Decision tree model-based hospital guide method and device

Abstract

The invention provides a hospital guide method and device based on a decision-making tree model. The hospital guide efficiency can be improved. The hospital guide method based on the decision tree model comprises the following steps: constructing a department data set based on hospital data; wherein the hospital data comprises hospital medical records and diagnosis and treatment records, and the department data set comprises standard symptoms; constructing a decision tree model based on the department data set; wherein a sample of the decision tree model is from the hospital medical record and the diagnosis and treatment record; patient information is collected; wherein the patient information comprises basic sign information and symptoms selected in the department data set; and inputting patient information into the decision tree model, and outputting a recommended doctor seeing department.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a diagnosis guidance method and device based on a decision tree model. Background Art

[0002] The complex department settings in hospitals have brought great difficulties to patients' registration. When seeking medical services, patients often need to go through multiple consultations and referrals before they can find a suitable department. Although some hospitals have deployed guidance systems, the data of these systems often do not match the actual departments, resulting in patients registering according to the system results and finding that there is actually no such department. Or traditional guidance systems are mostly based on diagnosis and treatment functions, which speculate possible diseases and provide corresponding departments. This method may make patients overly trust the results and ignore further diagnosis and treatment, thereby increasing the risk of misdiagnosis and wrong diagnosis. Summary of the invention

[0003] In order to solve the above technical problems, the present invention is proposed. The embodiments of the present invention provide a diagnosis guidance method and device based on a decision tree model, which can improve the diagnosis guidance efficiency.

[0004] According to one aspect of the present invention, a medical guidance method based on a decision tree model is provided, comprising: constructing a department data set based on hospital data; wherein the hospital data includes hospital medical records and medical treatment records, and the department data set includes standard symptoms; constructing a decision tree model based on the department data set; wherein the samples of the decision tree model are derived from the hospital medical records and the medical treatment records; collecting patient information; wherein the patient information includes basic physical sign information and symptoms selected from the department data set; inputting the patient information into the decision tree model, and outputting a recommended medical department.

[0005] In one embodiment, a department data set is constructed based on hospital data, including: preprocessing the hospital data; the preprocessing includes cleaning the noise in the medical data and processing missing values; normalizing the preprocessed medical data to extract multiple symptoms; sorting the multiple symptoms in reverse order according to the frequency of occurrence to obtain a sorted list; and selecting a preset number of symptoms from the front to the back in the sorted list as the standard symptom list.

[0006] In one embodiment, constructing a department data set based on hospital data also includes: organizing the standard symptoms and departments in the standard symptom list into a character separated value format to construct a department data set; wherein each line is a piece of data; constructing a department data set based on the collected real measurement values, evaluation values, the standard symptoms and the department; wherein the evaluation value represents that for standard symptoms for which the real measurement values ​​cannot be collected, a scoring standard is constructed to obtain an evaluation value.

[0007] In one embodiment, constructing a scoring standard to obtain an evaluation value includes: obtaining the status degree of the symptoms in the standard symptoms; when the status degree is greater than or equal to a first preset degree value, determining the score to be the highest score; when the status degree is less than the first preset degree value and greater than or equal to a second preset degree value, determining the score to be an intermediate score; when the status degree is less than the second preset degree value and greater than or equal to a third preset degree value, determining the score to be the lowest score.

[0008] In one embodiment, a decision tree model is constructed based on the department data set, including: calculating the information entropy increase of each standard symptom based on the department data set; selecting the standard symptom with the maximum gain in information entropy increase as the splitting point for separation; recursively continuing to calculate the child nodes until the stopping condition is reached to complete the construction of the decision tree model; wherein the stopping condition includes that the tree reaches the maximum preset depth or the subsets belong to the same treatment department.

[0009] In one embodiment, based on the department data set, the information entropy of each standard symptom is calculated, including: calculating the information entropy based on the department data set; calculating the conditional entropy of the data set with the standard symptoms as features; calculating the difference between the information entropy and the conditional entropy as the information entropy; wherein, the greater the information entropy, the more information the corresponding standard symptom can provide when dividing the data set.

[0010] In one embodiment, calculating information entropy based on the department data set includes: calculating information entropy based on the department data set by formula 1: In formula 1, H(D) represents the information entropy of the department data set, p i It represents the probability of samples belonging to the i-th category in the department data set, and n represents the total number of all possible values ​​of D.

