Continuous Parkinsonian Tremor Estimation Using Motion Data

Tremor is one of the main symptoms of Parkinson's Disease (PD) that reduces the quality of life of affected patients. Tremor is measured as part of the Unified Parkinson Disease Rating Scale (UPDRS) part III. However, the assessment is based on onsite physical examinations and do not necessarily represent the patients' tremor experience in their day-to-day life. In this work, we developed two methods based on deep long short-term memory (LSTM) networks and gradient tree boosting to estimate Parkinsonian tremor using gyroscope sensor signals collected as the patients performed a variety of free body movements. The developed methods were assessed on data from 24 PD subjects. Subject-based, leave-one-out cross-validation demonstrated that the method based on gradient tree boosting provided a high correlation (r=0.93 (p<0.0001)) between the estimated and clinically assessed tremor subscores in comparison to the LSTM-based method with (r=0.77 (p<0.0001)).