Parkinson’s disease remotely monitorable with machine learning

Researchers have developed an algorithm to help assess the response of Parkinson’s patients to medication while clinicians are at work and the patients are typing on personal computers in their homes.

Lead author Michele Matarazzo, MD, of the University of British Columbia and colleagues in Spain describe their work in a study published online June 18 in Movement Disorders.

The team remotely monitored the typing patterns of 31 participants recently diagnosed with Parkinson’s disease as they began treatment with dopaminergic drugs and as they continued with the medication regimen for 24 weeks.

The researchers simultaneously monitored 30 age-matched controls who were similarly diagnosed but weren’t prescribed the drugs.  

From the typing data, Matarazzo and colleagues created a novel algorithm based on recursive neural networks to help detect the medication’s effects.

They further tested their algorithm’s accuracy at predicting the medication’s efficacy as early as 21 weeks before the final clinical outcome at the project’s six-month mark.

Using as their clinical benchmark a widely used rating scale for evaluating the severity of Parkinson’s, the team found the algorithm had only moderate to fair agreement with the scale.

However, study participants classified as responders at the final visit—indicating their placement in the cohort that received the medication—had higher scores on the algorithm when compared with the participants whose ratings on the scale remained steady.

“This preliminary study suggests that remotely gathered unsupervised typing data allows for the accurate detection and prediction of drug response in Parkinson’s disease,” the authors concluded.

Recent advances in AI and related technologies “are opening a new opportunity to remotely evaluate motor features in people with Parkinson’s disease (PD),” they commented. “Typing on an electronic device … could allow for objectively and unobtrusively monitoring parkinsonian features and response to medication in an athome setting.”