CN111524598A - Perioperative complication prediction method and system - Google Patents

Abstract

The invention discloses a perioperative complication prediction method and system. The method comprises: acquiring existing perioperative complication data, constructing a perioperative complication database; , stratified random sampling according to the three stages of preoperative, intraoperative and postoperative; the sampling data of each stage is divided into training set, validation set and test set according to the proportion, and the complication probability prediction model and complications are obtained by training respectively. Occurrence time prediction model; Complication probability prediction model and complication time prediction model were used to predict the perioperative complication probability and complication time of patients. The present invention can avoid increasing the risk of perioperative complications due to the limitation of doctors, personal knowledge level and clinical experience, and can predict the occurrence probability and occurrence time of perioperative complications.

Description

A method and system for predicting perioperative complications

technical field

The invention relates to the field of medical data processing, in particular to a perioperative complication prediction method and system.

Background technique

The perioperative period refers to the whole process surrounding the operation, starting from the patient's decision to receive surgical treatment, to the surgical treatment until the basic recovery, including a period of time before, during and after the operation. Until the treatment related to this surgery is basically finished, the time is about 5-7 days before surgery to 7-12 days after surgery.

There are many factors that affect the incidence of perioperative complications in different patients, including patient age, gender, nutritional status, co-morbidities, and surgical methods. The incidence of complications in different perioperative periods is also different.

At present, the prediction and prevention of complications in the perioperative period is determined by the doctor’s personal experience, limited by the doctor’s personal knowledge level and clinical experience, and it is often impossible to judge the probability of complications in different periods of the perioperative period and whether active prevention is required, especially It is a rare complication that is often overlooked.

Therefore, it is necessary to study a computer prediction method based on an existing database to scientifically capture the probability and regularity of complications in different periods of the perioperative period.

SUMMARY OF THE INVENTION

The invention provides a perioperative complication prediction method and system, which are used to solve the technical problem of inaccurate prediction and prevention of perioperative complications by doctors.

In order to solve the above-mentioned technical problems, the technical scheme proposed by the present invention is:

A method for predicting perioperative complications, comprising the following steps:

Obtain the existing perioperative complication data and construct a perioperative complication database;

From the perioperative complications database, stratified random sampling was performed according to the three stages of preoperative, intraoperative and postoperative;

Divide the sampling data of each stage into a training set, a validation set and a test set in proportion, and train the complication probability prediction model and complication time prediction model for each stage separately;

The complication probability prediction model and the complication time prediction model were used to predict the perioperative complication probability and complication time of patients.

Preferably, the method further includes: acquiring corresponding inspection items and preventive measures data of various complications in the perioperative period, and establishing a perioperative complication prevention database;

After predicting the perioperative complication probability and complication time of the patient, according to the top three complications with complication probability, query from the perioperative complication prevention database and return the corresponding examination of the complication project;

And, after obtaining the inspection conclusion of the corresponding inspection item, it is determined whether to query and return the preventive measure data from the perioperative complication prevention database according to the inspection conclusion.

Preferably, the sampling data of each stage is divided into a training set, a validation set and a test set according to the ratio, and the ratio of the training set, the validation set and the test set is 7:2:1.

Preferably, the existing perioperative complication data include:

The basic information of the patient, including: age, gender, height, weight, nutritional status, mental status, past medical history and family medical history;

Various relevant medical examination data of the patient; the operation that the patient is going to perform; and, the time of the patient's complication, the type of complication, and the incidence rate corresponding to the type of complication.

Preferably, the method further includes: after constructing a perioperative complication database, preprocessing the original data in the perioperative complication database, and the preprocessing includes: cleaning, feature screening, and feature combination.

Preferably, the complication probability prediction model and the complication time prediction model for each stage are obtained by training, including the following steps:

The training set under each stage is trained with the GBDT model to obtain the primary prediction model;

The primary prediction model is cross-validated with the validation set data, and the secondary prediction model with the best hyperparameter combination is selected;

Use the test set to test the prediction results of the secondary prediction model. If the prediction results of the secondary prediction model on the test set meet the requirements, it is determined as the final prediction model; If the prediction results meet the requirements, it is determined as the final prediction model; the final prediction model is the prediction model of the probability of occurrence of complications and the prediction model of the time of occurrence of complications.

The present invention also provides a computer system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of any of the above methods when the processor executes the computer program.

The present invention has the following beneficial effects:

1. The perioperative complication prediction method and system of the present invention adopts the method of training the prediction model, and with the help of the powerful computing power of the computer and the algorithm of artificial intelligence, it can capture and learn from the massive complication database that different patients perform different treatments. Based on the difference in the probability of complications during surgery, a predictive model for complications at different stages during the perioperative period was constructed for different patients to predict the probability and time of complications.

