Fatemeh Vakili, Zahra Vakili, M. Kargari, Mehran Ghaffari
{"title":"Supervised Machine Learning Models for Multiple Sclerosis Relapse Probability Prognosis using Clinical Data","authors":"Fatemeh Vakili, Zahra Vakili, M. Kargari, Mehran Ghaffari","doi":"10.1109/ICWR57742.2023.10139225","DOIUrl":null,"url":null,"abstract":"Multiple sclerosis (MS) is an autoimmune disease that causes physical disabilities over time. The prevalence of this disease is increasing in the world. Since this disease causes physical disabilities, the prognosis of relapse is vital. Currently, doctors use magnetic resonance images to diagnose the disease and its process. But this method suffers from some problems such as operator fatigue, the difficulty of using magnetic resonance images, the complications of its use, etc. For this reason, machine learning techniques can be used as diagnostic aids for doctors. Some data were from the hospital’s HIS and electronic health records. Some other physical files of patients have been used. Overall, 492 records of 84 patients with relapsing-remitting type were collected. We studied the clinical data of patients with disease onset for two years. The data were collected from the MS clinic of Imam Hossein Hospital in Tehran. The patient's clinical symptoms were recorded from February 2021 to February 2023. Mainly, the patient's symptoms were classified into 7 groups of relapses. Also, the patient’s personal information and relapse symptoms were examined. With the help of machine learning techniques, we will design a model that predicts the probability of relapse in the next year. The data were unbalanced, so the SMOTE oversampling method has been used. The results show that these models predicted relapse in patients with high accuracy and precision. This research used standard machine learning models SVC, K-NN, Logistic Regression, Decision Tree, Random Forest, Naive Bayes, MLP, XGBoost, Ada Boost, and Gradient Boost. The results of evaluating the models based on accuracy and precision criteria showed that the SVC with 98.837% precision and 97.973% accuracy and followed by 97.674% precision and 97.297% accuracy in both LogisticRegression and MLP models are the best model and can provide an appropriate prognosis.","PeriodicalId":146989,"journal":{"name":"2023 9th International Conference on Web Research (ICWR)","volume":"210 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 9th International Conference on Web Research (ICWR)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICWR57742.2023.10139225","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Multiple sclerosis (MS) is an autoimmune disease that causes physical disabilities over time. The prevalence of this disease is increasing in the world. Since this disease causes physical disabilities, the prognosis of relapse is vital. Currently, doctors use magnetic resonance images to diagnose the disease and its process. But this method suffers from some problems such as operator fatigue, the difficulty of using magnetic resonance images, the complications of its use, etc. For this reason, machine learning techniques can be used as diagnostic aids for doctors. Some data were from the hospital’s HIS and electronic health records. Some other physical files of patients have been used. Overall, 492 records of 84 patients with relapsing-remitting type were collected. We studied the clinical data of patients with disease onset for two years. The data were collected from the MS clinic of Imam Hossein Hospital in Tehran. The patient's clinical symptoms were recorded from February 2021 to February 2023. Mainly, the patient's symptoms were classified into 7 groups of relapses. Also, the patient’s personal information and relapse symptoms were examined. With the help of machine learning techniques, we will design a model that predicts the probability of relapse in the next year. The data were unbalanced, so the SMOTE oversampling method has been used. The results show that these models predicted relapse in patients with high accuracy and precision. This research used standard machine learning models SVC, K-NN, Logistic Regression, Decision Tree, Random Forest, Naive Bayes, MLP, XGBoost, Ada Boost, and Gradient Boost. The results of evaluating the models based on accuracy and precision criteria showed that the SVC with 98.837% precision and 97.973% accuracy and followed by 97.674% precision and 97.297% accuracy in both LogisticRegression and MLP models are the best model and can provide an appropriate prognosis.