Projets novateurs réalisés

A feasibility and acceptability study of implementing a Collaborative service robot in long-term care

Service robots can offer personalized service for people with disabilities and empower the staff with efficient support. This study investigates using a collaborative service robot, Aether in a long-term care home. We work collaboratively with people living with disabilities, families, frontline staff, operation leaders, and industry) to Identify user experience, impact and challenges to inform future robot development and adoption. The study will be conducted through three phases: (1) Plan, (2) Adapt, and (3) Evaluate. We will recruit people with various disabilities, family members, and interdisciplinary staff. The focus will be on investigating the deployment process and lessons learned. The findings of this research will offer useful insights about user experiences, risk and mitigation strategies, and impact in LTC, in relation to adopting a robot (Aether) to support safety and quality of life in care homes.

To develop an AI model for predicting future lung cancer in low-dose screening CT

Screening with low-dose CT has been shown to significantly reduce lung cancer related mortality in high-risk ever-smokers. Interval cancer (IC) is a rising challenge in lung cancer screening because it usually presents in an advanced stage (stage III/IV non-small cell cancer) or is more biologically aggressive (i.e., small cell histology) and have a poorer prognosis than prevalent cancers. The dilemma is how to catch IC early because the regularly scheduled follow-up CT is often too late. We propose that artificial intelligence (AI) tools can identify sub-visual changes in the “normal” lung before a clinically detectable IC develops. We have access to three population-based screening datasets with ICs. Using the prior CT(s) before a diagnosed IC or benign nodule develops, we propose to build an AI algorithm that can distinguish a pre-IC “negative” CT from CTs of subjects that will remain negative. We will use this information to guide the follow-up interval for individual subjects. This AI tool has the potential to standardize the triage process, enable personalized follow-up intervals for high-risk individuals, detect IC earlier, and improve the outcome and cost-efficiency of lung cancer screening.

An Interactive Dashboard for Human-AI Detection of Anomalous Employee Accounts at Risk of Data Exfiltration

Confidential data is one of the most precious assets large organizations can have and data theft can be embarrassing and costly. In this research we will carry out innovative research on detecting employee accounts that are exhibiting risky behaviour that may lead to leakage of data, whether carelessly or through malicious intent. It is difficult to protect against careless actions of employees, or malicious people masquerading as employees. Machine Learning (ML) models have been used to make identification of anomalous data transfers more efficient. However, ML models are typically trained through supervised learning and require training with accurately assigned labels. In our past research with Sun Life we have focused on accurately labeled single events such as USB transfers and emails containing attachments and have shown that techniques such as visualization and Active Learning (AL) can be helpful. Active learning is an approach where human experts label the instances (e.g., email messages) that the machine finds hard to label. In this project we will take the next step of labelling employee accounts (rather than USB transfers, or emails) as potentially unusual (and possibly dangerous), based on more extensive analysis of the available data.

Managing tree models plasticity and mixing GLMs with regression trees for insurance ratemaking

Predicting policyholders’ claims over a year is crucial for a Property-Casualty insurance company. These expenditures, popularly called losses, are incurred by the insurer when reimbursing the policyholders’ claims. The insurance company is required to pay any legitimate claim made by a policyholder, in exchange the latter pays an amount of money, called the premium, to the company to buy this entitlement. Annual premium must be calculated with precision to ensure a fair deal on both sides.
It is the task of actuaries to set premiums for all policyholders; this is called ratemaking. Various classical statistical methods are used to set a policy premium rate, so maximize gain without losing existing customers to the competition. We explore here several data science techniques to help improve this ratemaking process.

Simulating the Use of Agriculture as a Nature-Based Carbon Capture Solution for Mitigating Climate Change

Unmitigated climate change is expected to have catastrophic impacts on our way of life, causing governments and corporations to step in to pledge to reduce our emissions. However, to prevent warming beyond 2°C above pre-industrial levels, emissions reductions will likely not be enough, so we will need to take carbon dioxide directly out of the atmosphere, such as through planting more forests or changing how we do agriculture. Using existing agricultural practices, such as adding the ash of leftover crops to the soil, we can account for a considerable amount of the carbon dioxide removal necessary to meet this goal. However, we know very little about how well this works, and the impacts it may have on the climate. In this study, we will answer these questions by performing numerical simulations of agricultural carbon dioxide removal under some likely future emissions and removal scenarios. We expect to find that it has a considerable effect on reducing the peak warming of the future climate. Furthermore, we will simulate how climate change affects agriculture itself. This research will be extremely beneficial to our understanding of climate change mitigation, and relationship between agriculture and the climate.

Development of new manufacturing processes for ionizable lipids and building blocks for same

In 2022, over 4.5 billion doses of the Pfizer-BioNTech and the Moderna COVID-19 vaccines were produced to combat a devastating pandemic. The above vaccines contain messenger RNA (“mRNA”) trapped inside lipid nanoparticles (“LNPs”), which are globules with a diameter of a few billionths of a meter, composed of fatty substances called lipids. A crucial type of lipid that is essential for the proper functioning of RNA-LNP medications is an “ionizable lipid” (IL): one that, once inside a living cell, can switch between an electrostatically neutral state and a charged one. The search for new, efficacious ILs is key to unlocking the potential of RNA therapies to treat a huge number of human diseases that currently are incurable. NanoVation Therapeutics, Inc. owns technology for the rapid preparation and evaluation of such ILs. This proposal aims to devise new practical, economical manufacturing methods for the building blocks of said ILs, as well as to produce quantities of benchmark ILs against which the efficacy of new ILs is evaluated. The methods thus developed are essential to unleash the full potential of RNA medicines well beyond the vaccine sphere.

Development and Evaluation of Antimicrobial Agents for Improving the Lifetime of Metalworking Fluid

Metalworking fluid is a widely used lubricant during the processing of metal (drilling, cutting, grinding, etc.). Even though it has been applied to industries for decades, microbial contamination is still considered a major problem during its application. The microbial contaminations not only reduce the effectiveness of the lubrication but also cause potential health issues. In this project, several eco-friendly biomaterials produced from low-cost agricultural- and food waste will be used for improving the resistance to microbial contamination. We expect it could extend the shelf-life of the in-use metalworking fluid as well as improve the biodegradability of the used fluid.

Using artificial intelligence to screen children suspected with listening difficulties

The goal of this project is to create a cutting-edge screening tool for children with listening difficulties, which can have a profound impact on their academic performance, social development, and overall well-being. Despite the current assessment process, which is time-consuming and requires extensive training and experience, early identification and intervention are critical for these children to reach their full potential.

Artificial Intelligence (AI) and Machine Learning (ML) algorithms have the potential to revolutionize healthcare by providing more information to inform diagnoses and improve diagnostic accuracy and efficiency. However, applying AI/ML to medical applications can be challenging due to the complexity and variability of medical data.

The project team has previous experience in developing fast and accurate infant ABR screening algorithms, which achieved 96% accuracy in detecting ABRs. In this project, the team will use deep learning approaches to create a fast and accurate AI-based screener for children with listening difficulties. The developed screener will be tested for accuracy, efficiency, and usefulness compared to the current assessment process. The work will be carried out in collaboration with Vivosonic Inc., a leading company in the field of hearing screening technologies.

By successfully completing this project, the team will contribute to filling the gap in research on developing efficient and reliable screening tools for children with listening difficulties and improving early identification and intervention. This project has the potential to improve the lives of many children by giving them a better chance to reach their full potential.

Non-destructive testing of utility poles in service

Wooden utility poles provide safe, economic, easily obtainable means of delivering power, communications, and cable television to the masses of industrial and residential locations worldwide. However, when a pole has been in service for a substantial number of years, its failure becomes more likely. Hence, the proposed project focuses on the strength estimation of in-service utility poles using non-destructive technology (NDT). The advantage of using NDT is that no further damage is caused to the poles during the investigation, which can further lead to reduced service life. The collected data will be statistically analyzed, and predictions will be made on the remaining service life of study area utility poles. The greatest economic benefit from regular inspection is in locating the decaying/serviceable group. Treating poles in this group can extend pole life, thereby saving emergency replacement costs. With the costs of replacing poles rising, the economics of extending service life is more favourable. To avoid costly failures of utility lines, many utility poles are condemned annually based on a precautionary basis. This represents a significant irresponsible waste of natural resources. Our industrial partner Stella-Jones Inc. is North America’s leading producer of industrial pressure-treated wood products and supplies utility poles to Canadian electric utilities companies. Keeping in view the environmental benefits of this study, Stella-Jones is happily supporting this proposed research. The outcomes will also help Stella-Jones expand the knowledge of the pole system, increase relative safety for the ones working on poles, and facilitate customer relations and quality perceptions.

Focused Ultrasound for Intraocular Hemmorhage

Intraocular hemorrhage is the most common cause of sudden vision loss from advanced diabetic eye disease. This occurs when blood from diseased retinal vasculature leak into the normally clear, gel-like substance in the eye called the vitreous, obscuring vision. The current management guidelines for these patients is a period of modified bedrest (up to 4 months) to encourage natural reabsorption. Apart from “watchful waiting”, the remaining treatment options are surgical. Focused ultrasound (FUS) represents a potential nonsurgical alternative to accelerate clearance of intraocular hemorrhage, and the present study aims to elucidate the biologic and mechanical effects of FUS through preclinical models of intraocular hemmorhage.

The FUS Lab at Sunnybrook Research Institute is 1 of 7 Centers of Excellence in FUS research worldwide, and Vitreosonic Inc. is excited to partner with this local academic powerhouse to assist in performing the rigorous scientific activities necessary to support the development of this sight-saving device.

Peptide-based material for heart muscle repair

With over 17 million deaths per year, heart diseases remain the top cause of mortality worldwide. Surgeries such as bypass restore blood supply and save lives. However, after a heart attack the capacity of the heart to pump blood is reduced. In Canada, over 2 million people aged 20+ live with heart disease, costing the healthcare system $2.8+ billion and thousands die each year. Approximately 1 in 4 of these patients develop heart failure, a number that increases by ~50,000/year. For many of those patients, a heart transplant is the only option. Thus, better treatments for repairing damaged hearts are urgent. In this project, we will develop a new generation of peptide-based materials to deliver stem cells to the damaged heart muscle and help rehabilitate the organ. This new therapeutic approach will be tested in small animals with infarcted hearts that will expedite advancing to first in human evaluation in the years to come.

Behavioral Model of Charge-Trap Transistors

Conventional von Neumann architectures rely heavily on communication between memory and compute elements,
making them power hungry. In recent years, therefore, neuromorphic computing based on low-power compute-inmemory
devices has been gaining interest. One of the essential aspects of such systems is the hardware modeling
of synapses that are expected to store weights that represent the strength of connections among neurons. Various
devices have been proposed as candidates for analog synapses. In this research, charge-trap transistors (CTTs)
that support the non-volatile analog adjustment of synaptic weights, are investigated. Accurately modeling the
physical phenomena of CTTs is critical to the design of CTT-based neuromorphic systems. A model of the weight
adjustment of CTTs will be developed and verified using experimental data. This model will enable efficient design
of CTT-based synaptic arrays, and the ability to simulate complete neuromorphic systems.