Assessment of Tremor Severity in Patients with Essential Tremor Using Smartwatches

Abstract

[Abstract] This paper presents a classification model for the automatic quantification of tremor severity in patients with essential tremor (ET). The system is based on the signals measured by two commercial smartwatches that the patients wear on their wrist and ankle. The smartwatches register acceleration and angular velocity in these body segments. A set of nine tremor features were used to train the classification algorithm. The proposed algorithm is based on a C4.5 decision tree classifier. It is able to assess rest and kinetic (postural or action) tremor. The method was evaluated using data collected from thirty-four patients with ET. The algorithm classifies the severity of tremor in five levels 0-4 corresponding to those in the Fahn-Tolosa-Marin tremor rating scale with a 94% accuracy. The method can be implemented in a networked platform for the remote monitoring and assessment of movement disorders such as ET or Parkinson’s disease.