Archives for March 14, 2024

by Sara Elhawash

Wearable devices have long been praised as the future of healthcare, fitness tracking, and athletics. However, the reliability of these devices has been hampered by the presence of motion artifacts, which greatly diminish the quality of data collected. 

The Data Sciences Institute (DSI) and the Tanenbaum Institute for Science in Sport (TISS) at the University of Toronto are pleased to award a Catalyst Grant for research in the Development of Convolutional Neural Network for Motion Artifact Mitigation in Wearable PPG Devices. Co-Led by Professors Daniel Franklin (Institute of Biomedical Engineering, Temerity Faculty of Medicine, University of Toronto) and Chris McIntosh (Department of Medical Biophysics, Temerity Faculty of Medicine and University Health Network), the project aims to revolutionize athletics and sports medicine by integrating novel sensors with advanced machine learning algorithms. 

The researchers propose a novel approach to overcome motion artifacts in wearable devices by enabling real-time motion artifact cancellation in optical wearables. This includes the development of a multimodal sensor coupled with deep learning models. The sensor will combine force and multiwavelength optical measurements to capture relative motion at the sensor interface, addressing a critical limitation of current wearable devices. 

“Conventional wearable devices capture global motion, but our approach focuses on capturing relative motion at the sensor-skin interface, which is crucial for accurate data interpretation. Wearable technologies offer a unique glimpse into patient function and biology outside of episodes of care. If AI is the present, wearables with AI are the next frontier,” says Professor Daniel Franklin. 
 
The research project will progress through several phases, including controlled lab experiments and real-world examples of motion. By collecting a novel multi-modal motion artifact dataset, the team aims to develop a robust algorithm for real-time optical motion artifact cancellation. 

“We are thrilled to partner with the DSI to award this seed grant. This project has the potential to significantly advance wearable health monitoring technologies for applications to the healthcare, fitness, and sports sectors,” says Dr. Ira Jacobs, Director of the Tanenbaum Institute for Science in Sport. 

The implications of this research extend beyond healthcare into consumer health, sports, and athletics. By enhancing the usability and interpretability of wearable device datasets, the project promises to advance remote health management and athletic performance tracking. 

“We envision a future where wearable devices provide more accurate and actionable insights, leading to improved patient care and athletic performance,” added Professor McIntosh. 

The DSI and TISS partner to co-sponsor Catalyst Grants focused on innovative and novel data science in sport and sport analytics.  

by Sara Elhawash

The Data Sciences Institute (DSI) and the Temerty Centre for AI Research and Education in Medicine (T-CAIREM) at the University of Toronto join efforts for the second consecutive year to co-fund two 2024 Catalyst Grant Awards focused on innovative and novel data science methodologies in medicine and health. 

Each catalyst grant provides up to $100,000 in seed funding for multidisciplinary researchers forming Collaborative Research Teams (CRTs) that are developing novel statistical or computational tools that address important societal needs. 

“These jointly funded catalyst grants are directed at highly innovative initiatives that have the potential to transform healthcare with data science, and this year’s winners are no exception,” says Muhammad Mamdani, executive director T-CAIREM. “It’s rewarding to see initiatives that not only focus on specific segments of our population to improve their quality of life but also those that have far reaching implications for society at large.” 

Examining biases due to confounders and colliders in observational health data using individual-based simulation models

Sharmistha Mishra (St. Michael’s Hospital, Unity Health Toronto and Rafal Kustra (Dalla Lana School of Public Health, University of Toronto)  

In the realm of medical research, observational studies leveraging large health-administrative datasets are crucial. However, bias in the data, including residual confounding and collider bias, can produce misleading results, potentially skewing policy decisions, resource allocation, and clinical management.  

This research aims to enhance public health outcomes during infectious disease outbreaks by employing simulation modeling combined with causal inference and statistical learning methods to identify and address different types of biases that could undermine inference drawn studies of health using observational data. 

Specifically, the researchers plan to generate synthetic datasets using simulation models that replicate the complex dynamics of the 2022 Mpox outbreak in Toronto, in collaboration with clinicians, public health teams, and community-based organizations. They intend to use statistical learning methods to predict how big the problem of residual confounding and collider biases could get when inferring risk factors and the effectiveness of interventions during an evolving outbreak. They will then pilot-test analytic approaches to reduce these biases. 

“This work has the potential for applicability across health conditions by helping to improve validity in estimating risks and intervention impact,” says Professor Mishra. 

