CN113724854A - Machine learning-based grading triage method and system and computer equipment - Google Patents
Machine learning-based grading triage method and system and computer equipment Download PDFInfo
- Publication number
- CN113724854A CN113724854A CN202110860445.7A CN202110860445A CN113724854A CN 113724854 A CN113724854 A CN 113724854A CN 202110860445 A CN202110860445 A CN 202110860445A CN 113724854 A CN113724854 A CN 113724854A
- Authority
- CN
- China
- Prior art keywords
- target patient
- grade
- target
- disease
- triage
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/247—Thesauruses; Synonyms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/20—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Physics & Mathematics (AREA)
- Epidemiology (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- General Business, Economics & Management (AREA)
- Business, Economics & Management (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention relates to a machine learning-based grading triage method and a machine learning-based grading triage system, wherein the method comprises the steps of receiving a questionnaire filled by a target patient to be graded and triaged, and summarizing answers of the questionnaire; extracting disease characteristics of the target patient from answers of the questionnaire, wherein the disease characteristics comprise symptoms, onset time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; preprocessing the disease characteristics of a target patient according to a preset characteristic preprocessing strategy to obtain target disease characteristics expressed in a digital form, wherein the target disease characteristics comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; the method and the system integrate various physical factors of the patient and invisible problems of diseases, and accurately divide the disease level of the patient based on machine learning.
Description
Technical Field
The invention relates to the field of grading triage, in particular to a grading triage method and system based on machine learning and computer equipment.
Background
At present, when a patient goes to a hospital for a doctor, a nurse in the hospital guides a doctor according to the symptoms of the patient, preliminarily determines a department where the patient should see a doctor and judges the level of the patient to allocate hospital resources. However, the efficiency is low and the human resources are insufficient when the classification triage is carried out manually, and the problem is not solved, so that an intelligent emergency classification triage system is provided in the prior art, the classification diagnosis is carried out on emergency patients, and the registration information of each emergency patient is stored in a server; the first client acquires corresponding patient registration information from the server according to the disease condition data of the emergency patients, associates the disease condition data with the corresponding patient registration information to form associated data, and distributes grades according to the associated data corresponding to the disease condition data and then transmits the grades back to the server. However, the system only uses common customer service end and service end butt joint, and only adopts a threshold form to carry out grading triage, and although a part of problems in human resources can be solved, the grading mode is too mechanized, and various physical factors of patients and recessive problems of diseases, such as the age, the sex, the disease onset time and the like of the patients are not comprehensively considered, so that the grading triage grading is unscientific and has a limitation problem.
Disclosure of Invention
In order to solve the problems, the invention provides a machine learning-based grading triage method, which comprises the following steps: receiving a questionnaire filled by a target patient to be classified and triaged, and summarizing answers of the questionnaire; extracting disease characteristics of the target patient from answers of the questionnaire, wherein the disease characteristics comprise symptoms, onset time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; preprocessing the disease characteristics of a target patient according to a preset characteristic preprocessing strategy to obtain target disease characteristics expressed in a digital form, wherein the target disease characteristics comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and inputting the target disease characteristics into a preset grading triage model to obtain grading information of the target patient, wherein the grading triage model is a machine learning model constructed by using an XGB model.
Further, the grading classification level has multiple levels, and the grading information of the target patient includes the probability of the target patient at each level and the final classification level of the target patient, wherein the final classification level of the target patient is the highest probability level.
Further, the method further comprises the steps of obtaining the number of the patients waiting for diagnosis with the same diagnosis grade as the target patient at the current time according to the grading information of the target patient, and determining the waiting time of the target patient according to the number of the patients waiting for diagnosis with the same diagnosis grade as the target patient and the preset corresponding relation between the number of the patients waiting for diagnosis and the waiting time.
Further, during waiting period of the target patient, corresponding symptom knowledge and disease prevention knowledge are pushed to the target patient according to the disease characteristics of the target patient.
Further, preprocessing the disease characteristics of the target patient according to a preset characteristic preprocessing strategy to obtain the target disease characteristics represented in a digital form comprises: replacing the symptoms of the target patient according to a preset synonym replacement comparison table, and converting the symptoms of the target patient into Arabic numerals according to a preset symptom conversion strategy; using a regular expression to exclude full-angle punctuations and half-angle punctuations in the disease onset duration of the target patient, carrying out synonym replacement on the disease onset duration of the target patient, converting the synonym replacement into Arabic numerals, and converting the unit of the disease onset duration of the target patient into minutes; irregular punctuation marks and Chinese in the body temperature of the target patient are excluded by utilizing a regular expression, and the body temperature which is not in a preset normal body temperature range is replaced by 37 degrees; subtracting the diastolic pressure from the systolic pressure of the target patient to obtain the pulse pressure difference of the target patient; converting the sex of the target patient into Arabic numerals according to the preset corresponding relation between the sex and the numerals; obtaining the heart rate grade of the target patient according to the preset corresponding relation between the heart rate grade and the heart rate of the target patient; acquiring the systolic pressure grade of the target patient according to the preset corresponding relation between the systolic pressure grade and the systolic pressure of the target patient; determining a risk level of the target patient according to a preset risk level determination strategy, the heart rate level and the systolic pressure level of the target patient; and analyzing the symptoms of the target patient according to a preset severe grade determination strategy, and determining the severe grade of the target patient.
