient according to the condition model and condition hypotheses by using the large model to generate one or more triage plans; S5, sorting out the triage report of the patient according to the triage plan; the triage report at least includes multiple items of the following information: the basic information of the patient, the description of the condition, triage suggestions, urgency, treatment suggestions, required medical resources, and precautions.

[0005] The existing publicly disclosed patent document II, with the publication number CN114678113B, discloses an intelligent emergency pre - triage system based on a convolutional neural network. The intelligent emergency pre - triage system includes the following steps: The first step: constructing an evaluation index feature cluster matrix; The second step: constructing a grayscale matrix fusion; The third step: training a convolutional neural network model; The fourth step: determining the grading and division of the consultation channels; The fifth step: differential event handling and re - evaluation; The sixth step: emergency outcome and re - training; It solves the problem that in the prior art, the triage efficiency for adult emergency patients is relatively low, and it is easy to cause the situation that patients with more serious conditions do not receive priority treatment, resulting in the aggravation of the condition, and improves the efficiency of rescuing emergency patients.

[0006] However, in the existing methods of introducing artificial intelligence into the pre - triage of emergencies, the input data of the model is mostly of a single input data type, that is, the actual condition data before. However, emergencies are mostly urgent situations, and problems such as unclear descriptions may occur. Therefore, a single input data type is likely to cause inaccurate model output results. Summary of the Invention

[0007] The present invention provides an emergency pre - triage method and device, which overcome the above - mentioned deficiencies of the prior art and can effectively solve the problem that in the existing traditional medical model, the manual triage method is prone to triage errors, resulting in the delay of the conditions of some emergency patients.

[0008] One of the technical solutions of the present invention is achieved by the following measures: An emergency pre - triage method includes: Fusing the actual condition data and historical medical treatment data of emergency patients to obtain a data set of emergency patients, where the data set includes identity characteristics and condition characteristics; Inputting the data set of emergency patients into a triage model to obtain the corresponding degree of urgency of the condition and the preliminary triage department, where the triage model is obtained by machine learning using a number of sample data, and each sample data includes the data set of a certain emergency patient and identification information marking the degree of urgency of the condition and the triage department, and the data set includes identity characteristics and condition characteristics; Combining the degree of urgency of the condition, the preliminary triage department and the actual reception department list at the current time, and implementing a dynamic adjustment strategy based on dynamic adjustment rules to obtain the final triage department and the waiting queue result in the consulting room, where the dynamic adjustment rules include: a. If the preliminary triage department is normally receiving patients, re - sort the waiting queue result according to the degree of urgency of the condition to determine the waiting queue result in the consulting room; b. If the preliminary triage department is not normally receiving patients, adjust to the consulting room with the highest relevance to the consulting room or the highest relevance to the attending doctor, and re - sort the waiting queue result according to the degree of urgency of the condition to determine the waiting queue result in the consulting room.

[0009] The following is a further optimization or / and improvement of the above - mentioned technical solution of the invention: The construction steps of the above - mentioned triage model include: Obtain a number of historical medical treatment data and divide them into a training set and a test set according to a ratio, where each piece of historical medical treatment data is a data set of a certain emergency patient, and the data set includes identity characteristics and disease conditions characteristics; Randomly select in the training set to obtain a number of random subsets, and establish a decision tree corresponding to each random subset; Based on each random subset, train the corresponding decision tree, and form a random forest model by combining multiple trained decision trees, where the hyperparameters of the random forest model are adjusted using a validation method during training; Use the test set to evaluate the random forest model. If the evaluation result does not meet the preset standard, retrain and adjust the parameters of the random forest model. If the evaluation result meets the preset standard, use this random forest model as the triage model.

[0010] The construction steps of the above triage model include: Obtain a historical medical treatment data training set and a historical medical treatment data test set, where both the historical medical treatment data training set and the historical medical treatment data test set include a number of historical medical treatment data sets, and each piece of historical medical treatment data is a data set of a certain emergency patient, and the data set includes identity characteristics and disease conditions characteristics; Divide the historical medical treatment data training set into multiple sub-data sets, and use each sub-data set to train different machine learning algorithms respectively to obtain multiple primary triage models; Use the historical medical treatment data test set to test each primary triage model, and evaluate the test results of each primary triage model based on the model regression evaluation index, and select the primary triage model with the best evaluation result as the final triage model.

[0011] The above process of fusing the actual disease condition data and historical medical treatment data of emergency patients to obtain the data set of emergency patients includes: Obtain the identity information of the emergency patient, and establish the actual disease condition data set and the historical medical treatment data set of the emergency patient; Based on the set feature types, extract the corresponding feature values from the actual disease condition data set and the historical medical treatment data set respectively to form an actual disease condition feature set and a historical medical treatment feature set; Combine the set feature fusion rules to fuse the actual disease condition feature set and the historical medical treatment feature set to obtain the feature value set of the emergency patient, where the feature fusion rules include: a. When the feature values of the same fixed field coincide, the feature values remain unchanged; b. When the feature values of the same fixed field do not coincide, the feature values are mainly based on the actual disease condition feature set; c. For different fixed fields, both are retained; Convert the feature value set of the emergency patient to generate a feature vector matrix.

