BEAT-PD is a data challenge designed to benchmark new methods to predict Parkinson’s disease progression. Specifically, its objective is to determine whether disease severity and progression can be assessed from passive sensor data collected during daily life.
View the winners at https://www.synapse.org/#!Synapse:syn20825169/wiki/596118.
Teams participating in the Challenge had access to raw sensor (accelerometer and gyroscope) time-series data that can be used to predict individual medication state and symptom severity. Using the sensor data, participants were asked to tackle specific subchallenges. Teams could submit methods to predict (1) on/off medication status, (2) dyskinesia severity, and/or (3) tremor severity.
Training/test set partitioning was done within study subjects to facilitate building of personalized predictive models. Due to the wide range in number of observations by subject, accuracy was assessed within study subject. The subject-specific accuracy was then weighted across subjects to arrive at a final score.
The Challenge was conducted by MJFF and Sage Bionetworks in partnership with Evidation Health, Northwestern University, Radboud University Medical Center, and BRAIN Commons.