Decoding unintelligible speech: a conversational context-aware assistive technology for children with complex communication needs (CCN) 

Tom Chau (Holland Bloorview Kids Rehabilitation Hospital) and Monika Molnar (Temerty Faculty of Medicine, Department of Speech-Language Pathology, University of Toronto)  

Children with CCN often prefer to vocalize, but their sounds are typically unintelligible to those unfamiliar with them. They are often excluded from fully participating in education, society, and eventual employment. 

This Catalyst Grant proposal is dedicated to helping children with complex communication needs (CCN), potentially leading to the development of assistive devices. 

The research team plan to utilize machine learning to decode the unintelligible speech of these children using an existing audio-video dataset of speech samples. This project could pave the way for the development of artificial intelligence-driven electronic devices tailored for children with CCN. 

“There are currently no assistive technologies that can accurately decode their speech sounds,” explains Professor Chau. “As a result, children with CCN remain excluded from full participation in education, society, and eventual employment.” 

The researchers hope this project will accelerate the impact of data sciences in the fields of rehabilitation and biomedical engineering, driving positive social change for children with CCN. 

The DSI’s Catalyst Grants, co-funded by T-CAIREM, play a crucial role in supporting these research projects by providing the essential seed funding and fostering the collaboration among research teams needed to realize this impactful work and apply for external funding in the future. 

by Sara Elhawash

The Data Sciences Institute (DSI) is pleased to announce the 2024 recipients of the annual DSI Catalyst Grant competition. Fourteen interdisciplinary teams across all three campuses and external funding partners received grants, for research that focusses on harnessing the transformative nature of data sciences.  

Catalyst Grants are awarded to teams working on the development of novel statistical or computational tools, as well as the use of existing methodology in innovative ways to address questions of major societal importance and effect positive social change. The intent is for the grants to serve as seed funds that bring cross-institutional multidisciplinary teams together including data science leadership, to innovate in traditional disciplines and position the teams for external competitive research funds.  

“The 2024 DSI Catalyst Grant recipients exemplify our commitment to multidisciplinary collaboration, uniting researchers to tackle pressing societal issues. This year’s projects promise innovative solutions and showcase the collective expertise driving positive change,” says Gary Bader, Associate Director, Data Sciences Institute. 

The DSI-funded research spans disciplinary areas and includes collaborations tackling critical issues in urban road safety and forest management amidst escalating wildfire risks (see full list below). This year, several Catalyst Grants are co-funded by the Temerty Centre for AI Research and Education in Medicine (T-CAIREM) with a focus on innovative and novel data science methodologies in medicine and health and the Tanenbaum Institute for Science in Sport (TISS) on innovative and novel data science in sport and sport analytics. 
 
Eye on the street: Using computer vision to capture the determinants of road safety 

In urban road safety research, comprehensive datasets detailing road network modifications are essential for evaluating intervention effectiveness and informing evidence-based policy decisions. 

With their project, Professors Brice Batomen (Dalla Lana School of Public Health) and Marianne Hatzopoulou (Department of Civil and Mineral Engineering, Faculty of Applied Science and Engineering) aim to address the critical public health issue posed by traffic collisions, which are a leading cause of premature death. 

They begin by compiling detailed information on road modifications in Canadian cities, starting with Toronto and Montreal, with the ultimate goal of promoting safer urban environments. 

“This research aims to impact public and environmental health by analyzing the effectiveness of road safety interventions,” says Batomen. By employing advanced causal inference methods and creating comprehensive datasets, the project aims to inform policy-making, reduce traffic-related fatalities and injuries, and foster safer, more equitable urban environments. 

“Through interdisciplinary collaboration facilitated by the DSI, the project brings together epidemiologists, computer scientists, and transportation engineers, laying the groundwork for impactful research with broader implications,” says Batomen. 

Effect of forest management on insurable wildfire risk in Northern Ontario 

Amidst the escalating frequency and severity of wildfires, particularly impacting regions like Northern Ontario, the intersection of forest management and wildfire risk assessment emerges as a critical focal point for research and policy intervention. 

Professors Rasoul Yousefpour (John H. Daniels Faculty of Architecture, Landscape, and Design) and Silvana Pesenti (Department of Statistical Sciences, Faculty of Arts and Science) flag that “Forest fires are occurring at an alarming rate, posing a significant challenge to the insurability of affected landscapes in Ontario, including indigenous communities.”  

Their research endeavors to unravel the intricate connections between forest management practices and wildfire risk assessment, essential for informing policy decisions and fostering equitable wildfire insurance mechanisms. “The DSI Catalyst grant represents precisely the resource required to pioneer innovative big data-driven technologies and models aimed at unraveling the impact of forest management on the insurability of forest fires in Ontario,” say Yousefpour and Pesenti. 

