York University is advancing predictive healthcare with two AI projects backed by almost CA$2 million from the Canadian Institutes of Health Research. Both studies aim to reduce complications for liver-transplant recipients and build on GraftIQ, a hybrid neural network recently published in Nature Communications.
GraftIQ combines clinician input with machine learning to spot six major causes of graft injury without a biopsy. External validations at the Mayo Clinic, Singapore’s NUHS and Hannover Medical School produced an AUC of 0.902, outperforming traditional models.
Why predictive AI matters in transplant care
Liver-transplant patients face risks such as fibrosis, graft failure and cirrhosis, which today are often diagnosed by invasive biopsy. York’s Team Liver AI project merges clinical records with social-determinant data, which includes income and geography, to create fairer risk models. A second study, DynaGraft, will predict graft fibrosis by fusing imaging, pathology and patient history.
York and the University Health Network manage the work with Canadian transplant centres. Dr Mamatha Bhat, a UHN hepatologist and co-lead, says the model enhances clinical judgement and speeds treatment. Each project aims to cut complications and improve access for underserved communities.
Next steps for York’s liver-AI team
Over the five-year grant period, researchers will refine the models and embed them in live clinical workflows. Goals include national deployment across transplant centres and new protocols that replace biopsy with non-invasive monitoring. Success would place Canada among the leading centres for responsible, predictive medical AI and could set a blueprint for other organ-care applications.
