On automated assessment of Levodopa-induced dyskinesia in Parkinson's disease

A method for the analysis of accelerometer and gyroscope signals in order to automatically assess the Levodopa-induced dyskinesia (LID) in patients with Parkinson's disease is presented in this paper. Several accelerometers and gyroscopes are placed on certain positions on the subject's body and the obtained signals are analyzed in order to extract several features that depict the energy distribution over the frequency spectrum and the non-linear properties of the signal. These features are fed into an artificial neural network which is used for LID detection and severity classification. The method has been evaluated using a group of 29 subjects. Results are presented related to the body locations that the accelerometers and the gyroscopes are placed. The obtained results indicate high classification ability (84.3% average classification accuracy).