[0011] In one embodiment, the conditional entropy of a data set is calculated using the standard symptoms as features, including: using the standard symptoms as features, calculating the conditional entropy of the data set by formula 2: In formula 2, H(D|A) represents the conditional entropy of feature A, V(A) represents all the values ​​of feature A, and D v It represents the subset whose feature A value is v, and D represents the department data set.

[0012] In one embodiment, the basic vital sign information includes at least one of age, gender, body temperature and blood pressure; wherein, collecting patient information includes: collecting at least one of age, gender, body temperature and blood pressure input by the patient; and collecting the standard symptoms selected by the patient according to his or her own condition; wherein, inputting the patient information into the decision tree model and outputting the recommended medical department includes: taking at least one of age, gender, body temperature and blood pressure input by the patient and the standard symptoms selected by the patient according to his or her own condition as inputs to the decision tree model, and obtaining the recommended medical department output by the decision tree model.

[0013] According to another aspect of the present invention, a medical guidance device based on a decision tree model is provided, comprising: a first construction module, used to construct a department data set based on hospital data; wherein the hospital data includes hospital medical records and medical treatment records, and the department data set includes standard symptoms; a second construction module, used to construct a decision tree model based on the department data set; wherein the samples of the decision tree model are derived from the hospital medical records and the medical treatment records; a collection module, used to collect patient information; wherein the patient information includes basic vital sign information and symptoms selected from the department data set; an output module, used to input patient information into the decision tree model and output a recommended department for medical treatment.

[0014] The medical guidance method and device based on the decision tree model provided by the present invention combine the hospital's historical data and patient information, use the decision tree model to perform comprehensive analysis, and recommend the most suitable medical department. Without predicting any disease, the appropriate medical department is recommended according to symptoms, and the doctor gives a disease diagnosis, which can reduce the risk of misdiagnosis and wrong diagnosis and improve the efficiency of medical guidance. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other purposes, features and advantages of the present invention will become more apparent by describing the embodiments of the present invention in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a flowchart of a diagnosis guidance method based on a decision tree model provided by an exemplary embodiment of the present invention.

[0017] Figure 2 It is a structural schematic diagram of a diagnosis guidance device based on a decision tree model provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0018] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0019] Figure 1 is a flow chart of a diagnosis guidance method based on a decision tree model provided by an exemplary embodiment of the present invention, such as Figure 1 As shown in FIG. 1 , the diagnosis guidance method based on the decision tree model includes:

[0020] S100: Build department data sets based on hospital data.

[0021] Among them, hospital data include hospital medical records and treatment records, and department data sets include standard symptoms.

[0022] Obtain historical data from the hospital, such as hospital medical records and medical treatment records. Hospital medical records and medical treatment records are detailed records of the patient's condition, diagnosis, treatment and rehabilitation process by doctors. They contain rich medical information and symptom descriptions. By carefully analyzing these records, various symptoms of the patient can be extracted. According to the department where the symptoms come from, a standard symptom-department dataset can be established as a sample for model construction. In addition, since hospital data comes from the hospital where the patient is treated,

[0023] S200: Build a decision tree model based on department data set.

[0024] Among them, the samples of the decision tree model come from hospital medical records and treatment records.

[0025] Decision Tree is a machine learning model for classification and regression, which helps decision-making or prediction by building a tree structure. It divides data into multiple subsets based on feature tests until the prediction result is finally obtained. The main features of decision tree are that it is easy to understand and explain, can handle various data types, does not require massive data sets to train decision tree models, saves a lot of manpower and material costs, and recommends medical departments based on decision tree models. The recommendation results are highly interpretable. Decision tree does not predict any disease, but directly recommends medical departments, reducing the risk of misdiagnosis and wrong diagnosis.

[0026] S300: Collect patient information.

[0027] Among them, patient information includes basic vital signs information and symptoms selected from the department data set.

[0028] The basic vital signs information in the patient information may be information such as age, gender, body temperature and blood pressure, which are actively input by the patient. Body temperature and blood pressure are real-time monitored data. The symptoms selected in the department data set indicate that the patient makes corresponding choices from the standard symptoms based on his or her own symptoms. For example, if the patient has symptoms such as coughing and chest tightness, the corresponding symptoms are selected from the standard symptoms.

[0029] S400: Input the patient information into the decision tree model and output the recommended treatment department.

