At IIT Roorkee, I built a CNN-Transformer ECG biometric model and investigated whether stolen biometric templates could reveal heart disease. My work found diagnostic information leakage across four classes, with the strongest signal for conduction disturbance.
In my clinical risk prediction project, I combined patient biomarkers and MRI analysis to evaluate prostate cancer risk and identify opportunities to reduce unnecessary biopsies. I’ve also analyzed brain tumor MRI scans, assessing model reliability with Grad-CAM, Integrated Gradients, and MC-dropout.