[0012] The steps of extracting corresponding feature values from the actual condition dataset and the historical visit dataset based on the set feature types to form the actual condition feature set and the historical visit feature set are the same. Among them, extracting the corresponding feature values from the actual condition dataset based on the set feature types to form the actual condition feature set includes: The set feature type extracts feature values from the actual condition dataset by using the fixed field matching method; After the fixed field matching ends, it is judged whether the list area corresponding to the fixed field is empty; In response to no, credibility weight values are assigned to the extracted feature values, and the actual condition feature set is formed; In response to yes, the actual condition data segment is searched by using the keyword search method to obtain the suspected feature values corresponding to the fixed field with an empty list area; The suspected feature values are subjected to format verification. If the feature value format standard is met, they are retained and written into the list area corresponding to the fixed field. If the feature value format standard is not met, the keyword search method is continued to search until all actual condition data segments are searched, credibility weight values are assigned to the extracted feature values, and the actual condition feature set is formed.

[0013] The second technical solution of the present invention is realized by the following measures: An emergency pre-triage device, including: A feature extraction unit that fuses the actual condition data and the historical visit data of the emergency patient to obtain a dataset of the emergency patient, where the dataset includes identity features and condition features; A model triage unit that inputs the dataset of the emergency patient into the triage model to obtain the corresponding disease emergency level and the preliminary triage department, where the triage model is obtained by machine learning using a number of sample data, and each sample data includes the dataset of a certain emergency patient and the identification information marking the disease emergency level and the triage department, and the dataset includes identity features and condition features; A triage adjustment unit that combines the disease emergency level, the preliminary triage department, and the actual reception department list at the current time, and executes a dynamic adjustment strategy based on the dynamic adjustment rule to obtain the final triage department and the consultation room queuing result, where the dynamic adjustment rule includes: a. If the preliminary triage department normally receives patients, the queuing result is re-sorted according to the disease emergency level to determine the consultation room queuing result; b. If the preliminary triage department does not normally receive patients, it is adjusted to the consultation room with the highest consultation room relevance or doctor relevance, and the queuing result is re-sorted according to the disease emergency level to determine the consultation room queuing result.

[0014] The following is a further optimization or / and improvement of the above-mentioned invention technical solution: The above-mentioned feature extraction unit includes a first extraction unit, a second extraction unit, and a fusion unit. The first extraction unit extracts corresponding feature values from the actual condition dataset based on the set feature type to form an actual condition feature set. The second extraction unit extracts corresponding feature values from the historical visit dataset based on the set feature type to form a historical visit feature set. The structures of the two are the same. Among them, the first extraction unit includes: A first extraction module that extracts feature values from the actual condition dataset using the fixed field matching method based on the set feature type; A judgment module that, after the fixed field matching is completed, judges whether the list area corresponding to the fixed field is empty; An output module that, in response to "no", assigns a credibility weight value to the extracted feature values and forms an actual condition feature set; A second extraction module that, in response to "yes", searches the actual condition data segment using the keyword search method to obtain suspected feature values corresponding to the fixed field with an empty list area, performs format verification on the suspected feature values, retains them if they meet the feature value format standard and writes them into the list area corresponding to the fixed field, and continues to search using the keyword search method if they do not meet the feature value format standard until all actual condition data segments are searched, assigns a credibility weight value to the extracted feature values, and forms an actual condition feature set.

[0015] The above also includes a first model construction unit or / and a second model construction unit; The first model construction unit includes: Obtain a historical visit data training set and a historical visit data test set, where both the historical visit data training set and the historical visit data test set include several historical visit datasets. Each historical visit data is a dataset of a certain emergency patient, and the dataset includes identity characteristics and condition characteristics; Divide the historical visit data training set into multiple sub-datasets, and use each sub-dataset to train different machine learning algorithms respectively to obtain multiple primary triage models; Use the historical visit data test set to test each primary triage model, and evaluate the test results of each primary triage model based on the model regression evaluation index, and select the primary triage model with the best evaluation result as the final triage model; The second model construction unit includes: Obtain a historical visit data training set and a historical visit data test set, where both the historical visit data training set and the historical visit data test set include several historical visit datasets. Each historical visit data is a dataset of a certain emergency patient, and the dataset includes identity characteristics and condition characteristics; The historical medical visit data training set is divided into multiple sub-datasets, and each sub-dataset is used to train different machine learning algorithms to obtain multiple primary triage models; Use the historical medical visit data test set to test each primary triage model, and evaluate the test results of each primary triage model based on the model regression evaluation index. Select the primary triage model with the best evaluation result as the final triage model.

[0016] The third technical solution of the present invention is achieved by the following measures: A storage medium, on which a computer program that can be read by a computer is stored, and the computer program is set to execute the steps in the emergency pre-triage when running.

[0017] The fourth technical solution of the present invention is achieved by the following measures: An electronic device, including a processor and a memory, and a computer program is stored in the memory. The computer program is loaded and executed by the processor to implement the steps in the emergency pre-triage.

[0018] The present invention can fully combine the actual condition data and historical medical visit data of emergency patients to obtain the input data for model prediction, provide accurate data support for predicting the urgency of the condition and the triage department, further quickly obtain the corresponding condition urgency and preliminary triage department based on machine learning, and dynamically adjust the triage department and the order of medical treatment in combination with the actual reception department list of the condition urgency, preliminary triage department and current time, so that the triage result is more in line with the actual situation, ensure that emergency patients seek medical treatment accurately and in time, reduce the waiting time of emergency patients, and improve the medical efficiency. Brief Description of the Drawings

[0019] Att Figure 1 It is a schematic diagram of an implementation environment provided by the present invention.

[0020] Att Figure 2 It is a schematic diagram of the emergency pre-triage method flow provided by the present invention.

