Brain Canada Brain Canada
FR Donate

Detecting Behaviours of Risk in Nursing Homes using Deep Learning

Principal Investigator:
  • Shehroz Khan, University Health Network
  • Alzheimer's Association

Project Overview

Alzheimer’s Association International Research Grant Program: Alzheimer’s Association Research Grant (AARG)

The main research question is to develop algorithms to detect behaviours of risk from videos recorded in the common area of a dementia care unit. People with dementia can have changes in judgement and behavior leading to events of risk, such as agitation. In residential settings, video surveillance is often times used in the common areas to help monitor the environment, but they are not continuously monitored. In this study, we propose to use video data collected and annotated with events of risk as part of a previous study. Our aim is to develop deep learning algorithms that are able to distinguish the patterns of usual everyday life in a dementia unit from “anomalous” events that place people with dementia at risk. We will develop anomaly detection algorithms from multiple cameras placed in the unit and combine and compare these algorithms to those that have been developed using wearable devices. The ultimate goal is to unburden staff by teaching a computer to monitor the video stream and flag when an event of risk is happening.