Further, the step of constructing the hierarchical triage model by using the XGBOOST model comprises: collecting case data of a hospital as sample cases, wherein the number of the sample cases is multiple; extracting disease characteristics from each sample case according to a preset characteristic extraction strategy, wherein the disease characteristics of the sample case comprise symptoms, disease starting time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; marking grading levels for each sample case in advance according to the disease characteristics of each sample case; preprocessing the disease characteristics of the sample case according to a characteristic preprocessing strategy to obtain target disease characteristics of the sample case in a digital form, wherein the target disease characteristics of the sample case comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and inputting the target disease state characteristics of the sample case into the XGBOOST model as input layer training samples by using the XGBOOST model, and training and learning by using the grading grade of the sample case as output layer training samples to obtain a grading triage model.
The invention also provides a grading triage system based on machine learning, which comprises a receiving module, a feature extraction module, a feature preprocessing module and a grading triage module, wherein: the receiving module is connected with the characteristic extraction module and used for receiving the questionnaire filled by the target patient to be classified and triaged and summarizing answers of the questionnaire; the characteristic extraction module is connected with the receiving module and the characteristic preprocessing module and is used for extracting the disease characteristics of the target patient from the answers of the questionnaire, wherein the disease characteristics comprise symptoms, onset time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; the characteristic preprocessing module is connected with the characteristic extraction module and the grading diagnosis module, and is used for preprocessing the disease characteristics of the target patient according to a preset characteristic preprocessing strategy to obtain target disease characteristics expressed in a digital form, wherein the target disease characteristics comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and the grading triage module is connected with the characteristic preprocessing module and is used for receiving the target disease symptoms and inputting the target disease symptoms into a preset grading triage model to obtain grading information of the target patient.
Further, the characteristic preprocessing module comprises a receiving unit, a symptom preprocessing unit, an onset time preprocessing unit, a body temperature preprocessing unit, a pulse pressure difference determining unit, a gender preprocessing unit, a heart rate grade determining unit, a systolic pressure grade determining unit, a danger grade determining unit and an severe grade determining unit, wherein: the receiving unit is connected with the symptom preprocessing unit, the onset time preprocessing unit, the body temperature preprocessing unit, the pulse pressure difference determining unit, the sex preprocessing unit, the heart rate grade determining unit, the systolic pressure grade determining unit, the danger grade determining unit and the severe grade determining unit and is used for receiving the disease characteristics of the target patient; the symptom preprocessing unit is used for replacing the symptoms of the target patient according to a preset synonym replacement comparison table and converting the symptoms of the target patient into Arabic numerals according to a preset symptom conversion strategy; the onset duration preprocessing unit is used for eliminating full-angle punctuations and half-angle punctuations in the onset duration of the target patient by using a regular expression, carrying out synonym replacement on the onset duration of the target patient, converting the synonym replacement into Arabic numerals, and converting the unit of the onset duration of the target patient into minutes; the body temperature preprocessing unit is used for removing irregular punctuation marks and Chinese from the body temperature of a target patient by using a regular expression, and replacing the body temperature which is not in a preset normal body temperature range with 37 degrees; the pulse pressure difference determining unit is used for subtracting the diastolic pressure from the systolic pressure of the target patient to obtain the pulse pressure difference of the target patient; the gender preprocessing unit is used for converting the gender of the target patient into Arabic numerals according to the preset correspondence between the gender and the numerals; the heart rate grade determining unit is used for obtaining the heart rate grade of the target patient according to the preset corresponding relation between the heart rate grade and the heart rate of the target patient; the systolic pressure grade determining unit is used for obtaining the systolic pressure grade of the target patient according to the preset corresponding relation between the systolic pressure grade and the systolic pressure of the target patient; the risk level determination unit is used for determining the risk level of the target patient according to a preset risk level determination strategy, the heart rate level and the systolic pressure level of the target patient; and the severe grade determining unit is used for analyzing the symptoms of the target patient to be treated according to a preset severe grade determining strategy and determining the severe grade of the target patient to be treated.
The system further comprises a grading triage model construction module, wherein the grading triage model construction module is connected with the grading triage module and is used for acquiring the case data of the hospital as a plurality of sample cases; extracting disease characteristics from each sample case according to a preset characteristic extraction strategy, wherein the disease characteristics of the sample case comprise symptoms, disease starting time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; marking grading levels for each sample case in advance according to the disease characteristics of each sample case; preprocessing the disease characteristics of the sample case according to a characteristic preprocessing strategy to obtain target disease characteristics of the sample case in a digital form, wherein the target disease characteristics of the sample case comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and inputting the target disease state characteristics of the sample case into the XGBOOST model as input layer training samples by using the XGBOOST model, and training and learning by using the grading grade of the sample case as output layer training samples to obtain a grading triage model.
The invention also provides a computer device, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the steps of the grading triage method.
