All work Healthcare AI · Research

Deep Learning + CFD for Coronary Artery Disease

Assessing coronary artery disease from CT angiography by combining deep learning with computational fluid dynamics.

RoleCo-author
TypePeer-reviewed research
DomainCardiology · Medical imaging
StatusPublished, 2022
Context

Coronary CT angiography (CCTA) is a non-invasive way to image the coronary arteries, but assessing the functional significance of a narrowing typically relies on invasive measurement or expensive fluid-dynamics simulation.

Problem

Bringing physically meaningful, flow-based assessment of coronary artery disease closer to routine, non-invasive imaging.

Objective

Assess CAD from CCTA by pairing deep learning with computational fluid dynamics.

Stakeholders

Cardiologists and patients being assessed for coronary artery disease.

My responsibility

Contributing author on the deep-learning methodology and experiments.

Approach

Couple learned models with CFD so that image-derived geometry informs a physically grounded assessment of the disease.

Solution

A deep-learning + CFD pipeline for CAD assessment from CCTA.

Outcome

Published in Applications of Medical AI (AMAI 2022), held with MICCAI 2022 (Springer LNCS).

Current status

Published, 2022.

Technical deep dive Implementation detail
Stack
Deep learningComputational fluid dynamicsMedical imaging