ABLE Music and Family
ABLE is a cross-faculty, interdisciplinary research project, located in Pulse Lab (Humanities), developing an interactive platform transforming low-level activity into an art experience. ABLE Music uses wearable sensors to engage older adults with dementia and caregivers in interactive play and art creation. Designed in response to the COVID-19 pandemic, ABLE Family uses an interactive web-based platform to provide collaborative gaming and artmaking opportunities for families in different living spaces.
Monitoring My Mobility (MacM3)
MacM3 is a Cross-faculty, interdisciplinary research project at McMaster that measures diverse mobility data of older adults to build a mobility monitoring tool to assist older adults and stakeholders to accurately mobility assets and needs.
#Caremongering
#Caremongering is a cross-faculty, interdisciplinary research project that seeks to understand how to adapt the #Caremongering solution to support older adults impacted by COVID-19 and facing social isolation.
MobilityAI (Elder people basic mobility analysis with AI)
Older adults hospitalized for acute medical problems are at risk of significant functional and mobility decline from restricted physical activity (e.g., extensive bed rest) during hospitalization. Many do not return to their pre-hospitalization level leading to an increase in the length of hospital stay (LOS), hospital readmission rates, and the need for post-discharge institutionalized care. The elder people basic mobility analysis with artificial intelligence (MobilityAI) project is a wearable sensor technology-based mobility assessment solution to in-hospital older adults. With a body attached accelerometer, the AI model recognizes activities such as sitting, standing, sleeping, stand up and go, and 10-meter walk. The project is currently in its first stage.
Smart Home
The Westdale Smart Home project aims to help senior citizens live independently by using sensing devices to monitor their movement and health in a renovated house. The technology detects anomalies in behavior and health and alerts caregivers or medical professionals to intervene before the situation worsens. The project also monitors medication usage and detects early symptoms of diseases for early treatment. The project uses information technology, wireless communication, web-based technologies, and autonomics to develop cost-effective solutions for senior citizens’ health and wellness and has the potential to revolutionize how we care for them.
Human Activity Recognition and Assessment
The use of inertial measurement unit (IMU) sensors has become increasingly important in monitoring the mobility status of elderly populations, particularly in Canada, which is facing a significant increase in senior population. A team of experts in deep learning and kinesiology have proposed a framework that combines machine learning with IMU and motion capture data to estimate three-dimensional knee joint kinematics with high accuracy and low cost. The approach aims to enable primary care providers to accurately assess mobility status and reduce the burden of musculoskeletal disorders on the national healthcare system. The team plans to incorporate meta-learning and multi-task learning to improve the performance of the machine learning model and simulate IMU data from motion capture and video data for data shortage limitations. This project has the potential to lay a foundation for a comprehensive and low-cost biomechanical data estimation framework.
Annotating Virtual Tai Chi Instruction to Improve Learning Outcomes for Older Adults
A focus group study was conducted to create and evaluate annotations for improving online Tai Chi classes. They used modern computer vision techniques to design visual annotations that can be added to a Tai Chi instructor’s video during live classes to enhance participant understanding and engagement, especially for older adults. The focus group consisted of older adults with varying online Tai Chi experience and a Tai Chi instructor who contributed their feedback and expertise to the annotation designs. The study provided valuable guidelines for future annotation designs.
FuSeARTHRO – Functional Severity Assessments for Joint Arthroplasties
This research project aims to gather functional movement data using cameras and wearable sensors from patients with knee and hip osteoarthritis who are seeking joint replacement surgery at St. Joseph’s Healthcare Hamilton orthopaedic clinic. The data collected will be used to create a more comprehensive profile of the functional severity of patients before and after surgery. The project aims to improve understanding of the changes in functional severity and to determine whether these changes are linked to surgical outcomes or treatment success. State-of-the-art technologies such as motion capture systems and wearable sensors will be utilized in this research.
Cognition, frailty, and falls: Should they be part of fracture risk assessment?
The study aimed to find out if there was an association between cognition, frailty, falls, and self-reported incident fractures in mid- to older adults in Canada. The study analyzed data from participants in the Canadian Longitudinal Study on Aging who completed the baseline assessment and three-year follow-up questionnaires. The results showed that higher frailty index scores and a history of falls were associated with an increased risk of incident fractures in all participants. However, falls were not associated with incident fractures in those aged 65 or older. There were no interactions or multicollinearity between cognition, frailty, and falls. The study’s findings should be interpreted with caution as the cohort was relatively young, and individual older than 65 years may react differently.
Benefits of Additive Noise in Composing Classes with Bounded Capacity
Machine learning systems used to monitor elderly individuals’ health often rely on time-series analysis, but it’s unclear if such models will perform consistently across different subjects. Researchers developed a new theory to determine the number of training samples needed to guarantee model performance. The key is adding a bit of noise, which prevents the network from “memorizing” and becoming too rigid. The team’s experiments showed this approach achieved tighter bounds than existing methods.