Duke's digital twin technology could take hemodynamic assessments to a whole other level
Researchers at Duke University are developing a cardiovascular digital twin that could move hemodynamic assessment beyond the fractional flow reserve (FFR) measurements used by clinicians today. The goal is to be able to predict hemodynamics days, weeks or months into the future to better inform care decisions and to enable virtual testing of different implantable devices.
Amanda Randles, PhD, director of the Duke Center for Computational and Digital Health Innovation and professor of biomedical engineering, said her laboratory has developed HARVEY, a computational fluid dynamics platform that creates personalized models of blood flow from medical imaging.
HARVEY can use CT, MRI or 2D angiography imaging to reconstruct a patient’s 3D vascular anatomy. It then applies physics-based fluid dynamics calculations to simulate blood flow and generate measures, including pressure, velocity, vorticity and FFR.
Unlike FFR-CT, which provides an assessment based on the conditions represented by a particular imaging study, Randles said HARVEY is being developed to model how a patient’s hemodynamics change over much longer periods of time.
The system can incorporate information from commercial wearable devices to model changes associated with everyday activity.
The Duke team has developed mathematical and machine-learning approaches that allow simulations to scale from individual heartbeats to millions of them. Randles said the laboratory has simulated approximately 4.5 million heartbeats, representing roughly six weeks.
That continuous data stream is an important distinction between a personalized model and a true digital twin, Randles said. A digital twin must be continually informed by sensor data so the model can evolve with the patient in real time. Heart rate, for example, can be used to update inflow conditions and model how coronary blood flow changes with activity.
The approach could eventually provide cumulative measures of cardiovascular risk rather than relying on a single measurement threshold or snapshot. Randles noted that the same heart rate and cardiac output can produce different hemodynamic risk in various patients because of differences in vascular anatomy.
Rather than applying a universal FFR threshold such as 0.80, future models could potentially track changes from an individual's personalized baseline and measure how long a patient remains exposed to potentially harmful hemodynamic conditions. Those measurements could help evaluate interventions such as medications or changes in exercise.
The technology also could support treatment planning. Duke researchers are studying applications involving coronary atherosclerosis, heart failure and carotid disease. In coronary bifurcation disease, for example, clinicians could potentially model different stenting strategies, simulate months of subsequent patient activity and compare the resulting cumulative risk before performing the procedure.
Computing advances have been critical to making this possible. Randles said a full coronary simulation in 2010 required the world's largest supercomputer at the time for about six hours to generate a single heartbeat. Today, GPUs, reduced-order models and machine learning allow a full heartbeat simulation to run on a much smaller computing system in minutes. The Duke team has demonstrated the ability to generate months of modeled data in about half an hour, while also using brute-force simulations and clinical measurements to validate the results.
Randles believes these digital twins could be used clinically within five to 10 years, with cardiology likely to be among the first specialties to adopt them.
Dave Fornell has covered healthcare for more than 17 years, with a focus in cardiology and radiology. Fornell is a 5-time winner of a Jesse H. Neal Award, the most prestigious editorial honors in the field of specialized journalism. The wins included best technical content, best use of social media and best COVID-19 coverage. Fornell was also a three-time Neal finalist for best range of work by a single author. He produces more than 100 editorial videos each year, most of them interviews with key opinion leaders in medicine. He also writes technical articles, covers key trends, conducts video hospital site visits, and is very involved with social media. E-mail: [email protected]