The grading triage method, the grading triage system and the computer equipment based on the machine learning provided by the invention at least have the following beneficial effects: the information of the target patient is obtained through questionnaire, the disease characteristics of the target patient are extracted, and the characteristics of the sex, the age, the symptoms, the onset time, the body temperature, the pulse pressure difference, the heart rate level, the systolic pressure level, the risk level and the severe level of the target patient are input into a hierarchical diagnosis division model constructed by using an XGBOST model, so that accurate classification is realized. According to the grading triage method, the grading triage system and the computer equipment, when patients are graded and triaged, various physical factors and recessive problems of diseases of the patients, such as the ages, the sexes, the attack times and the like of the patients are comprehensively considered, scientific grading is achieved, and grading grades with higher reference values are provided for doctors.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of a machine learning-based hierarchical triage method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a machine learning-based hierarchical triage method according to another embodiment of the present invention;
FIG. 3 is a flow diagram of a feature preprocessing method in one embodiment of the invention;
FIG. 4 is a flow chart of a method for constructing a hierarchical triage machine learning model in an embodiment of the present invention;
FIG. 5 is a schematic diagram of a hierarchical triage system based on machine learning according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a feature preprocessing module according to an embodiment of the present invention;
201-receiving module, 202-feature extraction module, 203-feature preprocessing module, 204-grading diagnosis module, 205-grading diagnosis model construction module, 2031-receiving unit, 2032-symptom preprocessing unit, 2033-disease onset duration preprocessing unit, 2034-body temperature preprocessing unit, 2035-pulse pressure difference determining unit, 2036-gender preprocessing unit, 2037-heart rate grade determining unit, 2038-systolic pressure grade determining unit, 2039-risk grade determining unit and 20310-severe grade determining unit.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In an embodiment of the present invention, a machine learning-based hierarchical triage method is provided, as shown in fig. 1, the method comprising the steps of:
step S101: and receiving a questionnaire filled by the target patient to be graded and classified, and summarizing answers of the questionnaire.
Specifically, in the present scenario, the questions in the questionnaire are preset by the professional.
Step S102: the disease characteristics of the target patient including symptoms, length of onset, body temperature, heart rate, systolic blood pressure, diastolic blood pressure, sex, and age are extracted from the answers to the questionnaire.
Specifically, in the questionnaire, the patient inputs his or her own complaint, current body temperature, heart rate, systolic pressure, diastolic pressure, sex, and age into the questionnaire, wherein the current body temperature, heart rate, systolic pressure, and diastolic pressure may be obtained by a previous basic examination after the patient arrives at a hospital, and the complaint is automatically filled by the patient according to his or her own symptoms.
In one embodiment, when extracting disease symptoms of a target patient from answers of a questionnaire, the disease chief complaints of the target patient are extracted by using a regular expression, some subject objects such as punctuation marks, numbers, letters, I, you and he and verbs are excluded, and word segmentation is carried out to obtain one or more symptom words.
Step S103: and preprocessing the disease symptoms of the target patient according to a preset characteristic preprocessing strategy to obtain the target disease symptoms represented in a digital form.
Target condition characteristics include gender, age, symptoms, length of onset, body temperature, pulse pressure differential, heart rate rating, systolic blood pressure rating, risk rating, and severity rating.
Step S104: and inputting the target disease characteristics into a preset grading triage model to obtain grading information of the target patient.
Specifically, in this embodiment, the hierarchical triage model is a machine learning model constructed by using an XGBOOST model.
In the embodiment, firstly, the symptom, the disease onset duration, the body temperature, the heart rate, the systolic pressure, the diastolic pressure, the gender and the age of the target patient are extracted, in addition, the severe level, the dangerous level, the heart rate level, the systolic pressure level and the pulse pressure difference of the target patient are further determined according to the symptom, the heart rate, the systolic pressure and the diastolic pressure of the target patient, and then, the gender, the age, the symptom, the disease onset duration, the body temperature, the pulse pressure difference, the heart rate level, the systolic pressure level, the dangerous level and the severe level of the target patient are utilized to carry out classification so as to comprehensively integrate various body factors and recessive problems of diseases of the target patient, meanwhile, a machine learning model-a classification diagnosis classification model is utilized to carry out scientific classification, the classification accuracy is high, and a classification result with higher reference can be brought to doctors.
In yet another embodiment of the present invention, the grading level has multiple levels, and the grading information of the target patient includes the probability of the target patient at each level and the final grading level of the target patient, wherein the final grading level of the target patient is the highest probability level.
Specifically, taking 4 levels of the grading triage as an example, the grading information of the target patient includes 0% of the probability that the target patient is graded as one level, 20% of the probability that the target patient is graded as two levels, 10% of the probability that the target patient is graded as three levels, and 70% of the probability that the target patient is graded as four levels, and the severity of the patient corresponding to each level is different, for example, the severity of the patient increases from one level to four levels. In this embodiment, the highest probability level is four, and thus the final triage level of the target patient is four.
In yet another embodiment of the present invention, as shown in fig. 2, after step S104, the method further comprises the steps of:
step S105: and acquiring the number of patients waiting for diagnosis with the same diagnosis grade as the target patient at the current time according to the grading information of the target patient.
Step S106: and determining the waiting time of the target patient according to the number of waiting patients with the same triage level as the target patient and the corresponding relation between the preset waiting patient number and the waiting time.
Specifically, the more patients waiting for treatment at the same level, the longer the waiting time. In this embodiment, after the grade of the graded triage of the target patient is confirmed, the waiting time of the target patient is determined, so that the target patient can know the waiting time of the target patient, and if the waiting time is long, the target patient can move freely in the waiting time, and does not need to wait in a hospital all the time. In addition, the waiting time, the disease characteristics of the target patient, the grading information and the like can be sent to the backstage doctor end, so that the doctor can know the condition of the patient conveniently.
