AI Algorithms Detect Parkinson’s Based on Watch Movements Verily Study

A new analysis using machine learning and sensor data from the Parkinson’s Progress Markers Initiative (PPMI), a digital health research program sponsored by The Michael J. Fox Foundation for Parkinson’s Disease Research (MJFF), were able to distinguish people with and without Parkinson’s disease, according to a small study conducted by Cohen Veterans Biosciences (CVB).

Specific information, such as movement data, was recorded using the Verily Study Watch, a wrist-worn device used by participants up to 23 hours a day for several months.

“This study demonstrates the possibility of harnessing unconstrained, unlabeled wearable sensor data to accurately detect Parkinson’s disease using powerful deep learning methods,” said Lee Lancashire, PhD, Principal Investigator. of the study and director of information at the CVB, in a Press release.

“With this combination of wearables and [artificial intelligence]we are moving closer to monitoring individual healthcare-related activities such as motor function outside of the clinic, unlocking the potential for early detection and diagnosis of diseases such as Parkinson’s disease,” he said. he adds.

recommended reading

The study, “Deep Learning for Daily Monitoring of Parkinson’s Disease Outside the Clinic Using Wearable Sensorswas published in the journal Sensors.

Parkinson’s disease is a progressive neurological condition characterized by motor symptoms, such as bradykinesia (slow movements), gait disturbances and tremors. The diagnosis can be difficult, in particular because there are no objective biomarkers of the disease. Better methods of tracking disease progression are also needed so that clinicians can provide individualized care and treatment.

According to the researchers, sensor technology is developing rapidly. Many studies have focused on finding digital biomarkers related to specific movement characteristics of Parkinson’s disease. However, the data from these studies were recorded in controlled laboratory environments and do not reflect patient movements in their everyday environment.

Now, a team of researchers has sought to collect data with the Verily Study watchworn daily by a subset of participants in the PPMI investigate and determine if their newly developed computer-generated algorithms can be used to identify people with Parkinson’s disease based on walking-like events.

The PPMI (NCT01141023) is a longitudinal observational study of people with and without Parkinson’s disease. Its objective is to identify biomarkers associated with the risk, onset and progression of Parkinson’s disease.

In 2018, PPMI launched a sub-study using the Verily Study Watch at sites in the United States; all subjects enrolled in the PPMI were invited to participate.

“So compared to other data types associated with the full PPMI dataset, wearable data is not just about tracking the progress of [Parkinson’s] beginning at an early and untreated stage, but can begin at any time along the trajectory,” the researchers wrote.

For the new analysis, the researchers extracted participant data from 11 people from the PPMI database in June 2021: seven patients clinically diagnosed with Parkinson’s disease and four controls. Of the people with Parkinson’s disease, five had genetic risk variants in the LRKK2, ACSand SNCA genes, and two were newly diagnosed patients who remained untreated during the study.

Patients were asked to wear the Verily Study Watch for up to 23 hours a day for several months to two years during their daily activities.

100% accuracy

The new algorithm showed 100% accuracy for a diagnosis of Parkinson’s disease based solely on participants’ walking motion data accumulated over a day. According to the researchers, this can be interpreted “as the ability to identify subtle changes in gait related to [Parkinson’s] that are not counted in the UPDRS [Unified Parkinson’s Disease Rating Scale] scores. The UPDRS is a commonly used scale for assess the severity of symptoms of Parkinson’s disease.

It could also distinguish with nearly 90% accuracy between people with and without a diagnosis of Parkinson’s disease on five-second walking movements.

“While further studies are needed, we are excited about the potential for using sensor data obtained through a patient’s normal activity to allow physicians to monitor and classify [Parkinson’s] symptoms through easy-to-obtain objective measures that can be used to improve clinical decision-making and guide therapeutic interventions,” said Mark Frasier, PhD, study co-author and scientific director of MJFF.

Funds from CVB and a grant from MJFF supported this study.

Sharon D. Cole