Evaluation Method of Cognitive Level Based on Machine Learning and Wearable Device

Dementia is an incurable syndrome, but early detection and treatment of dementia can help slow down the progress of dementia. Recent studies have highlighted the association between overall cognitive function, motor function, and gait abnormalities. In this study, we enrolled 302 participants from the Rehabilitation Hospital Affiliated to the National Rehabilitation Aids Research Center, screening 193 according to set criteria, including 137 patients with Mild Cognitive Impairment (MCI) and 56 healthy controls (HC). Gait parameters were collected using a wearable device while participants performed single and dual tasks. Focusing on gait cycle, kinematics parameters, and time–space parameters, we employed recursive feature elimination to identify crucial features. Utilizing the participant's Moca score as the response variable, we established a machine learning model that quantitatively evaluates cognitive levels based on gait features. The temporal and spatial parameters of toe-off angle and heel strike angle emerged as pivotal markers with significant clinical relevance for evaluating cognitive levels. These parameters hold substantial clinical application value in the prevention or delayed onset of dementia.