Analytics Initiatives: Cohort 3

OBI's Centre for Analytics (CfA) supports the development of cutting-edge analytical tools that accelerate brain health research and improve care. Cohort 3 brings together eight projects spanning neurodevelopmental conditions, concussion, epilepsy, neurodegeneration, and mental health. Leveraging OBI's secure computing infrastructure — including the NeuroFL federated learning platform and Brain-CODE data assets — these teams are building the next generation of AI-powered solutions that protect patient privacy while driving meaningful discoveries across the brain health ecosystem.

Multi-modal Federated Learning for Precision Care in Neurodevelopmental Conditions

Dr. Azadeh Kushki of Holland Bloorview Kids Rehabilitation Hospital and the University of Toronto is developing a privacy-preserving federated learning pipeline to identify unique biological subgroups in children with brain-based differences and disabilities. By analyzing complex data without centralizing sensitive information, this scalable tool moves beyond "one-size-fits-all" care to provide more personalized support for children and youth.

Predicting Concussion Recovery through Multimodal Neuroinformatics

Co-funded with The Branch Out Foundation

Dr. Michael Cusimano of St. Michael's Hospital and the University of Toronto is leveraging the Brain-CODE CONNECT concussion dataset to build predictive models that generate individualized recovery risk profiles for concussion patients. These tools aim to reduce diagnostic uncertainty and guide safer, more targeted return-to-work and return-to-learn decisions, ensuring rehabilitation resources are allocated where they are needed most.

Federated Learning Testbed for Cognitive Health Prediction in Neurodegenerative Disease

Dr. Erin Dickie of the Centre for Addiction and Mental Health (CAMH) and Dr. Jean-Baptiste Poline of McGill University are building a secure federated learning testbed to benchmark AI models for neurodegenerative diseases like Alzheimer's and Parkinson's. This project evaluates how privacy-preserving tools can scale across large neuroimaging datasets to ensure AI remains accurate and reliable in real-world clinical settings.

Quantifying Multiple Mental Health Conditions in Neurodevelopmental Disorders (Q-MiND)

Dr. Clement Ma of the Centre for Addiction and Mental Health and the University of Toronto is developing a new scoring tool to more accurately measure and predict the impact of co-occurring mental health conditions, such as anxiety and depression, in youth with neurodevelopmental disorders. By focusing on severity and functional impact rather than just a diagnosis count, this tool helps clinicians make more precise, evidence-based treatment decisions tailored to each young person's needs.

A Standardized Neuroanalytics Platform for Multi-Modal Wearable Data

Dr. Benicio Frey of McMaster University is developing a cloud-based pipeline to standardize data from wearable devices, such as body temperature and light exposure, to better understand the link between "body clocks" and mental health. This user-friendly platform addresses a top 2025 research priority by providing researchers and clinicians with validated, comparable metrics to measure circadian disruption across different brain health conditions.

Multi-Modal Biomarkers of Epilepsy: EEG Synchrony Signatures and Genetic Predictors of Drug Response

Mark Aquilino and the team at Epiloid Biotechnology Inc. are building a machine learning framework that combines brain activity (EEG) and genetic data to predict how epilepsy patients will respond to specific treatments. By identifying unique biological markers of drug response, this tool aims to eliminate the "trial-and-error" approach to epilepsy care, helping clinicians prescribe the most effective therapies sooner.

AI-Powered FLAIR MRI Biomarkers for Early Detection of Neurodegeneration

Dr. April Khademi, Canada Research Chair in AI for Medical Imaging at Toronto Metropolitan University and lead of the Image Analysis in Medicine Lab (IAMLAB), is developing AI-powered tools that detect early signs of neurodegeneration and vascular disease from brain scans before symptoms even appear. By ensuring scan results are reliable and comparable across different types of imaging equipment, her work provides clinicians with actionable data from routine scans to inform earlier treatment and faster care.

Generation and Optimization of Synthetic Methylation Data for Brain Disorder Classification

Paul Wambe and the team at EpiSign Inc. are developing an open-source tool to generate synthetic biological data, allowing AI models to be trained for rare disease diagnosis without using actual patient records. By bypassing traditional privacy barriers and data scarcity, this approach aims to create a practical roadmap for detecting brain disorders through a simple blood test.