Seminar Dr.Shehroz Khan “Multimodal Sensors in the Wild: Case Studies from Dementia Care and Post-hip Surgery”

Abstract

Wearable sensors provide a great opportunity to detect physical and physiological indicators of human body, including motion, heart rate, electro-dermal activity, muscle movements, and EEG. These indicators are of great value to clinicians and people who are using them to detect adverse events (e.g. falls), monitoring their progress (e.g. mobility), or changes in health conditions. A major concern with wearable devices is the user compliance in wearing them for intended times to collect meaningful data. If researchers are interested in measuring more than one vital indicator, then traditionally speaking, more than one wearable devices may be needed. In long term monitoring studies in the wild, this type of approach is unrealistic and may not fetch desired results. A plausible solution to this problem is the use of a wearable device that can combine multiple sensors in one gadget or use wearable device with single/multiple sensors along with other sensors, such as ambient sensors and computer vision. Besides detecting different types of vital signs, multi-modal sensors could also lead to more robust classifiers for a specific problem. In this talk, I will discuss two studies – one completed and one up-coming, which involve deploying multi-modal sensors in the wild for dementia care and post-hip surgery patients.

People living with dementia (PwD) often exhibit behavioral and psychological symptoms of dementia, with agitation and aggression being the most notable. Agitated PwD can harm themselves, other residents, and staff in the long-term care. These long-term care centers are often under-staffed and such incidents prohibit them to focus on care of other residents. The traditional method to assess agitation is to use clinical scoring methods; however, those methods are retrospective and cannot detect or predict agitation events. We conducted a study at the Specialized Dementia Unit at Toronto Rehabilitation Institute and collected more than 600 days’ worth data from 20 PwD using Empatica E4 watch that collects body acceleration, blood volume pulse, electro-dermal activity, and skin temperature. The predictive results give strong evidence that multi-modal sensing approach performs significantly better than single sensors in detecting agitation in PwD.

Patients normally get great care after post-hip surgery in a rehabilitation hospital. However, once they are discharged, it is very hard to track their progress. Social isolation and physical mobility are major risk factors for cognitive decline in older adults (OAs) post-hip fracture. The primary aim of this project is to develop a clinically validated multi-modal sensor system to assess changes in social isolation and physical function in OAs following discharge from inpatient rehabilitation. We are currently recruiting a sample of OA patients with hip fracture and will install a suite of sensors in their homes to collect various types of data (e.g. biometric, mobility, sleep quality) over 2-months. A trained research assistant will conduct clinical assessments of social isolation and physical mobility every 2 weeks during this period. This project is the first step to building a scalable sensor system that can collect and assess social and physical aspects of OAs in the community, which will ultimately support cognitive health and aging-in-place.

The lessons learnt from the dementia care project led to several innovations in terms of eliminating humans from data collection process and streaming data directly to the cloud, development of a novel software suite to handle multi-modal sensors and providing maintenance and feedback to researchers / users in real time.

Biography

Dr, Shehroz Khan is a Scientist in the Artificial Intelligence and Robotics in Rehab Lab at the KITE, Toronto Rehabilitation Institute, University Health Network, Canada. He is also cross appointed as an Assistant Professor at the Institute of Biomedical Engineering, University of Toronto, Canada. He holds a PhD Degree from the University of Waterloo, Canada in Computer Science with specialization in Machine Learning. Dr. Khan’s main research focus is the development of zero-effort machine learning and deep learning algorithms within the realms of Aging, Rehabilitation and Intelligent Assisted Living (ARIAL). As a Principal Investigator (PI) and Co-PI, his research program has been funded through NSERC, CIHR, AGEWELL NCE, AMS Healthcare, SSHRC, CABHI, and UAE University. Currently, he is developing a novel cloud-based multimodal sensor platform to assess various health indicators among the people living in the community. He has also started a new research program to facilitate telerehabilitation using advanced computer vision techniques to determine patient engagement, therapy compliance and their dropout rates. He has published 41 research papers in top international journals and conference that have garnered more than 2200 citations on Google scholar. He is the founder and organizer of the peer-reviewed International Workshop on Artificial Intelligence for ARIAL that is held in conjunction with top conferences in the field in 2017, 2018 and 2019.

Join Zoom Meeting

https://mcmaster.zoom.us/j/93692008212?pwd=dmVwMlpycUpyaytKeitrY0ExOHBZdz09

Meeting ID: 936 9200 8212

Passcode: 030988