Study On Diagnosis And Treatment Assistance Decision Support Based On Electronic Medical Records Of Rheumatoid Arthritis Patients

As an important component of medical data,electronic medical record data contains rich clinical diagnosis and treatment information.Using machine learning methods to mine potentially effective information from patient electronic medical record data can provide a reference for doctors’ diagnosis and treatment decisions.Rheumatoid arthritis(RA),as a highly disabling disease requiring lifelong treatment,brings people not only great pain physically and mentally,but also poses a serious economic burden to families and society.Currently,there are about 5 million RA patients in China,but the number of doctors in the department of rheumatism is seriously insufficient.Therefore,the research on diagnosis and treatment decisionmaking based on electronic medical records of RA patients has important practical significance.Based on this,this thesis takes the electronic medical record of RA patients as the research object and uses machine learning methods to carry out research from two aspects:RA comorbidity diagnosis decision-making and RA treatment decisionmaking.The main work and conclusions of this thesis are as follows:Firstly,for the decision-making task of diagnosing RA comorbidities,the third chapter of this thesis aims to construct a diagnostic model for RA comorbidities to assist doctors in diagnosing RA comorbidities and reducing the occurrence of missed diagnoses.In our study,association rules and chi-square independence tests are firstly used to explore the correlation between the eight common RA comorbidities:type 2 diabetes,hypertension,osteoarthritis,osteoporosis,secondary Sjogren’s syndrome,interstitial lung disease,cervical spondylosis and lumbar disc herniation.Then,feature data is extracted from the electronic medical records and these features are preprocessed.Next,RF-RFE algorithm and RF algorithm are used to select features for each disease.Finally,the effect of five typical multi-label classification algorithms in the diagnosis of RA comorbidities is compared and analyzed.The results show that in the overall evaluation indicators selected in this study,the CC algorithm performs well on most evaluation indicators,and the BR algorithm can also achieve good results.The RAKELD algorithm and RAKELO algorithm can also outperform the BR algorithm and CC algorithm on some evaluation indicators,while the LP algorithm performs worst on most indicators.In addition,both CC and BR algorithms perform well in the diagnosis of most diseases,while LP algorithm performs worst in the diagnosis of most diseases.Secondly,for the treatment decision-making task of RA,the fourth chapter of this thesis aims to construct a RA treatment model based on MDP to provide corresponding RA treatment strategy reference.Our study proposes to apply Markov Decision Process(MDP)to the treatment of RA.For the parameters needed for establishing MDP,our study gives the definition and uses clinical data to infer them.Firstly,our study uses laboratory indexes of patients to measure the health states,and then regards the traditional Chinese medicine used by patients as the basis of action.And next the sum of the improvement degree of patients’ indexes and the length of hospital stay between two laboratory indexes tests are regarded as the treatment reward and treatment cost respectively.Finally,the relative value iterative algorithm is used to solve the problem,and the corresponding treatment strategy,treatment reward and treatment cost are obtained.The experimental results show that the treatment reward obtained in our study is higher than that of the hospital,and the treatment cost is lower than that of the hospital.It has a certain clinical application value to apply MDP model in the treatment of rheumatoid arthritis in traditional Chinese medicine.Finally,based on the above research results,we propose some suggestions for constructing an RA diagnosis and treatment assistant decision support system.