Further, in another embodiment of the invention, during the waiting period of the target patient, corresponding symptom knowledge and disease prevention knowledge are pushed to the target patient according to the disease characteristics of the target patient. In this embodiment, knowledge corresponding to the symptoms of the target patient and prevention knowledge of diseases are pushed for the target patient during the waiting period, so that the target patient is popular in medical knowledge, the nervous emotion of the patient can be relieved, and the patient can be helped to wear away when the waiting time is too long.
In yet another embodiment of the present invention, as shown in fig. 3, the pre-processing of the disease condition characteristics of the target patient according to the preset characteristic pre-processing strategy to obtain the target disease condition characteristics represented in digital form comprises the following steps:
step S1031: and replacing the symptoms of the target patient according to a preset synonym replacement comparison table, and converting the symptoms of the target patient into Arabic numerals according to a preset symptom conversion strategy.
Specifically, a synonym replacement comparison table is preset, words with similar meanings are uniformly named to facilitate subsequent analysis, proper nouns are separated, and symptom words are collected and converted into Arabic numerals. Specifically, a comparison table of symptoms and arabic numerals can be preset, for example, fever corresponds to arabic numeral 1, cough corresponds to arabic numeral 2, and the like.
Step S1032: and (3) eliminating full-angle punctuations and half-angle punctuations in the disease starting time of the target patient by using a regular expression, carrying out synonym replacement on the disease starting time of the target patient, converting the synonym replacement into Arabic numerals, and converting the unit of the disease starting time of the target patient into minutes.
Specifically, when synonymy substitution is performed, the words indicating the onset time are all converted into arabic numerals in minutes in standard units of time, for example, 1 hour of fever, 60 minutes of fever, and the like.
Step S1033: irregular punctuation marks and Chinese in the body temperature of the target patient are eliminated by using the regular expression, and the body temperature which is not in the preset normal body temperature range is replaced by 37 degrees.
In particular, for example, pair 38..2, 38. 2,28.2+, etc., the body temperature is normalized to be lower than 22 degrees and higher than 50 degrees, and is not considered to be within the preset normal body temperature range, and is replaced by 37 degrees.
Step S1034: and subtracting the diastolic pressure from the systolic pressure of the target patient to obtain the pulse pressure difference of the target patient.
Step S1035: and converting the sex of the target patient into Arabic numerals according to the preset corresponding relation between the sex and the numerals.
For example, a male is represented by the number 0 and a female is represented by the number 1.
Step S1036: obtaining the heart rate grade of the target patient according to the preset corresponding relation between the heart rate grade and the heart rate of the target patient;
specifically, using expert consensus criteria, heart rates are graded, for example, heart rate of 0<40 or 180<200 for 1 grade, 40<50 or greater than 180 for 2 grade, 50<55 or 100<150 for 3 grade, and the rest for 4 grade.
Step S1037: and obtaining the systolic pressure grade of the target patient according to the preset corresponding relation between the systolic pressure grade and the systolic pressure of the target patient.
Using expert consensus criteria, systolic blood pressure is graded, for example, as 0<70 for 1 grade, greater than 200 for 2 grades, 80<90 or 180<200 for 3 grades, and the remainder for 4 grades.
Step S1038: and determining the risk level of the target patient according to a preset risk level determination strategy and the heart rate level and the systolic pressure level of the target patient.
For example, by comparing the heart rate level with the systolic pressure level, if one is not at level 4 or is above level 4, the risk level is level 1, and the rest is level 0.
Step S1039: and analyzing the symptoms of the target patient according to a preset severe grade determination strategy, and determining the severe grade of the target patient.
Through word analysis, for example, when the patient encounters dangerous words such as chest distress, chest pain and the like, the patient is in grade 1, the patient is in grade 2 when the patient is febrile and has the temperature of 40 degrees, the patient is in grade 3 when the patient is febrile and has the temperature of 38 degrees, and the patient is in grade 4.
In another embodiment of the present invention, before step S101, the machine learning-based hierarchical triage method further includes: and constructing a hierarchical triage model by using the XGB OST model.
Further, in this embodiment, as shown in fig. 4, the constructing the hierarchical triage model by using the XGBOOST model includes:
step S1061: the case data of the hospital is collected as a sample case, and the number of the sample cases is multiple.
Specifically, in the present embodiment, the greater the number of sample cases, the higher the accuracy of the classification triage machine learning model obtained by machine learning.
Step S1062: and extracting disease characteristics from each sample case according to a preset characteristic extraction strategy.
The disease characteristics of the sample case include symptoms, length of onset, body temperature, heart rate, systolic pressure, diastolic pressure, sex, and age.
Specifically, the patient's complaints, the patient's basic physical information, such as body temperature, heart rate, systolic pressure, diastolic pressure, sex, age, etc., are included in the sample medical record.
The patient's complaint includes the patient's symptoms, but the patient's complaint is generally a symptom of the patient's spoken expression itself, and therefore the language is not normative. It needs to be processed to obtain the desired symptom information. In this embodiment, a pkuseg word segmentation model can be trained in advance, so as to implement word segmentation and obtain the symptoms of the patient.