[0021] Att Figure 3 It is a schematic diagram of a triage model method flow provided by the present invention.

[0022] Att Figure 4 It is another schematic diagram of a triage model method flow provided by the present invention.

[0023] Att Figure 5 It is a schematic diagram of the method flow for constructing the dataset of emergency patients provided by the present invention.

[0024] Att Figure 6 It is a schematic diagram of the method flow for constructing the actual condition feature set provided by the present invention.

[0025] AttFigure 7 Schematic structural diagram of the emergency pre-triage device provided by the present invention.

[0026] Appendix Figure 8 Schematic structural diagram of the feature extraction unit provided by the present invention. Specific embodiments

[0027] The present invention is not limited by the following embodiments, and the specific implementation manners can be determined according to the technical solutions of the present invention and the actual situation.

[0028] Those skilled in the art of the present technology can understand that, unless specifically stated, the "module" or "unit" in the embodiments of the present application refers to a computer program with a predetermined function or a part of a computer program, which works together with other relevant parts to achieve a predetermined goal, and can be fully or partially implemented by using software, hardware (such as a processing circuit or a memory), or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of the overall module or unit that includes the function of the module or unit.

[0029] In addition, "a plurality of" in the embodiments of the present application means two or more, and "first" and "second" are used for distinguishing descriptions, and cannot be understood as implying relative importance.

[0030] In view of the problems in the traditional medical mode, in the emergency medical staff's manual subjective triage, it is very likely that the patient cannot communicate verbally, it is difficult to clearly describe the historical condition, and the levels of the emergency medical staff on duty are different, and the experience in the triage process judgment is also inconsistent. There may be triage errors, resulting in the delay of the conditions of some emergency patients.

[0031] An embodiment of the present application provides an emergency pre - triage method. The actual condition data and historical medical treatment data of emergency patients are integrated to obtain a data set of emergency patients, where the data set includes identity characteristics and condition characteristics. The data set of the emergency patient is input into a triage model to obtain the corresponding emergency level of the condition and the preliminary triage department. The triage model is obtained by machine learning using a number of sample data, and each sample data includes the data set of a certain emergency patient and identification information marking the emergency level of the condition and the triage department. The data set includes identity characteristics and condition characteristics. Combining the emergency level of the condition, the preliminary triage department and the actual reception department list at the current time, a dynamic adjustment strategy is executed based on dynamic adjustment rules to obtain the final triage department and the waiting queue result in the consulting room. The dynamic adjustment rules include: a. If the preliminary triage department is normally receiving patients, the waiting queue result is re - sorted according to the emergency level of the condition to determine the waiting queue result in the consulting room. b. If the preliminary triage department is not normally receiving patients, it is adjusted to the consulting room with the highest relevance to the consulting room or the highest relevance to the attending doctor, and the waiting queue result is re - sorted according to the emergency level of the condition to determine the waiting queue result in the consulting room.

[0032] Among them, the method provided by the embodiment of the present application may involve Artificial Intelligence (AI) technology and can be implemented based on artificial intelligence technology. For example, in the way of deep learning, a corresponding model is obtained by training with samples.

[0033] Machine Learning (ML) is an interdisciplinary subject that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence.

[0034] Deep Learning (DL) specifically refers to machine learning based on deep neural network models and methods. It is developed on the basis of algorithm models such as statistical machine learning and artificial neural networks, combined with the development of contemporary big data and high computing power. The most important technical feature of deep learning is the ability to automatically extract features.

[0035] The above - mentioned machine learning and deep learning usually include technologies such as neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and rote learning.

[0036] As shown in the appendix Figure 1 It shows a schematic diagram of the implementation environment provided by an embodiment of the present application. The implementation environment may include: training equipment and using equipment.

[0037] Both the training device and the using device are computer devices; optionally, the computer device is a terminal device, such as electronic devices like mobile phones, tablets, PCs (Personal Computers), etc.; or, the computer device is a server, which can be a single server, a server cluster composed of multiple servers, or a cloud computing service center, and the embodiments of the present application do not limit this.

[0038] The training device refers to a computer device with model training and learning capabilities. Optionally, the training device has the ability to obtain a model and trains and learns it according to application requirements. For example, the training device obtains a model from other devices through the network, and then trains it with training samples according to application requirements so that the model can have the ability to obtain the urgency of the condition and the triage department; optionally, the training device has the ability to build a model, which can build a model by itself according to application requirements and then train and learn it. For example, in order to obtain the urgency of the condition and the triage department based on the dataset of emergency patients, the training device builds a model by itself and then trains and learns it with samples according to application requirements.

[0039] The using device refers to a computer device with model usage requirements. Optionally, the using device obtains a model from other devices through the network according to application requirements. For example, the using device has the requirement of predicting the urgency of the condition and the triage department, and it can obtain a model that has completed training and learning to predict the urgency of the condition and the triage department from other devices through the network and use this model to predict the urgency of the condition and the triage department.

[0040] Based on this, the technical solutions of the present application will be introduced and illustrated below with several examples.

