UPDRS tracking using linear regression and neural network for Parkinson’s disease prediction

Abstract The Unified Parkinson’s Disease Rating Scale (UPDRS) is often used to track Parkinson's disease (PD) but it requires costly and logistically inconvenient for patient and clinical staff. In this work we present clinically useful accuracy replication of UPDRS, so we can classify the disease’s severity of the patients with, and predict the evolution of PD based on those results. We map the features extracted from the speech to UPDRS using Least-squares regression technique and neural network. We applied our techniques on large database of PD speech (~6,000 recordings from 42PD patients). And we compare our results with state of the art.