Digital Diagnostics: Noninvasive Technology Is Revolutionizing the Diagnosis of Coronary Artery Disease
Decades ago, the introduction of minimally invasive techniques to diagnose and treat coronary artery disease (CAD) was a major advance for both cardiologists and patients.
Today, the advent of an equally effective technology to do the same thing noninvasively represents another significant step forward.
Using leading-edge digital twin technology and advanced computational models, our research is helping to transform how cardiologists diagnose coronary artery disease, the most common type of heart disease.
Anatomical insights from digital twins and supercomputers, not dyes and guidewires
Research conducted in my lab has helped to develop individualized digital twins of patients’ unique blood flow, which enable clinicians to noninvasively evaluate the severity of coronary artery disease and guide treatment decisions.
Importantly, our digital twin method matches the accuracy of more invasive ways of measuring CAD and its impact on blood flow.
We created these groundbreaking models and tested them in collaboration with clinicians and researchers at Brigham and Women’s Hospital, UC San Diego, UC Denver, and Duke Health interventional cardiologist Schuyler Jones, MD, and Center members Manesh R. Patel, MD, and Sreekanth Vemulapalli, MD.
The models were developed over years with strong contributions from many Randles Lab members. Madhurima Vardhan and Cyrus Tanade led the large multi-site study to validate the model and compare the non-invasive measures to the gold standard invasive guidewires.
Accurate data for cardiologists — no catheter required
Any time cardiologists use minimally invasive catheter-based tools to enter and visualize a patient’s coronary arteries, there’s a bit of resulting downtime and a slight risk of complications.
Not only can our noninvasive technology potentially prevent patients from having to undergo more invasive CAD diagnostic procedures, but it also gives those with certain medical conditions a needed diagnostic option.
Many people with, for example, bleeding disorders, uncontrolled infections, severe kidney disease, and dye allergies, simply aren’t candidates for procedures that require the use of dyes, contrast agents, or guidewires.
The digital models are also valuable to cardiologists who wish to evaluate coronary artery disease because they instantly provide all available information quickly— particularly critical during cardiovascular events like heart attacks.
Because cardiologists often work on coronary arteries with several tight, plaque-covered areas, it can be challenging to decide which blockage requires treatment.
Cardiologists have historically made that choice by performing and interpreting angiograms. These minimally invasive procedures involve injecting contrast or dye — or sometimes inserting a small wire — into an affected artery to look for decreased blood pressure near blockages.
Using the high-throughput methods developed in my lab, we provide the ability to computationally derive that pressure.
This means that cardiologists can noninvasively predict pressure drops to detect hydrodynamic wall shear stress (friction on artery walls caused by blood flow that is linked to disease progression) and/or weak parts of a coronary artery — resulting in more informed, appropriate, and targeted patient care.
Made possible by the pioneering HARVEY-FFR framework
Our extended-reality fractional flow reserve (FFR) HARVEY framework — the foundation of our work in cardiovascular medicine — is informed in real time by hemodynamic (blood flow) conditions that come from clinical measurements and millions of HarVI simulations performed on supercomputers.
This framework — which can reconstruct a 3D computational model of a patient’s unique cardiovascular anatomy — demonstrates that we can precisely compute FFR without the use of invasive wires, the current clinical gold standard.
Especially noteworthy is that FFR-HARVEY is able to identify a new biomarker: vorticity, a measure of fluid rotation. This marker can help physicians decide which patients and which blockages should receive treatment when the appropriate therapy isn’t clear-cut.
FFR-HARVEY translates information gleaned from 2D angiograms into 3D blood-flow models, enabling the tool to digitally recreate the blood flowing through the coronary arteries.
It is also noteworthy that FFR-HARVEY models show cardiovascular anatomy that other tools don’t always pick up, such as complicated areas of plaque buildup and disease, side branches that other tools often overlook, and patient-specific 3D features.
And because FFR-HARVEY can accurately simulate hemodynamics in areas of the body large and small, it is helping to make sophisticated cardiovascular diagnostics more precise, practical, and accessible to the physicians and patients who rely on that information.
This pipeline is especially effective at assessing complex and serial lesions in the coronary arteries. Other computational tools often don’t evaluate these types of lesions.
The efficacy of the HARVEY framework has been clinically validated in a 160-plus patient cohort at Brigham and Women’s Hospital and Duke Health, including three years of follow-up data.
These frameworks and models are helping clinicians:
- Evaluate cardiovascular disease more thoroughly
- Diagnose conditions more accurately
- Choose treatments based on more and more detailed patient data
- Get a more complete understanding of each patient’s unique cardiovascular condition
Our team has also developed one-dimensional models in which blood flow is resolved along a single axis. These 1D models have proven to accurately recover FFR as well as more invasive measurements do.
We have used the 1D models to perform millions of simulations for uncertainty-quantification studies, as well — allowing us to better understand the sources of errors when computing FFR.
Our progress in noninvasive diagnostics that use both 3D and 1D computational models to accurately assess coronary artery disease has great potential. The goal is for these advances to better inform clinical decision-making and lead to improved patient outcomes.
Future applications
The impact of our non-invasive, high-fidelity, physics-based modeling extends far beyond fractional flow reserve (FFR) calculations for coronary artery disease.
We are actively expanding this framework to enable longitudinal hemodynamic mapping — capturing how blood flow changes over extended time periods, not just a single heartbeat. This evolution supports remote monitoring of vascular health and allows for the detection of subtle physiological shifts that may signal early disease progression or treatment response.
In parallel, we are applying our modeling techniques to other vascular territories, such as the peripheral arteries, where diagnostic ambiguity and variability in patient anatomy present major clinical challenges. By simulating patient-specific blood flow with submillimeter resolution, we aim to guide intervention decisions for peripheral arterial disease and beyond, using the same core principles of physics-based accuracy and clinical relevance.
Together, these efforts support the broader mission of the Duke Center for Computational and Digital Health Innovation: to develop scalable, non-invasive tools that find, track, and treat disease with unprecedented precision. To learn how your institution might join our work in computational health innovation, please contact us.