[0041] Example 1: As shown in the appendix Figure 2 The embodiments of the present invention disclose an emergency pre-triage method, including: Step S110, fusing the actual condition data and historical medical treatment data of emergency patients to obtain a dataset of emergency patients, where the dataset includes identity characteristics and condition characteristics; Step S120, inputting the dataset of emergency patients into a triage model to obtain the corresponding urgency of the condition and the preliminary triage department, where the triage model is obtained by machine learning using a number of sample data, and each sample data includes the dataset of a certain emergency patient and identification information marking the urgency of the condition and the triage department, and the dataset includes identity characteristics and condition characteristics; Step S130: Based on the actual reception department list at the current time in combination with the emergency level of the condition, the preliminary triage department, and the dynamic adjustment rules, execute the dynamic adjustment strategy to obtain the final triage department and the waiting queue results in the consultation room. The dynamic adjustment rules include: a. If the preliminary triage department is receiving patients normally, re - sort the waiting queue results according to the emergency level of the condition to determine the waiting queue results in the consultation room; b. If the preliminary triage department is not receiving patients normally, adjust to the consultation room with the highest relevance to the consultation room or the highest relevance to the attending doctor, and re - sort the waiting queue results according to the emergency level of the condition to determine the waiting queue results in the consultation room.

[0042] In this embodiment, the above - mentioned step S110 integrates the actual condition data and historical medical treatment data of the emergency patients to obtain a dataset of the emergency patients, enabling mutual supplementation between the two, thereby obtaining a more accurate dataset for subsequent triage and providing stable data support for accurate triage.

[0043] In this embodiment, the above - mentioned step S130 dynamically adjusts the preliminary triage department based on the emergency level of the condition and the preliminary triage department output in step S120, in combination with the actual reception department list at the current time, making the triage result more in line with the actual situation, and thus ensuring the effective medical treatment of emergency patients.

[0044] The actual reception department list at the current time in step S130 includes the normally receiving departments at the current time, the attending doctor schedules of the normally receiving departments, the patient waiting queue situations of the normally receiving departments, etc.

[0045] In step S130, the dynamic adjustment strategy is executed based on the dynamic adjustment rules to obtain the final triage department and the waiting queue results in the consultation room. The dynamic adjustment rules include: a. If the preliminary triage department is receiving patients normally, re - sort the waiting queue results according to the emergency level of the condition to determine the waiting queue results in the consultation room; b. If the preliminary triage department is not receiving patients normally, adjust to the consultation room with the highest relevance to the consultation room or the highest relevance to the attending doctor, and re - sort the waiting queue results according to the emergency level of the condition to determine the waiting queue results in the consultation room.

[0046] Specifically, match the preliminary triage department with the actual list of consulting departments at the current time. If the preliminary triage department is in normal consultation, re - sort the current queuing results of this consulting room according to the urgency of the condition (for example, if the urgency of the condition is high, medium, or low, then arrange them at the head, middle, or rear of the queue, or determine the waiting time for the emergency patient to see a doctor according to the urgency of the condition, that is, the higher the urgency of the condition, the shorter the waiting time to see a doctor, and then sort according to the consultation time. Specifically, it can be set according to the actual situation, or manual sorting can be introduced, and medical staff can set the order of seeing a doctor by combining on - site judgment and the urgency of the condition). If the preliminary triage department is not in consultation, it can be adjusted to the consulting room with the highest relevance to the department or the consulting doctor. This adjustment method can be adjusted through the pre - set matching relationship between departments and doctors, or manual adjustment can be introduced.

[0047] In summary, an emergency pre - triage method disclosed in an embodiment of the present invention can fully combine the actual condition data and historical consultation data of emergency patients to obtain input data for model prediction, provide accurate data support for predicting the urgency of the condition and the triage department, further quickly obtain the corresponding urgency of the condition and the preliminary triage department based on machine learning, and dynamically adjust the triage department and the order of seeing a doctor in combination with the urgency of the condition, the preliminary triage department, and the actual list of consulting departments at the current time, making the triage result more in line with the actual situation, ensuring that emergency patients can see a doctor accurately and in a timely manner, reducing the waiting time of emergency patients, and improving medical efficiency.

[0048] Example 2: As shown in the appendix Figure 3 This embodiment of the present invention discloses an emergency pre - triage method, which is a further optimization of the above - mentioned embodiment. The construction steps of a triage model include: Step S210, obtain a number of historical consultation data and divide them into a training set and a test set according to a certain proportion. Each historical consultation data is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics.

[0049] Step S220, randomly extract in the training set to obtain a number of random subsets, and establish a decision tree corresponding to each random subset.

[0050] In this step, m samples can be randomly sampled with replacement from the training set using the Bootstraping method, and sampling is performed n_tree times to generate n_tree random subsets.

[0051] Step S230, train the corresponding decision tree based on each random subset, and form a random forest model with the trained multiple decision trees. When training, use a validation method to adjust the hyperparameters of the random forest model.

[0052] In this step, a decision tree model is trained for each random subset. When each decision tree is split, the optimal feature is selected for splitting according to the information gain, information gain ratio, or Gini index until all training examples at this node belong to the same class.

[0053] In this step, multiple trained decision trees are combined to form a random forest model. For the model classification problem, the final classification result is determined by voting. For the regression problem, the final prediction result is determined by the mean of the predicted values of multiple trees.

[0054] In this step, the validation method used during training can include, but is not limited to, five-fold cross-validation. Five-fold cross-validation divides the data into five equal parts. Each time, one part is taken for testing, and the remaining parts are used for training. After five experiments, the average value is calculated.

[0055] Step S240: Use the test set to evaluate the random forest model. If the evaluation result does not meet the preset standard, retrain and adjust the parameters of the random forest model. If the evaluation result meets the preset standard, use this random forest model as the triage model.

