Discovering gene-disease relationships with deep learning & Using AI for intracranial hemorrhages

T-CAIREM Trainee Rounds
PRESENTERS: Anastasia Razdaibiedina and Michael Balas
DATE: August 10, 2021 (Tuesday)
TIME: 12pm to 1pm
VENUE: Zoom
_________________________________
Anastasia Razdaibiedina
PhD student, Computational Biology and Machine Learning, University of Toronto

PRESENTATION: Discovering gene-disease relationships with deep learning

Understanding the genetic causes of diseases is one of the central goals in medicine. Most diseases have a complex genetic basis, and genes often act in ‘modules’ to determine phenotypes. An effective way to discover a module of disease-associated genes, is to use biological networks, or interactomes, that describe interactions between genes and proteins. Here we use deep learning methods to infer an interactome computationally from microscopy imaging data, and subsequently discover gene-disease relationships from the constructed interactome.
_________________________________
Michael Balas
Medical student, Temerty Faculty of Medicine, University of Toronto

PRESENTATION: Using Artificial Intelligence to Identify Intracranial Hemorrhage and Predict Patient Outcomes

Intracranial hemorrhage (ICH), or bleeding in the skull, is one of the most frequently encountered emergencies and represents an important triaging task in the neurosurgical workflow. We used deep learning algorithms trained on hundreds of thousands of CT brain scans to automatically detect ICH, and we achieved near-human level accuracies. Furthermore, integrating our AI models into risk assessment tools yielded accurate predictions of mortality. Soon, this work will be deployed into automated clinical decision support systems in ICH management.