Step S1063: a grading level is previously marked for each sample case according to the disease characteristics of each sample case.
In particular, the labeling can be done manually by a professional physician here depending on the disease characteristics of the sample case.
Step S1064: and preprocessing the disease characteristics of the sample case according to a characteristic preprocessing strategy to obtain target disease characteristics of the sample case expressed in a digital form.
The target condition characteristics of the sample case include gender, age, symptoms, length of onset, body temperature, differential pulse pressure, heart rate rating, systolic blood pressure rating, risk rating, and severity rating. Specifically, the method for preprocessing the disease condition characteristics of the sample case according to the characteristic preprocessing strategy is consistent with the above-mentioned steps S1031 to S1039, and is not described in detail in this embodiment.
Step S1065: and inputting the target disease state characteristics of the sample case into the XGBOOST model as input layer training samples by using the XGBOOST model, and training and learning by using the grading grade of the sample case as output layer training samples to obtain a grading triage model.
Specifically, in this embodiment, the target function of the XGBOOST model is:
The taylor function second derivative can be approximately expressed as:
definition of
The objective function is expanded using the taylor formula second derivative as:
the core algorithm idea of the XGboost is as follows: (1) and continuously adding trees, and continuously performing feature splitting to grow a tree, wherein a new function f (x) is learned to fit the residual error of the last prediction each time a tree is added. (2) When training is completed to obtain k trees, a score of a sample is predicted, namely, according to the characteristics of the sample, a corresponding leaf node is fallen in each tree, and each leaf node corresponds to a score. (3) Finally, the score corresponding to each tree only needs to be added up to be the predicted value of the sample.
As shown in the following table, in the present example, MAE (Mean Absolute Error) and MAPE (Mean Absolute Percentage Error) were used as evaluation criteria for each grade in training of the model. MAE and MAPE are good or bad indexes of a model and are used for evaluating the grading accuracy, and when the grading accuracy meets the preset accuracy requirement, the training can be stopped.
| MAE | MAPE(％) | |
| First-level | 0.8 | 7.101 |
| Second-level | 3.246 | 17.039 |
| Third-level | 19.870 | 15.237 |
| Fourth-level | 21.612 | 21.714 |
| Total | 31.213 | 14.249 |
In an embodiment of the present invention, there is also provided a hierarchical triage system based on machine learning, as shown in fig. 5, the system includes a receiving module 201, a feature extraction module 202, a feature preprocessing module 203, and a hierarchical triage module 204, where: the receiving module 201 is connected with the feature extraction module 202 and is used for receiving a questionnaire filled by a target patient to be graded and triaged and summarizing answers of the questionnaire; the feature extraction module 202 is connected with the receiving module 201 and the feature preprocessing module 203, and is configured to extract disease features of the target patient from answers of the questionnaire, where the disease features include symptoms, onset duration, body temperature, heart rate, systolic pressure, diastolic pressure, gender, and age; the characteristic preprocessing module 203 is connected with the characteristic extraction module 202 and the grading diagnosis module 204, and is used for preprocessing the disease characteristics of the target patient according to a preset characteristic preprocessing strategy to obtain target disease characteristics expressed in a digital form, wherein the target disease characteristics comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and the grading triage module 204 is connected with the characteristic preprocessing module 203 and is used for receiving the target disease symptoms and inputting the target disease symptoms into a preset grading triage model to obtain grading information of the target patient.
In still another embodiment of the present invention, as shown in fig. 6, the feature preprocessing module 203 includes a receiving unit 2031, a symptom preprocessing unit 2032, an onset duration preprocessing unit 2033, a body temperature preprocessing unit 2034, a pulse pressure difference determining unit 2035, a gender preprocessing unit 2036, a heart rate level determining unit 2037, a systolic blood pressure level determining unit 2038, a risk level determining unit 2039, and an severity level determining unit 20310, wherein: a receiving unit 2031 connected to the symptom preprocessing unit 2032, the onset time preprocessing unit 2033, the body temperature preprocessing unit 2034, the pulse pressure difference determining unit 2035, the sex preprocessing unit 2036, the heart rate grade determining unit 2037, the systolic blood pressure grade determining unit 2038, the risk grade determining unit 2039, and the severe grade determining unit 20310, and configured to receive symptoms of the target patient; a symptom preprocessing unit 2032, configured to replace the symptom of the target patient according to a preset synonym replacement comparison table, and convert the symptom of the target patient into an arabic numeral according to a preset symptom conversion policy; the onset duration preprocessing unit 2033 is configured to exclude full-angle punctuations and half-angle punctuations in the onset duration of the target patient by using a regular expression, perform synonym replacement on the onset duration of the target patient, convert the synonym replacement into an arabic number, and convert the unit of the onset duration of the target patient into minutes; the body temperature preprocessing unit 2034 is configured to exclude irregular punctuation marks and chinese characters in the body temperature of the target patient by using a regular expression, and replace the body temperature that is not within the preset normal body temperature range with 37 degrees; a pulse pressure difference determining unit 2035 for subtracting the diastolic pressure from the systolic pressure of the target patient to obtain a pulse pressure difference of the target patient; a gender preprocessing unit 2036 for converting the gender of the target patient into arabic numerals according to the preset correspondence between gender and numerals; the heart rate grade determining unit 2037 is configured to obtain the heart rate grade of the target patient according to the preset heart rate grade and heart rate correspondence and the heart rate of the target patient; a systolic pressure grade determination unit 2038, configured to obtain a systolic pressure grade of the target patient according to a preset corresponding relationship between the systolic pressure grade and the systolic pressure of the target patient; a risk level determination unit 2039, configured to determine a risk level of the target patient according to a preset risk level determination strategy, the heart rate level of the target patient, and the systolic pressure level; an importance level determination unit 20310 configured to analyze the symptoms of the target patient and determine the importance level of the target patient according to a preset importance level determination strategy.
