Develops weakly supervised and physics-informed AI frameworks for anomaly segmentation and biomarker extraction across multi-modal medical imaging. Applications include diabetic retinopathy, stroke, intracerebral haemorrhage, lacunas, and microbleeds. The work integrates lesion simulation, domain adaptation, federated learning, and explainable AI to enable robust precision diagnostics in low-data clinical settings.
Faculty: Vanathi Sundaresan
Click image to view enlarged version