[0030] The patient's basic physical sign information and selected standard symptoms are input into the decision tree model, and finally a recommended treatment department is obtained, which solves the problem of mismatch between the recommended department of the guidance system and the actual department of the hospital. Through this system, patients can get reliable treatment department recommendations in a short time, improving the medical experience. Hospitals can also allocate and utilize medical resources more efficiently, laying the foundation for the realization of intelligent medical services.

[0031] In one embodiment, S100 may include: preprocessing hospital data; the preprocessing includes cleaning noise in the medical data and processing missing values; normalizing the preprocessed medical data to extract multiple symptoms; sorting the multiple symptoms in reverse order according to the frequency of occurrence to obtain a sorted list; and selecting a preset number of symptoms from the front to the back in the sorted list as a standard symptom list.

[0032] Preprocessing hospital medical records or treatment records aims to extract valuable information from large, complex and unstructured text data for subsequent machine learning. First, it is necessary to collect data from multiple sources such as the hospital's electronic medical record system (EMR), paper medical record transcription, laboratory test results, imaging reports, etc. The cleaning steps can include removing duplicate records, processing missing values ​​(such as through interpolation, mean filling or marking as missing), correcting erroneous data (such as spelling errors, inconsistent formats, etc.), and unifying data formats (such as date format, unit conversion, etc.). Since medical records and treatment records are usually in text form, natural language processing can be used to parse the text and identify key information such as disease names, drug names, and test results. After normalizing the preprocessed medical data, extract each symptom. Arrange the symptoms in reverse order of frequency, select the first preset number (for example, 150) of symptoms as the standard symptom list, and discard the remaining symptoms and do not include them in the standard symptom list.

[0033] In one embodiment, S100 may also include: organizing the standard symptoms and departments in the standard symptom list into a character separated value format to construct a department data set; wherein each row is a piece of data; constructing a department data set based on the collected real measurement values, evaluation values, standard symptoms and departments; wherein the evaluation value represents that for standard symptoms for which real measurement values ​​cannot be collected, a scoring standard is constructed to obtain an evaluation value.

[0034] The relationship between symptoms and the department visited is organized into a character-separated value (CSV) format. The CSV file consists of any number of records separated by some kind of line break character; each record consists of fields, and the separators between fields are other characters or strings, the most common being commas or tabs, with one piece of data per line. Among them, the value of the corresponding data under the standard state can be a real measurement value or an evaluation value. The standard of the evaluation value can be freely formulated, but the score must be strictly in accordance with the standard during collection. For example, body temperature is the value obtained by real measurement, and the severity of cough is the evaluation value: 0 indicates that there is no cough symptom, 1 indicates that there is a mild cough symptom, 2 indicates that there is an intermittent severe cough symptom, and 3 indicates that there is a frequent and severe cough symptom.

[0035] In one embodiment, constructing a scoring standard to obtain an evaluation value may include: obtaining the status degree of the symptoms in the standard symptoms; when the status degree is greater than or equal to a first preset degree value, determining the score to be the highest score; when the status degree is less than the first preset degree value and greater than or equal to a second preset degree value, determining the score to be an intermediate score; when the status degree is less than the second preset degree value and greater than or equal to a third preset degree value, determining the score to be the lowest score.

[0036] Scoring criteria can be freely formulated according to actual conditions. For example, if the number of diarrhea episodes is greater than 5 within 30 minutes, the score is 2; if the number of diarrhea episodes is less than 5 and greater than 3 within 30 minutes, the score is 1; if the number of diarrhea episodes is 0 within 30 minutes, the score is 0. After constructing the evaluation value, when the patient selects diarrhea as a symptom, the score is filled in according to the patient's own diarrhea situation as input for model judgment. By constructing scoring criteria, patients can not only judge their own symptoms simply and efficiently, but also provide a data basis for the subsequent decision tree model, which can more effectively handle multi-category classification problems and efficiently classify in multidimensional data.

[0037] In one embodiment, S200 may include: calculating the information entropy increase of each standard symptom based on the department data set; selecting the standard symptom with the largest gain in the information entropy increase as the splitting point for separation; recursively continuing to calculate the child nodes until the stopping condition is reached to complete the construction of the decision tree model; wherein the stopping condition includes that the tree reaches the maximum preset depth or the subsets belong to the same treatment department.

