Disease Detection Using Machine Learning in Vital Sign Data Telemonitoring

Abstract

For elderly patients, it is very important that the sudden onset of acute illnesses or the unexpected worsening of chronic diseases be detected and related information be sent to doctors as soon as possible. Herein, we report on attempts to identify suitable detection methods by analyzing the time sequences of several vital data using two methods, principal component analysis (PCA) and support vector machine (SVM). Using PCA for vital data for four patients, we found that peak illness indicators could be detected by using first and second principal components in three cases, but that detection was difficult in one case. Using SVM, we could obtain an 86% accuracy level. These results show that it is possible to detect acute illness and chronic-disease-related symptoms more precisely by employing machine learning (ML)-based methods.