PETRA (Pedal Elevated Temperature Risk Assessment) is a thermal-AI screening tool designed to predict diabetic foot ulcers 4–5 weeks before ulceration.
I co-founded and developed PETRA as part of my work in India and Kenya, combining biomedical engineering with global health design principles. PETRA uses a smartphone-mounted thermal camera to capture plantar temperature maps, applies calibration-free thermal asymmetry algorithms, and detects clinically significant hotspots (≥2.2°C) while restricting noise outside foot boundaries.
I built the data pipeline, hotspot detection model, Grad-CAM explainability layer, and worked with clinicians at MV Diabetes and AMTZ to validate the workflow. PETRA is currently moving toward expanded clinical testing, publication, and regulatory alignment as a low-cost, high-sensitivity triage tool for primary care settings.
The vision: prevent avoidable amputations by giving frontline providers an early-warning system that works anywhere.