[0056] In this step, when using the test set to evaluate the random forest model, the evaluation methods can include, but are not limited to, the ROC curve and AUC value, F1 score, etc.

[0057] This embodiment constructs a model using the random forest algorithm. The random forest can handle a large amount of data and complex features, capture the non-linear relationships and interaction effects in the data, and can effectively reduce the risk of overfitting and improve the generalization ability of the model, making the prediction of the emergency degree of the condition and the preliminary triage department more accurate.

[0058] Embodiment 3: As shown in the appendix Figure 4 This embodiment of the present invention discloses an emergency pre-triage method, which is a further optimization of the above embodiment. The construction steps of another triage model include: Step S310: Obtain a historical visit data training set and a historical visit data test set. Both the historical visit data training set and the historical visit data test set include several historical visit data sets. Each historical visit data is a data set of a certain emergency patient. The data set includes identity characteristics and condition characteristics. Step S320: Divide the historical visit data training set into multiple sub-data sets, and use each sub-data set to train different machine learning algorithms respectively to obtain multiple primary triage models. Step S330: Use the historical visit data test set to test each primary triage model, and evaluate the test results of each primary triage model based on the model regression evaluation index. Select the primary triage model with the best evaluation result as the final triage model.

[0059] In this embodiment, the historical medical visit data training set is divided into multiple sub-datasets, and each sub-dataset is used to train different machine learning algorithms respectively, as follows: The historical medical visit data training set is divided into multiple sub-datasets, and each sub-dataset is further divided into a training set and a test set; The training set in the sub-dataset is used to train different machine learning algorithms, and the test set is used to test the trained model, and the model that meets the test requirements is output as the primary triage model.

[0060] Furthermore, if the historical medical visit data training set is divided into 5 sub-datasets and there are 6 machine learning algorithms set, then 5×6 primary triage models will be obtained.

[0061] The above-mentioned multiple machine learning algorithms can include but are not limited to polynomial regression algorithm, neural network regression algorithm, nearest neighbor method regression algorithm, linear regression algorithm, decision tree regression algorithm, ridge regression algorithm, elastic net model regression algorithm, support vector machine regression algorithm, XGBoost algorithm. The polynomial regression algorithm, neural network regression algorithm, nearest neighbor method regression algorithm, linear regression algorithm, decision tree regression algorithm, ridge regression algorithm, elastic net model regression algorithm, support vector machine regression algorithm, and XGBoost algorithm are all existing well-known technologies. Each machine learning algorithm has its own advantages and disadvantages. The model trained by a single machine learning algorithm cannot be applicable to the prediction of the urgency of the condition and the triage department in various scenarios, and cannot effectively guarantee the prediction accuracy of the urgency of the condition and the triage department. However, the present invention combines multiple machine learning algorithms, and finds the one with the best regression effect, that is, the highest prediction accuracy, in different scenarios as the final model, effectively guaranteeing the prediction accuracy of the urgency of the condition and the triage department.

[0062] Example 4: As shown in the appendix Figure 5 This embodiment of the present invention discloses an emergency pre-triage method, which is a further optimization of the above embodiment. The dataset of emergency patients is obtained by integrating the actual condition data and historical medical visit data of emergency patients, including: Step S410, obtain the identity information of the emergency patient, and establish the actual condition dataset and historical medical visit dataset of the emergency patient; In this step, establishing the actual condition dataset and historical medical visit dataset of the emergency patient includes: combining the patient's own condition description data, basic physical sign data collected by basic detection equipment, and condition description data collected by on-site doctors and patients to form the actual condition dataset. It should be noted here that the condition description data can be voice data. When forming the actual condition dataset, the voice data can be converted into text data; according to the identity information of the emergency patient, the corresponding historical medical visit data is extracted from the hospital medical visit database to form the historical medical visit dataset.

[0063] Further, establish the actual condition dataset and historical visit dataset of emergency patients, and the actual condition dataset and historical visit dataset can be preprocessed, such as removing null values, data that has nothing to do with the condition, etc.

[0064] Step S420, extract the corresponding feature values from the actual condition dataset and historical visit dataset respectively based on the set feature types to form an actual condition feature set and a historical visit feature set; In this step, as shown in the appendix Figure 6 extract the corresponding feature values from the actual condition dataset based on the set feature types to form an actual condition feature set, including: Step S421, use the fixed field matching method to extract feature values from the actual condition dataset based on the set feature types; This step specifically includes: (1) Use the field customization function to determine several fixed fields based on the set feature types, where the fixed fields refer to structured fields used to store discrete data segments; Among them, the several fixed fields determined can be but are not limited to the following: { "Heart rate": "heart_rate", "Blood pressure": "blood_pressure", "Blood oxygen level": "blood_oxygen_level", "Body temperature": "body_temperature", "Pain location": "pain_location", "Pain intensity": "pain_intensity", "Medical history": "medical_history", "Age": "age", "Gender": "gender" } (2) Initially divide the data characters in the actual condition dataset according to the type and delimiter to obtain actual condition data segments; (3) Match the actual condition data segments with several fixed fields; (4) Write the successfully matched actual condition data segments (i.e., data characters) into the list area corresponding to the fixed fields as the feature values of the fixed fields, and set the list area corresponding to the fixed fields that do not match to be empty.

[0065] Step S422, after the fixed-field matching ends, determine whether the list area corresponding to the fixed field is empty; Step S423, in response to a negative determination, assign a credibility weight value to the extracted feature values and form an actual condition feature set; Step S424, in response to a positive determination, use the keyword search method to search the actual condition data segment to obtain the suspected feature values corresponding to the fixed fields with an empty list area; Step S425, perform format verification on the suspected feature values. If the feature value format standard is met, retain them and write them into the list area corresponding to the fixed field. If the feature value format standard is not met, continue to search using the keyword search method until all actual condition data segments are searched, and then return to Step S423 to assign a credibility weight value to the extracted feature values and form an actual condition feature set.

[0066] The rule for assigning a credibility weight value to the extracted feature values in the above steps can be that for fields obtained through the fixed-field matching method, the weight value is 1; for fields obtained through the keyword search method, the weight value is 0.5; and for fields that cannot be obtained, the weight value is 0.

[0067] In this embodiment, the same parts between the steps of forming the actual condition feature set and the steps of forming the historical visit feature set will not be elaborated. The difference is that the fixed fields used when adopting the fixed-field matching method for the two can be different, and the specific types can be set according to the actual situation. The formation of the historical visit feature set can be but is not limited to: { "Heart rate": "heart_rate", "Blood pressure": "blood_pressure", "Blood oxygen level": "blood_oxygen_level", "Body temperature": "body_temperature", "Pain location": "pain_location", "Pain intensity": "pain_intensity", "Past diagnosis": "past_diagnosis", "Medical history": "medical_history", "Age": "age", "Gender": "gender" } Step S430: Combine the actual condition feature set and the historical medical visit feature set according to the set feature fusion rules to obtain the feature value set of the emergency patient. The feature fusion rules include: a. When the feature values of the same fixed field coincide, the feature value remains unchanged; b. When the feature values of the same fixed field do not coincide, the feature value is mainly based on the actual condition feature set; c. For different fixed fields, all are retained. Step S440: Convert the feature value set of the emergency patient to generate a feature vector matrix.

[0068] In this step, the conversion of the feature value set of the emergency patient can use, but is not limited to, the principal component analysis method.

[0069] For example, if the feature values of the emergency patient are as follows: { "heart_rate": 80, "blood_pressure": "130 / 85", "blood_oxygen_level": 97, "body_temperature": 37.0, "pain_location": "chest", "pain_intensity": "severe", "medical_history": "hypertension", "age": 45, "gender": "male" } The feature value conversion method used is as follows: { "heart_rate": numerical correspondence method "blood_pressure": numerical correspondence method, generating two vectors, "blood_oxygen_level": numerical correspondence method "body_temperature": numerical correspondence method "pain_location": dictionary correspondence method, such as (chest: 1; abdomen: 2; limbs: 3) "pain_intensity": dictionary correspondence method, such as (mild: 1; moderate: 2; severe: 3) "medical_history": dictionary correspondence method, using the national ICD 10 dictionary, and then taking the absolute value of its number "age": 45, numerical correspondence method "gender": dictionary correspondence method, e.g., (female: 1; male: 2) } The generated feature vector matrix after conversion is as follows: [80, 130, 85, 97, 37, 1, 3, 10, 45, 2] Example 5: As shown in the appendix Figure 7 As shown, an emergency pre - triage device disclosed in an embodiment of the present invention includes: A feature extraction unit that fuses the actual condition data and historical medical treatment data of emergency patients to obtain a data set of emergency patients, where the data set includes identity features and condition features; A model triage unit that inputs the data set of emergency patients into a triage model to obtain the corresponding degree of illness urgency and preliminary triage department, where the triage model is obtained by machine learning using a number of sample data, and each sample data includes the data set of a certain emergency patient and identification information marking the degree of illness urgency and triage department, and the data set includes identity features and condition features; A triage adjustment unit that combines the degree of illness urgency, preliminary triage department, and the actual reception department list at the current time, and executes a dynamic adjustment strategy based on dynamic adjustment rules to obtain the final triage department and the waiting queue result in the consultation room, where the dynamic adjustment rules include: a. If the preliminary triage department is normally receiving patients, re - sort the waiting queue result according to the degree of illness urgency to determine the waiting queue result in the consultation room; b. If the preliminary triage department is not normally receiving patients, adjust to the consultation room with the highest relevance to the consultation room or the highest relevance to the attending doctor, and re - sort the waiting queue result according to the degree of illness urgency to determine the waiting queue result in the consultation room.

[0070] Example 6: As shown in the appendix Figure 8 As shown, an emergency pre - triage device disclosed in an embodiment of the present invention is a further optimization of the above - mentioned embodiment. The feature extraction unit includes a first extraction unit, a second extraction unit, and a fusion unit. The first extraction unit extracts corresponding feature values from the actual condition data set based on the set feature types to form an actual condition feature set. The second extraction unit extracts corresponding feature values from the historical medical treatment data set based on the set feature types to form a historical medical treatment feature set. The two have the same structure. Among them, the first extraction unit includes: A first extraction module that extracts feature values from the actual condition data set using the fixed - field matching method based on the set feature types; A judgment module that, after the fixed - field matching is completed, judges whether the list area corresponding to the fixed field is empty; An output module that, in response to "no", assigns credibility weight values to the extracted feature values and forms an actual condition feature set; The second extraction module, in response to yes, searches the actual condition data segments using the keyword search method to obtain the suspected feature values corresponding to the fixed fields with an empty list area, performs format verification on the suspected feature values, if they meet the feature value format standard, retains them and writes them into the list area corresponding to the fixed fields, if they do not meet the feature value format standard, continues to search using the keyword search method until all actual condition data segments are searched, assigns credibility weight values to the extracted feature values, and forms an actual condition feature set.