In another embodiment of the present invention, the system further includes a grading triage model building module 205, where the grading triage model building module 205 is connected to the grading triage module 204, and is configured to collect case data of a hospital as a plurality of sample cases; extracting disease characteristics from each sample case according to a preset characteristic extraction strategy, wherein the disease characteristics of the sample case comprise symptoms, disease starting time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; marking grading levels for each sample case in advance according to the disease characteristics of each sample case; preprocessing the disease characteristics of the sample case according to a characteristic preprocessing strategy to obtain target disease characteristics of the sample case in a digital form, wherein the target disease characteristics of the sample case comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and inputting the target disease state characteristics of the sample case into the XGBOOST model as input layer training samples by using the XGBOOST model, and training and learning by using the grading grade of the sample case as output layer training samples to obtain a grading triage model.
The invention also provides computer equipment which comprises a memory and a processor, wherein the memory stores computer programs, and the processor realizes the steps of the grading triage method when executing the computer programs.
The machine learning-based grading triage method, the machine learning-based grading triage system and the computer equipment comprehensively consider various physical factors of patients and recessive problems of diseases, such as the ages, the sexes, the attack times and the like of the patients when patients are graded triage, realize scientific grading and provide grading grades with higher reference values for doctors. Meanwhile, the waiting time of the patient can be determined according to the grading result, and related symptom knowledge and a disease prevention method can be pushed for the patient during the waiting period of the patient, so that the science popularization of medical knowledge can be performed for the target patient, the nervous emotion of the patient can be relieved, and the wearing time of the patient can be helped when the waiting time is too long.
It will be appreciated by those skilled in the art that changes could be made to the details of the above-described embodiments without departing from the underlying principles thereof. The scope of the invention is, therefore, indicated by the appended claims, in which all terms are intended to be interpreted in their broadest reasonable sense unless otherwise indicated.
Claims (10)
1. A machine learning based hierarchical triage method, the method comprising:
receiving a questionnaire filled by a target patient to be classified and triaged, and summarizing answers of the questionnaire;
extracting condition features of the target patient from the answers to the questionnaire, the condition features including symptoms, length of onset, body temperature, heart rate, systolic blood pressure, diastolic blood pressure, gender, and age;
preprocessing the disease characteristics of the target patient according to a preset characteristic preprocessing strategy to obtain target disease characteristics expressed in a digital form, wherein the target disease characteristics comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade;
inputting the target disease feature into a preset grading triage model to obtain grading information of the target patient, wherein the grading triage model is a machine learning model constructed by using an XGBOOST model.
2. The machine-learning-based triage method according to claim 1, wherein the levels of the hierarchical triage have multiple levels, and the hierarchical information of the target patient includes a probability of the target patient at each level and a final triage level of the target patient, wherein the final triage level of the target patient is the highest probability level.
3. The machine learning-based hierarchical triage method according to claim 1, further comprising:
according to the grading information of the target patient, the number of the patients waiting for diagnosis with the same diagnosis grade as the target patient at the current time is obtained, and the waiting time of the target patient is determined according to the number of the patients waiting for diagnosis with the same diagnosis grade as the target patient and the corresponding relation between the preset number of the patients waiting for diagnosis and the waiting time.
4. The machine learning-based hierarchical triage method according to claim 3, further comprising: and during waiting of the target patient, pushing corresponding symptom knowledge and disease prevention knowledge to the target patient according to the disease characteristics of the target patient.
5. The machine-learning-based hierarchical triage method according to claim 1, wherein the preprocessing the disease symptoms of the target patient according to a preset symptom preprocessing strategy to obtain the target disease symptoms in a digital form comprises:
replacing the symptoms of the target patient according to a preset synonym replacement comparison table, and converting the symptoms of the target patient into Arabic numerals according to a preset symptom conversion strategy;
using a regular expression to exclude full-angle punctuations and half-angle punctuations in the disease onset duration of the target patient, carrying out synonym replacement on the disease onset duration of the target patient, converting the synonym replacement into Arabic numerals, and converting the unit of the disease onset duration of the target patient into minutes;
using a regular expression to exclude irregular punctuation marks and Chinese characters in the body temperature of the target patient, and replacing the body temperature which is not within a preset normal body temperature range with 37 ℃;
subtracting the diastolic pressure from the systolic pressure of the target patient to obtain the pulse pressure difference of the target patient;
converting the sex of the target patient into Arabic numerals according to a preset corresponding relation between the sex and the numerals;
obtaining the heart rate grade of the target patient according to the preset corresponding relation between the heart rate grade and the heart rate of the target patient;
acquiring the systolic pressure grade of the target patient according to the preset corresponding relation between the systolic pressure grade and the systolic pressure of the target patient;
determining a risk level of the target patient according to a preset risk level determination strategy and the heart rate level and the systolic pressure level of the target patient;
and analyzing the symptoms of the target patient according to a preset severe grade determination strategy, and determining the severe grade of the target patient.
