Five Amii Fellows were awarded research grants as part of CIFAR’s AI and COVID Catalyst Grants initiative. The CIFAR AI Catalyst Grants Program is intended to catalyze new research areas and collaborations in machine learning, providing funding for innovative, high-risk, high-reward ideas and projects.
Amii researchers will collaborate across four projects, each in a distinct area of pandemic management: enabling drug discovery, creating a virtual data lab, detecting and monitoring illness, and tracking mental health.
Learn more about the projects below. You can also find out more about how Amii is lending our expertise in the global fight against COVID-19 at amii.ca/covid-19
Tracking Mental Health During the Coronavirus Pandemic
Alona Fyshe, Amii Fellow and Canada CIFAR AI Chair, has joined with a team of researchers from Western University and NYU in order to leverage machine learning and social media to better understand the drivers of mental health, both acute and long-term, as well as mitigating factors. With a recent uptick in social media use, the expanding discourse provides a view into the experiences of people that is not available by other means, especially for marginalized groups (e.g. people with limited access to health care, lower socioeconomic status, and undocumented immigrants).
With findings disseminated through an online visual analytics and rapid reporting system, the team will develop AI techniques for social media data to understand: 1) emerging challenges that affect people during the pandemic; and 2) how these challenges impact mental health.
Collaborators include Alona Fyshe (Canada CIFAR AI Chair, CIFAR Learning in Machines & Brains program, Amii, University of Alberta), Daniel Lizotte (Western University), Rumi Chunara, Brent Davis.
Learn more: amii.ca/mental-health-coronavirus
Accelerating Small Molecule Drug Discovery
Amii Fellow Matthew E. Taylor joins a collaborative team of researchers from Mila and 99andBeyond (a company that develops technological platforms for drug discovery) to accelerate the discovery of safe and effective small-molecule treatments against COVID-19 and to mitigate other future outbreaks. The team will apply recent advancements in machine learning to repurpose small molecules with proven safety (phase I trials) and identify novel candidates as anti-COVID-19 therapeutics.
The project seeks to advance the field of reinforcement learning in small molecule drug discovery, retain and monetize made-in-Canada IP, and most crucially, openly publish anti-COVID-19 candidates associated with promising in-vitro data for the benefit of the research community.
Collaborators include Sarath Chandar (Canada CIFAR AI Chair, Mila, Polytechnique Montréal), Matthew E. Taylor (Amii, University of Alberta), Sai Krishna (99andBeyond), Karam Thomas (99andBeyond).
Guarding At-Risk Demographics with AI (GuARD-AI)
GuARD-AI brings together Amii Fellows Randy Goebel and Martha White (also a Canada CIFAR AI Chair) with health informatics researchers, health information leaders and data ethics researchers to prototype a virtual data laboratory. The project aims to identify at-risk populations, predict disease course at an individual level, predict disease spread across an entire health system, and analyze insights gained to refine future use of virtual healthcare delivery models in crisis scenarios.
The project is working to develop best practices for health analytics in situations where time is of the essence and action-based decisions can be supported by extracting value from highly dynamic and time-sensitive data. GuARD-AI will help with immediate pandemic-related challenges facing Alberta and Canada and contribute to future adaptable systems for reacting to time-sensitive outbreaks.
Collaborators include Daniel C. Baumgart (University of Alberta), Geoffrey Rockwell (University of Alberta), Martha White (Canada CIFAR AI Chair, University of Alberta, Amii), Randy Goebel (Amii, University of Alberta), Robert Hayward (Chief Medical Information Officer, Alberta Health Services), Shy Amlani, (Virtual Health), Jonathan Choy (Virtual Health), Sara Webster (Virtual Health), and Sarah Hall (Virtual Health).
Detecting and Monitoring Pneumonia in COVID-19 Patients
Severe illness and death in COVID-19 patients is most often due to progression of the disease to an interstitial pneumonia resembling acute respiratory distress syndrome (ARDS), which requires hospitalization. Amii Fellow Russ Greiner joined with MEDO.ai, a machine learning diagnostics company, and health experts from New York state to produce a diagnostic tool through applying machine learning to ultrasound scans to automatically determine which patients have pneumonia. Ultrasound, rather than computerized tomography (CT), is proposed due to the portability and of the technology (reducing transmission risk) and because it does not carry the same risk of exposure to radiation.
Detecting changes to a patient’s condition before the patient requires emergent intervention may allow for the provision of supportive care earlier and more effectively. The final system, which researchers anticipate will outperform the average human reader (and potentially even exceed the performance of experts) will be integrated into ultrasound scanners to produce a tool that can be used effectively by a healthcare worker – even one with limited training.
Collaborators include Kumaradevan Punithakumar (University of Alberta), Russell Greiner (University of Alberta, Amii), Jacob Jaremko (University of Alberta), Nathaniel Meuser-Herr (Upstate Health Care Center, NY), Dornoosh Zonoobi (MEDO.ai).