Machine Learning in the Parkinson’s disease smartwatch (PADS) dataset

This paper introduces the Parkinson's Disease Smartwatch (PADS) dataset, a comprehensive collection of data from a three-year study involving 504 participants with Parkinson's disease (PD), differential diagnoses (DD), and healthy controls (HC). Data was collected using a multimodal smartphone app integrated with smartwatches, capturing over 5000 clinical assessment steps. Machine learning (ML) models, combining classical signal processing and deep learning techniques, achieved 91.16% balanced accuracy in classifying PD vs. HC and 72.42% for PD vs. DD. The PADS dataset, with its extensive annotations, offers valuable resources for further research into movement disorder biomarkers.

Julian Varghese, Alexander Brenner, Michael Fujarski, Catharina Marie van Alen, Lucas Plagwitz, Tobias Warnecke