Detecting Behaviours of Risk in Nursing Homes using Deep Learning
Project Overview
Many people with dementia (PwD) live in nursing homes, which often suffer from a lack of skilled staff to support and monitor the safety of residents. Behavioural and psychological symptoms of dementia are commonly seen as a result of this lack of support and can be associated with risk for residents and staff. Many nursing homes have surveillance cameras for the safety of residents and staff, which are usually unmonitored. In this project, we propose to develop deep learning algorithms that will allow a computer to monitor surveillance video streams for events of risk. In a previous project, we collected data from 15 video cameras at a Specialized Dementia Unit. These videos are annotated for agitation events and represent a rich source of data about events of risk occurring in the nursing homes. Using this existing dataset, we will develop an algorithm capable of detecting “anomalous events,” such as agitation or other risky behaviours, to help staff respond and intervene quickly.
Principal Investigator
Shehroz Khan , University Health Network
Partners and Donors
Alzheimer's Association