Over the two-year funding period, the grant will empower the recruitment of graduate students who will collaboratively establish connections between forest and fire data using advanced data science methodologies.  

“The findings of this research will not only inform forest management best practices but also raise awareness and contribute to the establishment of equitable wildfire insurance mechanisms for all citizens, including First Nations communities,” says Yousefpour and Pesenti.  

They envision the integration of cutting-edge technology to disseminate across both fields of study, providing guidance for future research and policy analysis in fire-prone forest landscapes.  

Congratulations to all the 2024 DSI Catalyst Grant collaborative research teams!  

Coronavirus in the Urban Built Environment (CUBE) 

• Michael Fralick (Department of Medicine, Temerty Faculty of Medicine, University of Toronto) David Guttman (Department of Cell and Systems Biology, Faculty of Arts and Science, University of Toronto) 

Decoding unintelligible speech: a conversational context-aware assistive technology for children with complex communication needs 

• Project co-funded by T-CAIREM  
• Tom Chau (Holland Bloorview Kids Rehabilitation Hospital) and Monika Molnar (Department of Speech-Language Pathology, Temerty Faculty of Medicine, University of Toronto) 

Developing Algorithms & Statistical Analysis Techniques for Adaptive Experimentation 

• Joseph Williams (Department of Computer Science, Faculty of Arts and Science, University of Toronto), Felix Cheung (Department of Psychology, Faculty of Arts and Science, University of Toronto), Anna Heath (The Hospital for Sick Children) and Michael Liut (Department of Mathematical and Computational Sciences, University of Toronto Mississauga) 

Development of Convolutional Neural Network for Motion Artifact Mitigation in Wearable PPG Devices 

• Project co-funded by Tanenbaum Institute for Science in Sport (TISS)  
• Daniel Franklin (Institute of Biomedical Engineering, Faculty of Applied Science and Engineering, University of Toronto) and Chris McIntosh (University Health Network, Toronto General Hospital Research Institute) 

Effect of forest management on insurable wildfire risk in Northern Ontario 

• Silvana Pesenti (Department of Statistical Sciences, Faculty of Arts and Science, University of Toronto) and Rasoul Yousefpour (John H. Daniels Faculty of Architecture, Landscape, and Design, University of Toronto) 

Enhancing the Reliability of Large Language Models for Structured Data Extraction in Chemical Sciences 

• Seyed Mohamad Moosavi (Department of Chemical Engineering and Applied Chemistry, Faculty of Applied Science and Engineering, University of Toronto) and David Sinton (Department of Mechanical and Industrial Engineering, Faculty of Applied Science and Engineering, University of Toronto)

Examining biases due to confounders and colliders in observational health data using individual-based simulation models 

• Project co-funded by T-CAIREM  
• Sharmistha Mishra (St. Michael’s Hospital, Unity Health Toronto and Rafal Kustra (Dalla Lana School of Public Health, University of Toronto) 

Eye on the street: Using computer vision to capture the determinants of road safety 

• Brice Batomen Kuimi (Dalla Lana School of Public Health, University of Toronto) and Marianne Hatzopoulou (Department of Civil and Mineral Engineering, Faculty of Applied Science and Engineering, University of Toronto)  

Interpretable and fair machine learning for equitable assessment of patient safety in hospitals 

• Eldan Cohen (Department of Mechanical and Industrial Engineering, Faculty of Applied Science and Engineering, University of Toronto), Sheila McIlraith (Department of Computer Science, Faculty of Arts and Science, University of Toronto), Amol Verma (St. Michael’s Hospital, Unity Health Toronto) and Fahad Razak (St. Michael’s Hospital, Unity Health Toronto) 

Investigating the biological function of the m6A epitranscriptome using Oxford Nanopore direct RNA sequencing 

• Ina Anreiter (Department of Biological Sciences, University of Toronto Scarborough) and Jared Simpson (Ontario Institute for Cancer Research) 

Scaling up highly multiplexed imaging with compressed sensing 

• Kieran Campbell (Lunenfeld-Tanenbaum Research Institute) and Hartland Jackson (Lunenfeld-Tanenbaum Research Institute) 

Toolkit for Improved Climate Hazard and Risk Assessment in Ontario 

• Robert Soden (Department of Computer Science, Faculty of Arts and Science, University of Toronto) and Paul Kushner (Department of Computer Science, Faculty of Arts and Science, University of Toronto). 

Using generative models to “fix” missing structures and artifacts in MRI images 

• Evdokia Anagnostou (Holland Bloorview Kids Rehabilitation Hospital) and David Duvenaud (Department of Computer Science, Faculty of Arts and Science, University of Toronto)