Physics World Weekly Podcast Physics‑based simulations help diagnose and treat disease
Feb 5, 2026
Margaret Harris, a Physics World journalist who steers the conversation, chats with Amanda Randles, a Duke computer scientist and biomedical engineer known for physics-based, high-performance simulations of the circulatory system. They discuss medical digital twins, building patient-specific vascular models from imaging and wearables, tackling multiscale physics and petabyte-scale data, and using simulations to study metastasis and blood disorders.
AI Snips
Chapters
Transcript
Episode notes
Lookup-Table Approach Scales Twins
- Randles' longitudinal hemodynamic mapping creates a finite set of hemodynamic units that act like a lookup table.
- Precomputing those units lets edge devices infer 3D flow states from wearable data in near real time.
Validate With Full Simulations First
- Run full, high-fidelity simulations once to generate the hemodynamic library, then deploy lightweight inference on cloud or edge.
- Use discovery-phase full-data runs to find novel biomarkers before creating AI surrogates for deployment.
Supercomputers For Ground Truth
- High-fidelity validation runs require the world's largest supercomputers but production use can run on a few cloud nodes.
- Ground-truth supercomputer simulations are essential for verification against experimental and clinical measurements.