6. The machine learning-based hierarchical triage method of claim 1, wherein constructing the hierarchical triage model using an XGBOOST model comprises:
acquiring case data of a hospital as a plurality of sample cases;
extracting disease characteristics from each sample case according to a preset characteristic extraction strategy, wherein the disease characteristics of the sample cases comprise symptoms, disease starting time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age;
marking a grading grade for each sample case in advance according to the disease characteristics of each sample case;
preprocessing the disease characteristics of the sample case according to the characteristic preprocessing strategy to obtain target disease characteristics of the sample case in a digital form, wherein the target disease characteristics of the sample case comprise sex, age, symptoms, onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade;
and inputting the target symptom characteristics of the sample case into the XGBOOST model as input layer training samples by using the XGBOOST model, and performing training and learning by using the grading grade of the sample case as an output layer training sample to obtain the grading triage model.
7. The grading triage system based on machine learning is characterized by comprising a receiving module, a feature extraction module, a feature preprocessing module and a grading triage module, wherein:
the receiving module is connected with the feature extraction module and used for receiving a questionnaire filled by a target patient to be graded and triaged and summarizing answers of the questionnaire;
the characteristic extraction module is connected with the receiving module and the characteristic preprocessing module and is used for extracting the disease characteristics of the target patient from the answers of the questionnaire, wherein the disease characteristics comprise symptoms, disease onset time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age;
the characteristic preprocessing module is connected with the characteristic extraction module and the grading diagnosis module, and is used for preprocessing the disease characteristics of the target patient according to a preset characteristic preprocessing strategy to obtain target disease characteristics expressed in a digital form, wherein the target disease characteristics comprise sex, age, symptoms, disease onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade;
the grading triage module is connected with the feature preprocessing module and used for receiving the target disease feature and inputting the target disease feature into a preset grading triage model to obtain grading information of the target patient.
8. The machine learning-based triage system according to claim 7, wherein the feature preprocessing module comprises a receiving unit, a symptom preprocessing unit, an onset duration preprocessing unit, a body temperature preprocessing unit, a pulse pressure difference determining unit, a gender preprocessing unit, a heart rate level determining unit, a systolic pressure level determining unit, a risk level determining unit, and an severity level determining unit, wherein:
the receiving unit is connected with the symptom preprocessing unit, the onset time preprocessing unit, the body temperature preprocessing unit, the pulse pressure difference determining unit, the gender preprocessing unit, the heart rate grade determining unit, the systolic pressure grade determining unit, the danger grade determining unit and the severe grade determining unit and is used for receiving the disease characteristics of the target patient;
the symptom preprocessing unit is used for replacing the symptoms of the target patient according to a preset synonym replacement comparison table and converting the symptoms of the target patient into Arabic numerals according to a preset symptom conversion strategy;
the onset duration preprocessing unit is used for excluding full-angle punctuations and half-angle punctuations in the onset duration of the target patient by using a regular expression, carrying out synonym replacement on the onset duration of the target patient, converting the synonym replacement into Arabic numerals, and converting the unit of the onset duration of the target patient into minutes;
the body temperature preprocessing unit is used for removing irregular punctuation marks and Chinese from the body temperature of the target patient by using a regular expression and replacing the body temperature which is not in a preset normal body temperature range with 37 ℃;
the pulse pressure difference determining unit is used for subtracting diastolic pressure from systolic pressure of the target patient to obtain pulse pressure difference of the target patient;
the gender preprocessing unit is used for converting the gender of the target patient into Arabic numerals according to a preset correspondence between the gender and the numerals;
the heart rate grade determining unit is used for obtaining the heart rate grade of the target patient according to the preset corresponding relation between the heart rate grade and the heart rate of the target patient;
the systolic pressure grade determining unit is used for obtaining the systolic pressure grade of the target patient according to the preset corresponding relation between the systolic pressure grade and the systolic pressure of the target patient;
the risk level determination unit is used for determining the risk level of the target patient according to a preset risk level determination strategy and the heart rate level and the systolic pressure level of the target patient;
and the severe grade determining unit is used for analyzing the symptoms of the target patient to be treated according to a preset severe grade determining strategy and determining the severe grade of the target patient to be treated.
9. The machine learning-based hierarchical triage system according to claim 7, further comprising a hierarchical triage model construction module connected to the hierarchical triage module for collecting hospital case data as a plurality of sample cases; extracting disease characteristics from each sample case according to a preset characteristic extraction strategy, wherein the disease characteristics of the sample cases comprise symptoms, disease starting time, body temperature, heart rate, systolic pressure, diastolic pressure, sex and age; marking a grading grade for each sample case in advance according to the disease characteristics of each sample case; preprocessing the disease characteristics of the sample case according to the characteristic preprocessing strategy to obtain target disease characteristics of the sample case in a digital form, wherein the target disease characteristics of the sample case comprise sex, age, symptoms, onset time, body temperature, pulse pressure difference, heart rate grade, systolic pressure grade, danger grade and severe grade; and inputting the target symptom characteristics of the sample case into the XGBOOST model as input layer training samples by using the XGBOOST model, and performing training and learning by using the grading grade of the sample case as an output layer training sample to obtain the grading triage model.