2. In the preferred solution, the perioperative complication prediction method and system of the present invention can call the existing corresponding data after the complication occurrence time and probability prediction results are obtained. The test results are combined with the previous patient-related data to recommend relevant preventive measures.

In addition to the objects, features and advantages described above, the present invention has other objects, features and advantages. The present invention will be described in further detail below with reference to the accompanying drawings.

Description of drawings

The accompanying drawings constituting a part of the present application are used to provide further understanding of the present invention, and the exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the attached image:

FIG. 1 is a schematic flowchart of a method for predicting perioperative complications according to a preferred embodiment of the present invention.

Detailed ways

The embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

Referring to Fig. 1, the perioperative complication prediction method of the present invention includes the following steps:

S1. Obtain the existing perioperative complication data, and construct a perioperative complication database (may also include various perioperative complication-related influencing factors and preventive measures); Some perioperative complications data include: the basic information of the patient, including: age, gender, height, weight, nutritional status, mental status, past medical history and family medical history; various relevant medical examination data of the patient; surgery; and, the time, type of complication, and rate of complication corresponding to the patient's complication.

In this embodiment, after the perioperative complication database is constructed, (before calling or sampling the database), the original data in the perioperative complication database can also be preprocessed. The preprocessing includes: cleaning, feature screening, and feature selection. combination. Data cleaning includes removing or modifying abnormal data, removing or complementing missing data, and removing duplicate data. Feature screening is to select features that are different and related to the learning objectives of the model (for example: a patient with high blood sugar, high blood pressure, high density lipoprotein, high uric acid, a complication (cerebral hemorrhage), However, high high-density lipoprotein and high uric acid are not associated with cerebral hemorrhage, and are features that are not associated with cerebral hemorrhage). Feature combination is the multiplication of a feature by itself or other features.

S2. From the perioperative complications database, stratified random sampling is carried out according to the three stages of preoperative, intraoperative and postoperative; period of time. Preoperative refers to the decision of surgical treatment until the operation; intraoperative refers to the operation process; postoperative refers to the closely related recovery time after the operation, generally about 7-12 days. Each disease of the patient may cause risk, and the probability of this risk occurring before, during and after surgery is different, and some only occur after surgery (some only occur before surgery, and some only occur after surgery). Intraoperative), that is, before sampling, the database can be divided into three sub-databases according to preoperative, intraoperative and postoperative.

S3. Divide the sampling data of each stage into a training set, a validation set and a test set according to the proportion (in this embodiment, the ratio of the training set, the verification set and the test set is 7:2:1), and obtain each stage by training separately. Complication probability prediction model and complication time prediction model. The following steps can be preferably used:

S301. Use the GBDT (Gradient Boosting Decision Tree, gradient boosting iterative decision tree) model to train the training set in each stage to obtain a primary prediction model;

S302. Cross-validate the primary prediction model with the validation set data, and select the secondary prediction model with the hyperparameter combination with the best prediction effect; the hyperparameters are the parameter values set before starting the learning, such as the learning rate, the number of iterations, and the The number of hidden layers, etc.

S303. Use the test set to test the prediction result of the secondary prediction model. If the prediction result of the secondary prediction model on the test set meets the requirements (during the model training and optimization process, the error is used as the optimization basis. The smaller the error, the better. When the error reaches When the set requirements are set, the iteration can be stopped; after the training is completed, the accuracy rate is used as the evaluation basis, and the accuracy rate of the model meets the set requirements), then it is determined as the final prediction model; otherwise, the hyperparameter combination is further adjusted until If the prediction results of the secondary prediction model on the test set meet the requirements, it is determined as the final prediction model; the final prediction models are the complication probability prediction model and the complication time prediction model.

After the final prediction model is obtained, the prediction model can be encapsulated into application software or web pages, only the user (usually the doctor input from the terminal) extracts or inputs the patient-related data, and the trainer automatically captures the difference between the probabilities of complications in different patients , so as to learn the relationship between the indicators of patients with complication rates at different stages, and then the prediction results can be obtained. The patient should try to provide all the indicators in the database as comprehensively as possible. Usually, the more comprehensive the patient's indicators are, the more accurate the prediction will be.

S4. Use the complication probability prediction model and the complication time prediction model to predict the perioperative complication probability and complication time of the patient.

Using the method of training the prediction model, with the help of the powerful computing power of the computer and the algorithm of artificial intelligence, it is possible to capture and learn the difference in the probability of complications of different patients when performing different operations from the massive complication database, and construct different patients to perform different operations. A predictive model of complications at different stages of the perioperative period was used to predict the probability and timing of complications.

