Digital Organ Twins from Single Imaging Scans: A Hybrid Physics-ML Framework for Predictive Medicine

Mahule Roy, Subhas Roy

Keywords: Digital twins, Medical imaging, Disease progression prediction, Therapeutic response modeling, Personalized medicine, Clinical decision support

TL;DR: We present OrganTwins, a framework that creates dynamic digital twins of human organs from a single clinical scan to predict disease progression and treatment response using deep learning and physics-based modeling.

Abstract: Medical imaging remains fundamentally limited to static anatomical assessment, lacking predictive capabilities for disease progression and treatment response. We introduce OrganTwins, a novel framework that constructs dynamic digital twins of human organs from single, routine clinical scans (MRI, CT, ultrasound). Unlike existing methods requiring longitudinal data, our approach synergistically integrates deep learning with physics-based modeling to simulate both structural and functional evolution over time, enabling virtual testing of therapeutic interventions through counterfactual analysis. Comprehensive validation across cardiac and hepatic datasets demonstrates consistent performance improvements of 14–22% in predictive accuracy compared to conventional baselines, establishing the framework’s potential for personalized clinical decision support and precision medicine.

Submission Number: 18