10. A computer device comprising a memory in which a computer program is stored and a processor which, when executing the computer program, carries out the method steps of the triage method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110860445.7A CN113724854A (en) | 2021-07-27 | 2021-07-27 | Machine learning-based grading triage method and system and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110860445.7A CN113724854A (en) | 2021-07-27 | 2021-07-27 | Machine learning-based grading triage method and system and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113724854A true CN113724854A (en) | 2021-11-30 |
Family
ID=78674236
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110860445.7A Pending CN113724854A (en) | 2021-07-27 | 2021-07-27 | Machine learning-based grading triage method and system and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113724854A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160012192A1 (en) * | 2013-03-04 | 2016-01-14 | Board Of Regents Of The University Of Texas System | System and method for determining triage categories |
| CN108922608A (en) * | 2018-06-13 | 2018-11-30 | 平安医疗科技有限公司 | Intelligent hospital guide's method, apparatus, computer equipment and storage medium |
| CN110047584A (en) * | 2019-04-23 | 2019-07-23 | 清华大学 | Hospital distributing diagnosis method, system, device and medium based on deep learning |
| CN111709233A (en) * | 2020-05-27 | 2020-09-25 | 西安交通大学 | Intelligent diagnosis guiding method and system based on multi-attention convolutional neural network |
| CN111785368A (en) * | 2020-06-30 | 2020-10-16 | 平安科技（深圳）有限公司 | Triage method, device, equipment and storage medium based on medical knowledge map |
| CN112652386A (en) * | 2020-12-25 | 2021-04-13 | 平安科技（深圳）有限公司 | Triage data processing method and device, computer equipment and storage medium |
-
2021
- 2021-07-27 CN CN202110860445.7A patent/CN113724854A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160012192A1 (en) * | 2013-03-04 | 2016-01-14 | Board Of Regents Of The University Of Texas System | System and method for determining triage categories |
| CN108922608A (en) * | 2018-06-13 | 2018-11-30 | 平安医疗科技有限公司 | Intelligent hospital guide's method, apparatus, computer equipment and storage medium |
| CN110047584A (en) * | 2019-04-23 | 2019-07-23 | 清华大学 | Hospital distributing diagnosis method, system, device and medium based on deep learning |
| CN111709233A (en) * | 2020-05-27 | 2020-09-25 | 西安交通大学 | Intelligent diagnosis guiding method and system based on multi-attention convolutional neural network |
| CN111785368A (en) * | 2020-06-30 | 2020-10-16 | 平安科技（深圳）有限公司 | Triage method, device, equipment and storage medium based on medical knowledge map |
| CN112652386A (en) * | 2020-12-25 | 2021-04-13 | 平安科技（深圳）有限公司 | Triage data processing method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107705839B (en) | Disease automatic coding method and system | |
| CN107731269B (en) | Disease coding method and system based on original diagnosis data and medical record file data | |
| CN111192680B (en) | Intelligent auxiliary diagnosis method based on deep learning and collective classification | |
| CN106874643B (en) | Method and system for automatically constructing knowledge base to realize auxiliary diagnosis and treatment based on word vectors | |
| CN107403068B (en) | Merge the intelligence auxiliary way of inquisition and system of clinical thinking | |
| CN107577826B (en) | Classification of diseases coding method and system based on raw diagnostic data | |
| CN110838368B (en) | Active inquiry robot based on traditional Chinese medicine clinical knowledge map | |
| CN112863630A (en) | Personalized accurate medical question-answering system based on data and knowledge | |
| CN110675944A (en) | Triage method and device, computer equipment and medium | |
| CN112365978A (en) | Method and device for establishing early risk assessment model of tachycardia event | |
| KR20170061222A (en) | The method for prediction health data value through generation of health data pattern and the apparatus thereof | |
| CN108492877B (en) | Cardiovascular disease auxiliary prediction method based on DS evidence theory | |
| WO2021147363A1 (en) | Text-based major depressive disorder recognition method | |
| CN110931137A (en) | Machine-assisted dialog system, method and device | |
| CN110931128A (en) | Method, system and device for automatically identifying unsupervised symptoms of unstructured medical texts | |
| CN112541066A (en) | Text-structured-based medical and technical report detection method and related equipment | |
| CN112542242A (en) | Data transformation/symptom scoring | |
| CN109410074A (en) | Intelligent core protects method and system | |
| CN112347257A (en) | Patient symptom spoken normalization method and device | |
| CN112037909A (en) | Diagnostic information rechecking system | |
| CN109907751B (en) | Laboratory chest pain data inspection auxiliary identification method based on artificial intelligence supervised learning | |
| KR20110098286A (en) | Self health diagnosis system of oriental medicine using fuzzy inference method | |
| Joshe et al. | Symptoms analysis based chronic obstructive pulmonary disease prediction in Bangladesh using machine learning approach | |
| CN113360643A (en) | Electronic medical record data quality evaluation method based on short text classification | |
| CN108831560B (en) | Method and device for determining medical data attribute data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |