Caracterização do tremor parkinsoniano de membros superiores por meio de sensores inerciais usando os métodos KNN e FCM

Abstract

Tremor and all the derivations of pathologies implicated in it have in the elderly setting the most representative population, being a primary source responsible for physical deterioration and in the social life of the subject. Nowadays, the elderly population is increasing and represents a great worldwide change, in Brazil, it has had a significant percentage increase in the last years. The elderly are more inclined to develop tremors, but tremor may indicate more complex syndromes, such as Parkinson's disease. Thus, this study focuses on the classification of tremor severity in upper limbs in subjects with Parkinson's disease through features extracted from inertial sensors that can be used to analyze the relationship between tremor and its severity using KNN and FCM classifier. For the analysis, inertial signals from the signal database of the Center for Innovation and Technological Evaluation in Health (NIATS) were used. The characteristics extracted from the signal were divided between: 1) amplitude; 2) frequency; 3) entropy; 4) variability. The data analysis consisted of classification of the various characteristics extracted from the sensor activity, using a classifier based on the K-nearest neighbor (KNN) and other based on Fuzzy C-Means (FCM) and the sensitivity, specificity, accuracy and precision of the classifier. In total, 27 individuals participated in the study, 15 of them with Parkinson's disease, and 12 healthy ones were divided into two groups and the same protocol was used in both groups. The results obtained show that it is possible to separate normal people, from those with Parkinson's disease with better KNN results over FCM with sensitivity and accuracy above 0,96.