[0038] There are many selection strategies for decision trees, the most common ones are: Information Gain and Gini Impurity. Information gain is based on entropy in information theory. It measures the uncertainty of the sample set. At each node, the decision tree algorithm calculates the information gain of each possible split point and selects the information gain that maximizes the overall tree structure as the optimal split point. It is suitable for processing numerical features because the information gain is calculated directly based on the value of the feature; Gini impurity is an indicator to measure the impurity or confusion of the sample in the node. At each node, the decision tree algorithm calculates the Gini impurity of each possible split point and selects the split point that can minimize the impurity as the optimal split point. When the number of samples in certain categories in the data set is small or unbalanced, Gini impurity can better handle this situation because it does not involve assumptions about probability distribution. Therefore, this application can be based on the C4.5 decision tree algorithm (C4.5 Decision Tree Algorithm), which uses information entropy increase (InformationGain) as the splitting criterion. Compared with Gini Impurity (Gini Impurity), information entropy increase has the following significant advantages: ① It can handle multi-category classification problems more effectively; ② It performs well in high-dimensional data; ③ It avoids bias towards multi-valued features through the information gain ratio (GainRatio). Based on these advantages, in the scenario of inferring the department of treatment based on symptoms, selecting information gain as the splitting criterion of the decision tree model can provide better performance. When constructing a tree model, the maximum gain of all standard symptoms is obtained by calculating the information entropy increase, and the feature with the largest information gain (such as body temperature) is selected for the first separation, and then the subnodes are recursively calculated (for example, the feature of the second gain is selected for subnode separation) until the stopping condition is reached, that is, the preset maximum depth of the tree is reached or the subsets all belong to the same department of treatment. For example, if the maximum depth of the tree is set to 10, a maximum of 10 symptoms are required to give a recommended department.

[0039] Each standard symptom can calculate multiple information entropies. The maximum value among the multiple information entropies is the maximum gain of the symptom. Information gain is used to measure the degree of improvement of the purity of a feature to the data set, and can also be understood as the importance of the feature. The calculation formula of information entropy is: information entropy = information entropy - conditional entropy, where conditional entropy represents the uncertainty of the data set when the value of a certain feature is known. For a feature, assume that it has m values, corresponding to m sub-datasets. Calculate the information entropy of these m sub-datasets respectively, and sum them according to the weighted sample proportion of each sub-dataset to obtain the conditional entropy. Assuming there is a feature, the data set can be divided into two sub-datasets according to the value of the feature, containing 40 samples and 60 samples respectively. 60 of them belong to category A and 40 belong to category B. Then the sample proportion of category A is 60 / 100=0.6, and the sample proportion of category B is 40 / 100=0.4. According to the calculation formula of information entropy, the information entropy of this data set is:

[0040] Information entropy = -(0.6×log20.6+0.4×log20.4) = 0.971 (the result is rounded to three decimal places);

[0041] One of the sub-datasets contains 40 samples, 30 of which belong to category A and 10 belong to category B; the other sub-dataset contains 60 samples, 30 of which belong to category A and 30 belong to category B. Then the information entropy of these two sub-datasets is:

[0042] (The result is rounded to three decimal places)

[0043] (Because the number of samples in the two categories is equal, the entropy reaches the maximum);

[0044] According to the weighted summation of the sample proportions of each sub-dataset, the conditional entropy is obtained as follows:

[0045]

[0046] Finally, according to the difference between information entropy and conditional entropy, the information entropy of the feature is calculated as:

[0047] Information entropy increase = information entropy - conditional entropy = 0.971-0.906 = 0.065.

[0048] In practical applications, the classification model can be constructed by calculating the information entropy of each feature and selecting the feature with larger information gain as the split node of the decision tree.

[0049] In one embodiment, based on the department data set, calculating the information entropy increase of each standard symptom may include: calculating the information entropy based on the department data set; calculating the conditional entropy of the data set with the standard symptoms as features; calculating the difference between the information entropy and the conditional entropy as the information entropy increase; wherein, the greater the information entropy increase, the more information the corresponding standard symptom can provide when dividing the data set.

[0050] As an example of calculating the information entropy of each standard symptom, based on the department data set, the information entropy can be calculated by formula 1:

[0051]

[0052] In formula 1, H(D) represents the information entropy of the department data set, p i It represents the probability of samples belonging to the i-th category in the department data set, and n represents the total number of all possible values ​​of D.

[0053] As an example of calculating the information entropy of each standard symptom, taking the standard symptom as a feature, the conditional entropy of the data set can be calculated by Formula 2:

[0054]

[0055] In formula 2, H(D|A) represents the conditional entropy of feature A, V(A) represents all the values ​​of feature A, and D v It indicates that feature A takes the value of v as a subset, and D represents the department data set. Feature A can be any standard symptom.

[0056] Information entropy is the difference between the information entropy and conditional entropy of a data set. The larger the information entropy, the more information feature A can provide when dividing the data set, making the splitting of the decision tree more effective. The information entropy of feature A can be calculated by formula three:

[0057] Information Gain(D, A)=H(D)-H(D|A) Formula 3;

[0058] There are three formulas, H(D) represents the information entropy of the department data set, and H(D|A) represents the conditional entropy of feature A.

[0059] Taking body temperature as an example, the information entropy of body temperature can be calculated as follows:

[0060] Calculate the total entropy, and the probability of each department is as follows:

[0061]

[0062] Substituting into formula 1, we get:

[0063]

[0064] The goal is to select a value that maximizes the increase in entropy. In the actual process, the optimal solution will be obtained by iteration many times. Here, an initial value of 37.5 is selected as an example. When the subset with body temperature ≤ 37.5 has three samples, all of whom are treated in the internal medicine department, their entropy is 0; the subset with body temperature > 37.5 has two samples, who are treated in the respiratory department and the cardiovascular department respectively, and their entropy is:

[0065]

[0066] Calculate the conditional entropy:

[0067]

[0068] Calculate information entropy:

[0069] Information Gain=0.971-0.4=0.571;

[0070] That is, the information entropy increase of body temperature is 0.571. Then the information entropy increase of other standard symptoms is calculated. The standard symptom corresponding to the maximum information entropy increase is selected from all standard symptoms as the split point for the first separation. The other symptoms are recursively calculated on the child nodes in turn until the stopping condition is reached.

[0071] In one embodiment, basic vital sign information includes at least one of age, gender, body temperature and blood pressure; wherein, S300 may include: collecting at least one of age, gender, body temperature and blood pressure input by the patient; and collecting standard symptoms selected by the patient according to his or her own condition; wherein, S400 may include: using at least one of age, gender, body temperature and blood pressure input by the patient and the standard symptoms selected by the patient according to his or her own condition as inputs to the decision tree model, and obtaining the recommended medical department output by the decision tree model.

[0072] If the maximum depth of the decision tree model is 10, basic vital signs information such as age, gender, body temperature, blood pressure, etc. can be set and actively input by the patient. The number of basic vital signs should be at least 7, and other symptoms should be selected from the standard symptom list, with at least 3 existing symptoms selected.

[0073] For example, the first split feature is body temperature:

[0074] Left node (body temperature ≤ 37.55℃): All samples belong to internal medicine and no further separation is required.

[0075] Right node (body temperature>37.55℃): contains samples from respiratory medicine and cardiovascular medicine.

[0076] Next, the right node is further divided into: cough severity

[0077] Left node (cough severity ≤ 1.5): no sample, no further separation required.

[0078] Right node (cough severity > 1.5): All samples belong to different departments and need to be further separated.

[0079] Continue segmenting until all samples are from the same department or the maximum depth of the tree is reached.

[0080] The final decision tree may not contain all symptoms. In the example given, according to the decision tree, if the patient's temperature is 37.5≤37.55, the patient can be directly advised to see an internal medicine doctor.

[0081] Figure 2 FIG. 1 is a schematic diagram of a medical guidance device based on a decision tree model provided by an exemplary embodiment of the present invention. Figure 2 As shown, the guidance device 2 based on the decision tree model includes: a first construction module 21, which is used to construct a department data set based on hospital data; wherein the hospital data includes hospital medical records and medical records, and the department data set includes standard symptoms; a second construction module 22, which is used to construct a decision tree model based on the department data set; wherein the samples of the decision tree model come from hospital medical records and medical records; an acquisition module 23, which is used to collect patient information; wherein the patient information includes basic vital signs information and symptoms selected from the department data set; an output module 24, which is used to input patient information into the decision tree model and output a recommended department for consultation.

[0082] In one embodiment, the first construction module 21 can be configured to: preprocess hospital data; the preprocessing includes cleaning noise in the medical data and processing missing values; normalizing the preprocessed medical data to extract multiple symptoms; sorting the multiple symptoms in reverse order according to the frequency of occurrence to obtain a sorted list; and selecting a preset number of symptoms from the front to the back in the sorted list as a standard symptom list.

[0083] In one embodiment, the first construction module 21 can also be configured to: organize the standard symptoms and departments in the standard symptom list into a character separated value format to construct a department data set; wherein each line is a piece of data; construct a department data set based on the collected real measurement values, evaluation values, standard symptoms and departments; wherein the evaluation value represents that for standard symptoms for which real measurement values ​​cannot be collected, a scoring standard is constructed to obtain an evaluation value.

[0084] In one embodiment, constructing a scoring standard to obtain an evaluation value includes: obtaining the status degree of the symptoms in the standard symptoms; when the status degree is greater than or equal to a first preset degree value, determining the score to be the highest score; when the status degree is less than the first preset degree value and greater than or equal to a second preset degree value, determining the score to be an intermediate score; when the status degree is less than the second preset degree value and greater than or equal to a third preset degree value, determining the score to be the lowest score.

[0085] In one embodiment, the second construction module 22 can be configured as follows: based on the department data set, calculate the information entropy increase of each standard symptom; select the standard symptom with the largest gain in information entropy increase as the splitting point for separation; recursively continue to calculate the child nodes until the stopping condition is reached to complete the construction of the decision tree model; wherein the stopping condition includes that the tree reaches the maximum preset depth or the subsets belong to the same treatment department.

[0086] In one embodiment, the second construction module 22 can be configured as follows: calculating information entropy based on the department data set; calculating the conditional entropy of the data set using standard symptoms as features; calculating the difference between the information entropy and the conditional entropy as information entropy increase; wherein, the greater the information entropy increase, the more information the corresponding standard symptoms can provide when dividing the data set.

[0087] In one embodiment, the second construction module 22 may be configured to calculate information entropy based on the department data set by formula 1: In formula 1, H(D) represents the information entropy of the department data set, p i It represents the probability of samples belonging to the i-th category in the department data set, and n represents the total number of all possible values ​​of D.

[0088] In one embodiment, the second construction module 22 may be configured to: use the standard symptoms as features and calculate the conditional entropy of the data set by using Formula 2: In formula 2, H(D|A) represents the conditional entropy of feature A, V(A) represents all the values ​​of feature A, and D v It represents the subset whose feature A value is v, and D represents the department data set.

[0089] In one embodiment, the basic vital sign information includes at least one of age, gender, body temperature and blood pressure; wherein, the collection module 23 can be configured to: collect at least one of the age, gender, body temperature and blood pressure input by the patient; and collect standard symptoms selected by the patient according to his or her own condition; wherein, the output module 24 can be configured to: use at least one of the age, gender, body temperature and blood pressure input by the patient and the standard symptoms selected by the patient according to his or her own condition as inputs to the decision tree model, and obtain the recommended medical department output by the decision tree model.

[0090] An embodiment of the present invention provides a guidance device based on a decision tree model. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. From a hardware perspective, in addition to the CPU, memory, network interface, and non-volatile memory, the device in the embodiment where the device is located can generally include other hardware, such as a forwarding chip responsible for processing messages, etc. Taking software implementation as an example, as a device in a logical sense, it is formed by the CPU of the device where it is located reading the corresponding computer program instructions in the non-volatile memory into the memory for execution.

[0091] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and the computer program is used to execute the diagnosis guidance method based on the decision tree model of any of the above embodiments.

[0092] In addition to the above-mentioned methods and devices, an embodiment of the present invention may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the decision tree model-based diagnosis guidance method according to various embodiments of the present invention described in the above-mentioned "Exemplary Method" section of this specification.

[0093] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: a processor; a memory for storing instructions executable by the processor; and a processor for executing the diagnosis guidance method based on the decision tree model of any of the above embodiments.

[0094] In addition, an embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enables the processor to execute the steps of the diagnosis guidance method based on a decision tree model according to various embodiments of the present invention described in the above “Exemplary Method” section of this specification.

[0095] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A diagnosis guidance method based on a decision tree model, characterized in that: include: Based on the hospital data, a department data set is constructed; wherein the hospital data includes hospital medical records and diagnosis and treatment records, and the department data set includes standard symptoms; Based on the department data set, a decision tree model is constructed; wherein the samples of the decision tree model are derived from the hospital medical records and the diagnosis and treatment records; Collecting patient information; wherein the patient information includes basic physical sign information and symptoms selected from the department data set; The patient information is input into the decision tree model, and the recommended medical department is output.

