Toward interpretable digital biomarkers of walking and reaching in Parkinson's disease. (9th September 2022)

• Record Type: Journal Article
• Authors: Ryu, Jihye; Torres, Elizabeth
• Is Part Of: Wearable technologies. Volume 3(2022)
• Publication Date: 2022-09-09
• Subjects: biomechanics -- human in the loop -- optimization -- human motor control -- performance characterization -- sensors; Wearable technology -- Periodicals; Robotics in medicine -- Periodicals; 610.28
• DOI: 10.1017/wtc.2022.16
• Languages: English
• ISSNs: 2631-7176

Abstract

Abstract: Multimodal digital data registered with wearable biosensors have emerged as highly complementary of clinical pencil-and-paper criteria, offering new insights in ways to detect and diagnose various aspects of Parkinson's disease (PD). A pressing question is how to combine both the clinical knowledge of PD and the new technology to create interpretable digital biomarkers easily obtainable with off-the-shelf technology. Several challenges concerning disparity in biophysical units, anatomical differences across participants, sensor positioning, and sampling resolution are addressed in this work, along with identification of optimal parameters to automatically differentiate patients with PD from controls. We combine data from a multitude of biosensors registering signals from the central (electroencephalography) and peripheral (magnetometry, kinematics) nervous systems, inclusive of the autonomic nervous system (electrocardiogram), as the participants perform natural tasks requiring different levels of intentional planning and automatic control. We find that magnetometer data during walking, across a variety of amplitude and timing signals, provide optimal separation of PD from neurotypical controls. We conclude that using multimodal signals within the context of actions that bear different levels of intent, can be revealing of features of PD that would scape the naked eye. Further, we add that clinical criteria combined with such optimal digital parameter spaces offer a far more complete picture of PD than using either one of these pieces of data alone.