Unobtrusive and Pervasive Monitoring of Geriatric Subjects for Early Screening of Mild Cognitive Impairment

摘要

The primary marker for on-boarding a neurological patient for engagement is old-age. Hence, cognitive impairment becomes a primary focus of geriatric care. Instrumented elderly care homes, providing ambient assisted living (AAL) use sensors for monitoring the activities of daily living (ADL) of users. Though the primary focus of such monitoring has been to find trends in the physical health of the subject, recent studies have indicated that the inferences can also be used for research on cognition. In this paper, we explore the use of unobtrusive, non-contact ADL sensors for early detection of Mild Cognitive Impairment in the geriatric population. We show the feasibility of using deep learning techniques to make such inferences. We handled the case of missing sensor data due to sensor failures using time-series prediction and based on the sensor data fea- tures, we perform an RNN and auto-encoder based methodology for screening subjects with probable Mild Cognitive Impairment. Further, this information is used to design a classifier to predict the future cases of illness.