2. The diagnosis guidance method based on the decision tree model according to claim 1, characterized in that: Based on hospital data, a department data set is constructed, including: Preprocessing the hospital data; the preprocessing includes cleaning noise in the medical data and processing missing values; Normalizing the preprocessed medical data to extract multiple symptoms; Sort multiple symptoms in descending order of frequency to obtain a sorted list; In the sorted list, a preset number of symptoms are selected from front to back as the standard symptom list.

3. The diagnosis guidance method based on the decision tree model according to claim 2, characterized in that: Based on hospital data, we build department data sets, which also include: Arrange the standard symptoms and departments in the standard symptom list into a character separated value format to construct a department data set, wherein each line contains one piece of data; A department data set is constructed based on the collected real measurement values, evaluation values, the standard symptoms and the department; wherein the evaluation value represents a scoring standard constructed to obtain the evaluation value for the standard symptoms for which the real measurement values ​​cannot be collected.

4. The diagnosis guidance method based on the decision tree model according to claim 3 is characterized in that: Constructing a scoring rubric to obtain the assessment value includes: Obtain the status degree of the symptoms in the standard symptoms; When the state degree is greater than or equal to a first preset degree value, determining the score to be the highest score; When the state degree is less than the first preset degree value and greater than or equal to the second preset degree value, determining the score to be an intermediate score; When the status degree is less than the second preset degree value and greater than or equal to the third preset degree value, the score is determined to be the lowest score.

5. The diagnosis guidance method based on the decision tree model according to claim 1, characterized in that: Based on the department data set, a decision tree model is constructed, including: Based on the department data set, calculating the information entropy increase of each standard symptom; Select the standard symptom with the maximum gain in information entropy increase as the split point for separation; The sub-nodes are recursively calculated until the stopping condition is reached to complete the construction of the decision tree model; wherein the stopping condition includes that the tree reaches a maximum preset depth or the subsets belong to the same medical department.

6. The diagnosis guidance method based on the decision tree model according to claim 5, characterized in that: Based on the department data set, the information entropy of each standard symptom is calculated, including: Based on the department data set, calculating information entropy; Using standard symptoms as features, the conditional entropy of the data set is calculated; The difference between the information entropy and the conditional entropy is calculated as the information entropy increase; wherein, the greater the information entropy increase, the more information the corresponding standard symptom can provide when dividing the data set.

7. The diagnosis guidance method based on the decision tree model according to claim 6, characterized in that: Based on the department data set, information entropy is calculated, including: Based on the department data set, the information entropy is calculated by formula 1: In formula 1, H(D) represents the information entropy of the department data set, p i It represents the probability of samples belonging to the i-th category in the department data set, and n represents the total number of all possible values ​​of D.

8. The diagnosis guidance method based on the decision tree model according to claim 1, characterized in that: The conditional entropy of the dataset is calculated using standard symptoms as features, including: Using standard symptoms as features, the conditional entropy of the data set is calculated using Formula 2: In formula 2, H(D|A) represents the conditional entropy of feature A, V(A) represents all the values ​​of feature A, and D v It represents the subset whose feature A value is v, and D represents the department data set.

9. The diagnosis guidance method based on the decision tree model according to claim 1, characterized in that: The basic physical sign information includes at least one of age, gender, body temperature and blood pressure; The collection of patient information includes: Collect at least one of the patient's age, gender, temperature and blood pressure input; and Collecting the standard symptoms selected by the patient according to his / her own condition; The patient information is input into the decision tree model, and the recommended treatment department is output, including: At least one of the age, gender, body temperature and blood pressure input by the patient and the standard symptoms selected by the patient according to his or her own condition are used as inputs of the decision tree model to obtain the recommended medical department output by the decision tree model.

10. A medical guidance device based on a decision tree model, characterized in that: include: The first construction module is used to construct a department data set based on hospital data; wherein the hospital data includes hospital medical records and diagnosis and treatment records, and the department data set includes standard symptoms; A second construction module is used to construct a decision tree model based on the department data set; wherein the samples of the decision tree model are derived from the hospital medical records and the diagnosis and treatment records; A collection module, used for collecting patient information; wherein the patient information includes basic physical sign information and symptoms selected from the department data set; The output module is used to input patient information into the decision tree model and output the recommended treatment department.

Images