Computer-aided Diagnosis Model For Acute Aggravation And Chronic Obstructive Pulmonary Disease

Chronic obstructive pulmonary disease(COPD)is a common and frequentlyoccurring disease that seriously endangers human health,ranking the fourth leading cause of death in the world.At present,the diagnosis method is comprehensive diagnosis through clinical symptoms,signs,laboratory examination indicators and so on.General diagnostic methods have long diagnostic time,high treatment cost,and lack of quantitative indicators in describing clinical symptoms,which can easily lead to missed diagnosis and misdiagnosis.In this paper,through the analysis of the specific clinical characteristics of acute exacerbated COPD and the use of natural language processing technology,we established an assistant diagnosis model for electronic medical records of acute exacerbated COPD.This study has very important academic value for the application of electronic medical records in the field of medical assistant diagnosis.This paper mainly studies from the following points and innovations are as follows:(1)Comparing with the current single use of data-driven or expert system for disease diagnosis in the medical field,this paper combines NLP technology,through data-driven and knowledge-driven,text categorization processing,to make features more interpretable for patients,which is also the innovation of this paper;(2)For knowledge-driven,it is mainly to read expert manuals manually,and communicate with experts.Fixing the existing features of AECOPD;(3)For data-driven,according to the clinical characteristics of AECOPD,entity extraction and word frequency feature extraction of electronic medical record data are combined;(4)Dictionary features are obtained from feature entities derived from data-driven and knowledge-driven,and different classifiers are designed according to different features.In this paper,we use SVM classifier based on word frequency and dictionary,and use Xgboostclassifier based on gradient lifting tree to fuse the features of the two classifiers;(5)After learning the classifier,we get the exact value of the acute aggravation interval,which can assist doctors to diagnose whether it is AECOPD or not.The electronic medical record data used in this paper were collected by Beijing Sino-Japanese Friendship Hospital from September 2017 to August 2018.Among them,467 cases were AECOPD patients and 531 cases were non-AECOPD patients.By sorting out the knowledge in the expert manual and preprocessing the original electronic case text,word segmentation and entity extraction,word frequency feature extraction,and finally using automatic classification algorithm to identify and classify the electronic case text.Among them,the sensibility ccof Xgboost classifier based on the fusion of the two features reaches88.6%.The experimental results show that it is feasible to use machine learning method to classify electronic medical records for assisting doctors in diagnosis.