[0071] Embodiment 7: The embodiment of the present invention discloses an emergency pre-triage device, which is a further optimization of the above embodiment, and further includes a first model construction unit, including: Obtain a historical medical treatment data training set and a historical medical treatment data test set, where both the historical medical treatment data training set and the historical medical treatment data test set include several historical medical treatment data sets, and each historical medical treatment data is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics; Divide the historical medical treatment data training set into multiple sub-data sets, and use each sub-data set to train different machine learning algorithms respectively to obtain multiple primary triage models; Use the historical medical treatment data test set to test each primary triage model, and evaluate the test results of each primary triage model based on the model regression evaluation index, and select the primary triage model with the best evaluation result as the final triage model.

[0072] Embodiment 8: The embodiment of the present invention discloses an emergency pre-triage device, which is a further optimization of the above embodiment, and further includes a second model construction unit, including: Obtain a historical medical treatment data training set and a historical medical treatment data test set, where both the historical medical treatment data training set and the historical medical treatment data test set include several historical medical treatment data sets, and each historical medical treatment data is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics; Divide the historical medical treatment data training set into multiple sub-data sets, and use each sub-data set to train different machine learning algorithms respectively to obtain multiple primary triage models; Use the historical medical treatment data test set to test each primary triage model, and evaluate the test results of each primary triage model based on the model regression evaluation index, and select the primary triage model with the best evaluation result as the final triage model.

[0073] Embodiment 9: The embodiment of the present invention discloses a storage medium, on which a computer program readable by a computer is stored, and the computer program is set to execute the emergency pre-triage method when running.

[0074] The above storage medium may include, but is not limited to: various media that can store computer programs such as USB flash drives, read-only memories, mobile hard disks, magnetic disks, or optical discs, etc.

[0075] Embodiment 10: An embodiment of the present invention discloses an electronic device, including a processor and a memory. A computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the emergency pre-triage method.

[0076] The above processor may be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in combination with the disclosure of the present application. It can also be a combination that implements computing functions, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, and so on. The memory may include, but is not limited to: various media that can store computer programs such as USB flash drives, read-only memories, mobile hard disks, magnetic disks, or optical discs, etc.

[0077] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0078] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0079] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means embodying the functionality specified in the flowchart Figure 1 one or more flowcharts and / or boxes Figure 1 specified in the box or boxes.

[0080] The above content is only for the various specific embodiments of the present application, which have strong adaptability and implementation effects. However, the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all such changes or substitutions should be covered by the protection scope of the present application. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. An emergency pre-examination and triage method, characterized in that: include: The actual condition data and historical medical data of emergency patients are integrated to obtain a data set of emergency patients, where the data set includes identity features and condition features; The data set of emergency patients is input into the triage model to obtain the corresponding urgency of the disease and the preliminary triage department. The triage model is obtained by machine learning using a number of sample data. Each sample data includes a data set of an emergency patient and identification information marking the urgency of the disease and the triage department. The data set includes identity characteristics and disease characteristics. Combined with the urgency of the disease, the preliminary triage department, and the actual list of receiving departments at the current time, the dynamic adjustment strategy is executed based on the dynamic adjustment rules to obtain the final triage department and clinic queuing results. The dynamic adjustment rules include: a. If the initial triage department is accepting patients normally, the queuing results will be re-sorted according to the urgency of the condition to determine the clinic queuing results; b. If the initial triage department is not accepting patients normally, the clinic will be adjusted to the clinic with the highest clinic relevance or the highest relevance to the receiving doctor, and the queuing results will be re-sorted according to the urgency of the condition to determine the clinic queuing results.

2. The emergency pre-examination and triage method according to claim 1, characterized in that: The steps of constructing the triage model include: Obtain a number of historical medical treatment data and divide them into a training set and a test set in proportion, where each historical medical treatment data is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics; Randomly extract from the training set to obtain several random subsets, and establish a decision tree corresponding to each random subset; Based on each random subset, the corresponding decision tree is trained, and the trained multiple decision trees are combined into a random forest model, wherein the hyperparameters of the random forest model are adjusted by the verification method during training; The random forest model is evaluated using the test set. If the evaluation result does not meet the preset standard, the random forest model is retrained and parameter adjusted. If the evaluation result meets the preset standard, the random forest model is used as the triage model.

3. The emergency pre-examination and triage method according to claim 1, characterized in that: The steps of constructing the triage model include: Obtain a historical medical visit data training set and a historical medical visit data test set, wherein the historical medical visit data training set and the historical medical visit data test set both include several historical medical visit data sets, each historical medical visit data set is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics; The historical medical data training set is divided into multiple sub-datasets, and different machine learning algorithms are trained using each sub-dataset to obtain multiple primary triage models; Each primary triage model was tested using a test set of historical medical data, and the test results of each primary triage model were evaluated based on the model regression evaluation index. The primary triage model with the best evaluation results was selected as the final triage model.

