Application of Machine Learning to the Development of a Quantitative Clinical Biomarker for the Progression of Parkinson’s Disease

Bambi R. Brewer, Sujata Pradhan, George Carvell, Patrick Sparto, Deborah Josbeno, and Anthony Delitto
University of Pittsburgh
Pittsburgh, PA 15260


Effective clinical trials for neuroprotective interventions for Parkinson’s disease (PD) require a way to quantify an individual’s motor symptoms and analyze the change in these symptoms over time. Clinical scales provide a global picture of function but cannot precisely measure specific aspects of motor control. We have used commercially available sensors to create a system called ASAP (Advanced Sensing for Assessment of Parkinson’s disease) to obtain a quantitative and reliable measure of motor impairment in early PD. The ASAP protocol measures grip force as an individual tracks a sinusoidal or pseudorandom target force. The individual performs the tracking task under three conditions of increasing cognitive load. Thirty individuals with PD have completed the ASAP protocol. The ASAP data were summarized in terms of 36 variables, and machine learning techniques were used to predict an individual’s score on the Unified Parkinson Disease Rating Scale based on this data. We observed a mean prediction error of approximately 5 UPDRS points, and the predicted score accounted for approximately 70% of the variability of the UPDRS. These results indicate that the ASAP protocol should be able to measure changes in an individual’s Parkinsonian symptoms over time.


Parkinson’s disease, outcome measure, biomarker, machine learning


This work was supported in part by the Foundation of Physical Therapy Promotion of Doctoral Studies II Scholarship 2007 and by resources and the use of facilities at the Human Engineering Research Laboratories, VA Pittsburgh Healthcare System.

Author Contact Information:

Bambi Brewer, PhD, Department of Rehabilitation Science and Technology, University of Pittsburgh, 5044 Forbes Tower, Pittsburgh, PA 15260, Office Phone (412) 383-6594 EMAIL: