HealthDay News — An artificial intelligence tool can help remotely assess the severity of Parkinson disease symptoms, according to a study published online Aug. 23 in npj Digital Medicine.
Saiful Islam, from the University of Rochester in New York, and colleagues developed an artificial intelligence system to remotely assess the motor performance of individuals with Parkinson disease. The analysis included 250 global participants (172 with Parkinson disease and 78 controls) who performed a standardized motor task involving finger-tapping with each hand in front of a webcam. Overall, 489 videos were analyzed (244 videos for the left hand and 245 for the right hand). Severity of Parkinsonian symptoms diagnosed by the finger-tapping task was compared to three expert neurologists’ ratings.
The researchers found that interrater reliability was excellent, with an intraclass correlation coefficient of 0.88. The machine learning model trained on these measures outperformed two certified raters of the Movement Disorder Society Unified Parkinson’s Disease Rating Scale (mean absolute error [MAE], 0.58 points versus raters’ average MAE of 0.83 points). However, expert neurologists (0.53 MAE) slightly outperformed the model.
“The methodology can be replicated for similar motor tasks, providing the possibility of evaluating individuals with Parkinson disease and other movement disorders remotely, objectively, and in areas with limited access to neurological care,” the authors write.