4. The emergency pre-examination and triage method according to any one of claims 1 to 3, characterized in that: The data set of emergency patients obtained by fusing the actual condition data and historical medical treatment data of emergency patients includes: Obtain the identity information of emergency patients and establish a dataset of their actual condition and historical medical history; Based on the set feature type, corresponding feature values ​​are extracted from the actual condition data set and the historical medical treatment data set to form an actual condition feature set and a historical medical treatment feature set; Combined with the set feature fusion rules, the actual condition feature set and the historical visit feature set are fused to obtain the feature value set of emergency patients. The feature fusion rules include: a. When the feature values ​​of the same fixed field overlap, the feature value remains unchanged; b. When the feature values ​​of the same fixed field do not overlap, the feature value is mainly based on the actual condition feature set; c. Different fixed fields are retained; The eigenvalue set of emergency patients is transformed to generate an eigenvector matrix.

5. The emergency pre-examination and triage method according to claim 4, characterized in that: The steps of extracting corresponding feature values ​​from the actual condition data set and the historical medical treatment data set based on the set feature type to form the actual condition feature set and the historical medical treatment feature set are the same, wherein the corresponding feature values ​​are extracted from the actual condition data set based on the set feature type to form the actual condition feature set, including: Based on the set feature types, the fixed field matching method is used to extract feature values ​​from the actual disease data set; After the fixed field matching is completed, it is determined whether the list area corresponding to the fixed field is empty; If the response is no, the extracted feature values ​​are assigned credibility weights to form an actual disease feature set; In response to this, a keyword search method is used to search the actual condition data fragments to obtain suspected feature values ​​corresponding to the fixed fields whose list areas are empty; The suspected eigenvalues ​​are format checked. If they meet the eigenvalue format standard, they are retained and written into the list area corresponding to the fixed field. If they do not meet the eigenvalue format standard, the keyword search method is continued to be used until all actual condition data fragments are searched, and the extracted eigenvalues ​​are assigned credibility weights to form an actual condition feature set.

6. An emergency pre-examination and triage device using the method according to any one of claims 1 to 5, characterized in that: include: A feature extraction unit integrates the actual condition data and historical medical data of emergency patients to obtain a data set of emergency patients, where the data set includes identity features and condition features; The model triage unit inputs the emergency patient data set into the triage model to obtain the corresponding disease urgency and preliminary triage department. The triage model is obtained by machine learning using a number of sample data. Each sample data includes a data set of an emergency patient and identification information marking the disease urgency and triage department. The data set includes identity characteristics and disease characteristics. The triage adjustment unit combines the urgency of the disease, the preliminary triage department and the actual receiving department list at the current time, and executes the dynamic adjustment strategy based on the dynamic adjustment rules to obtain the final triage department and clinic queuing results, where the dynamic adjustment rules include: a. If the preliminary triage department is receiving patients normally, the queuing results are re-sorted according to the urgency of the disease to determine the clinic queuing results; b. If the preliminary triage department is not receiving patients normally, the patient is adjusted to the clinic with the highest clinic relevance or receiving doctor relevance, and the queuing results are re-sorted according to the urgency of the disease to determine the clinic queuing results.

7. The emergency pre-examination and triage device according to claim 6, characterized in that: The feature extraction unit includes a first extraction unit, a second extraction unit and a fusion unit. The first extraction unit extracts corresponding feature values ​​in the actual condition data set based on the set feature type to form an actual condition feature set. The second extraction unit extracts corresponding feature values ​​in the historical medical treatment data set based on the set feature type to form a historical medical treatment feature set. The two have the same structure. The first extraction unit includes: The first extraction module extracts feature values ​​from the actual disease condition data set using a fixed field matching method based on the set feature type; The judgment module, after the fixed field matching is completed, determines whether the list area corresponding to the fixed field is empty; The output module, in response to "no", assigns credibility weights to the extracted feature values ​​and forms an actual disease feature set; The second extraction module, in response to "yes", uses the keyword search method to search the actual condition data fragments, obtains the suspected feature values ​​corresponding to the fixed fields whose list areas are empty, performs format verification on the suspected feature values, and if they meet the feature value format standard, they are retained and written into the list area corresponding to the fixed field; if they do not meet the feature value format standard, the keyword search method is continued to be used for searching until all actual condition data fragments are searched, and credibility weights are assigned to the extracted feature values ​​to form an actual condition feature set.

8. The emergency pre-examination and triage device according to claim 6 or 7, characterized in that: Also includes a first model building unit and / or a second model building unit; The first model building unit includes: Obtain a historical medical visit data training set and a historical medical visit data test set, wherein the historical medical visit data training set and the historical medical visit data test set both include several historical medical visit data sets, each historical medical visit data set is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics; The historical medical data training set is divided into multiple sub-datasets, and different machine learning algorithms are trained using each sub-dataset to obtain multiple primary triage models; Each primary triage model is tested using the historical medical data test set, and the test results of each primary triage model are evaluated based on the model regression evaluation index. The primary triage model with the best evaluation result is selected as the final triage model. The second model building unit includes: Obtain a historical medical visit data training set and a historical medical visit data test set, wherein the historical medical visit data training set and the historical medical visit data test set both include several historical medical visit data sets, each historical medical visit data set is a data set of a certain emergency patient, and the data set includes identity characteristics and condition characteristics; The historical medical data training set is divided into multiple sub-datasets, and different machine learning algorithms are trained using each sub-dataset to obtain multiple primary triage models; Each primary triage model was tested using a test set of historical medical data, and the test results of each primary triage model were evaluated based on the model regression evaluation index. The primary triage model with the best evaluation results was selected as the final triage model.

9. A storage medium, characterized in that: The storage medium stores a computer program that can be read by a computer, and the computer program is configured to execute the steps of the method according to any one of claims 1 to 5 when running.

10. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the steps in the method according to any one of claims 1 to 5.

Images