In practical application, on the basis of the above steps, the perioperative complication prediction method of the present invention can also add the following steps for optimization:

S5, acquiring the corresponding inspection items and preventive measures data of various complications in the perioperative period, and establishing a perioperative complication prevention database; this step can be performed together with step S1;

S6. After predicting the perioperative complication probability and complication occurrence time of the patient, according to the top three complications ranked by the complication probability, query from the perioperative complication prevention database and return the corresponding complications of the complication or directly query and return preventive measures data from the perioperative complication prevention database (see Figure 1, if no corresponding inspection items or subsequent inspection items are required, step S7 is not performed) ;

S7. After obtaining the inspection conclusions of the corresponding inspection items, determine whether to query and return the preventive measures data from the perioperative complication prevention database according to the inspection conclusions; if necessary, according to the newly added inspection conclusions, you can also return In step S6, the complication occurrence probability and complication occurrence time prediction of the complication is performed again.

Through the above steps, after obtaining the complication time and probability prediction results, the existing corresponding data can be called, and it is recommended to further improve the medical examination. measure.

This embodiment also provides a computer system, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the steps of the methods of the above embodiments when executing the computer program.

The present invention has the following beneficial effects:

1. The perioperative complication prediction method and system of the present invention, with the help of the powerful computing power of the computer and the algorithm of artificial intelligence, can capture and learn from the massive complication database of the complication probability of different patients during different operations. difference,

2. In the preferred solution, the perioperative complication prediction method and system of the present invention can call the existing corresponding data after the complication occurrence time and probability prediction results are obtained. The test results are combined with the previous patient-related data to recommend relevant preventive measures.

To sum up, the present invention constructs a prediction model of complications occurring in different perioperative stages when different patients perform different operations by means of training a prediction model, so as to predict the occurrence probability and occurrence time of complications. It can avoid the limitation of doctors and personal knowledge and clinical experience, or ill-consideration, and increase the risk of perioperative complications. Further, according to the prediction results, it is recommended to further improve the medical examination, so as to overcome the problem of inexperience of doctors, and to ensure the scientificity and pertinence of the follow-up recommendations. Complication prediction can improve doctors' vigilance and take corresponding preventive measures to reduce complications and perioperative risks.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (7)

1. a perioperative complication prediction method, is characterized in that, comprises the following steps: Obtain the existing perioperative complication data and construct a perioperative complication database; From the perioperative complications database, stratified random sampling was performed according to the three stages of preoperative, intraoperative and postoperative; Divide the sampling data of each stage into a training set, a validation set and a test set in proportion, and train the complication probability prediction model and complication time prediction model for each stage separately; The complication probability prediction model and the complication time prediction model are used to predict the perioperative complication probability and complication time of the patient. 2. The method for predicting perioperative complications according to claim 1, wherein the method further comprises: acquiring corresponding inspection items and preventive measures data of various complications in the perioperative period, and establishing the perioperative period complication prevention database; After predicting the perioperative complication probability and complication time of the patient, according to the top three complications with complication probability, query from the perioperative complication prevention database and return the corresponding complication of the complication inspection items; And, after acquiring the inspection conclusion of the corresponding inspection item, it is determined whether to query and return preventive measure data from the perioperative complication prevention database according to the inspection conclusion. 3. The method for predicting perioperative complications according to claim 1, wherein the sampling data of each stage is divided into a training set, a verification set and a test set in proportion, and the ratio is a training set, a verification set, and a test set. The ratio of set to test set is 7:2:1. 4. The method for predicting perioperative complications according to claim 1, wherein the existing perioperative complication data comprises: The basic information of the patient, including: age, gender, height, weight, nutritional status, mental status, past medical history and family medical history; Various relevant medical examination data of the patient; the operation that the patient is going to perform; and, the time of the patient's complication, the type of complication, and the incidence rate corresponding to the type of complication. 5 . The method for predicting perioperative complications according to claim 1 , wherein the method further comprises: after constructing a perioperative complication database, analyzing the original data in the perioperative complication database. 6 . Preprocessing is performed, and the preprocessing includes: cleaning, feature screening, and feature combination. 6. The perioperative complication prediction method according to any one of claims 1 to 5, characterized in that, the complication probability prediction model and the complication time prediction model of each stage are obtained by said training respectively, Both include the following steps: The training set under each stage is trained with the GBDT model to obtain the primary prediction model; The primary prediction model is cross-validated with the validation set data, and the secondary prediction model of the hyperparameter combination with the best prediction effect is selected; Test the prediction result of the secondary prediction model with the test set, if the prediction result of the secondary prediction model meets the requirements for the test set, it is determined as the final prediction model; otherwise, further adjust the hyperparameter combination until the If the prediction result of the secondary prediction model on the test set meets the requirements, it is determined as the final prediction model; the final prediction model is the complication probability prediction model and the complication time prediction model. 7. A computer system comprising a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor implements the above claims 1 to 6 when executing the computer program the steps of any of the methods.

Images