Parkinson’s disease (PD) is the fastest growing neurological disorder worldwide, with an expected doubling of patients to 12 million by 2040. PD is an invariably progressive condition, leading to severe impairments in the late stages. Disease-modifying treatments are being developed, but testing their efficacy is hampered by the lack of objective biomarkers.
Unobtrusive wearable sensors now allow us to capture objectively and continuously how patients function in their natural environment. However, because reference datasets were scarce, we lack models to extract relevant insights from the raw sensor signals. We now have access to two, worldwide unique studies: the Personalized Parkinson Project, continuously monitoring 650 PD patients for 2 years using a multi-sensor smartwatch, and the Parkinson@Home study, providing video-referenced sensor data during unscripted activities in the home environment of 25 PD patients and 25 controls.
To build new models for monitoring PD symptoms, we will exploit recent developments in unsupervised deep learning, combined with weakly supervised methods. The results will be integrated into a modular toolbox for clinical researchers to efficiently pre-process sensor data and extract digital biomarkers for PD progression. Such open-source tools have the potential to revolutionize clinical trial design, and create new opportunities for telemedicine.