Natalia Trayanova pulls up a digital model of a patient’s left and right ventricles, the heart’s lower chambers that pump blood to the lungs and the rest of the body. She watches on her computer how the waves of blue, orange, green, yellow and red move across it. In a healthy beating heart, the colors would move across the ventricles in a wave as the heart’s electrical signal propagates, followed by a contraction from the heart. But Trayanova sees a spiral pattern — not unlike a hurricane or tornado on a weather map — indicating ventricular tachycardia (VT), a heart rhythm disorder that leaves part of the organ quivering rather than contracting. In this patient, it is caused by a heart disease called arrhythmogenic right ventricular cardiomyopathy (ARVC).

VTs, which can be life-threatening, are treated with ablation, which destroys tissue considered to perpetrate the VT, but it can be difficult to locate that tissue, making the treatment ineffective in some cases.

Trayanova, a professor of biomedical engineering at Johns Hopkins University and director for the Alliance for Cardiovascular Diagnostic and Treatment Innovation (ADVANCE), is looking at a “genotype-specific digital heart twin.” The virtual replica of a patient’s heart, which her team has pioneered, was created from imaging such as contrast MRI. The digital twin can help with diagnosis, treatment management and prediction of adverse outcomes, such as sudden cardiac arrest, based on an individual patient’s genetics and heart structure.

“We have representation of how the cells behave in that digital twin, and that information is usually very difficult to glean from a patient,” Trayanova says. Her team is conducting a study to determine the effectiveness of the digital twin in precise targeting of ablations, with the hope that it will increase efficacy, save time during the procedure and reduce recurrence of VTs.

The study is a collaboration with the ARVC research group at The Johns Hopkins Hospital led by cardiologist Hugh Calkins.

The team took a major step toward making digital twins more accessible recently, when they integrated an advanced AI model called DIMON (Diffeomorphic Mapping Operator Learning). When they applied DIMON to 1,000 digital twins, the AI accurately predicted electrical signal propagation in each unique heart structure, significantly reducing computation time usually required for such predictions — from several hours to a few seconds, on a personal computer. This could mean integrating individualized cardiac assessments into clinical workflows, and improving diagnostics and treatment plans for heart conditions such as arrhythmia.

“With DIMON,” Trayanova says, “we have a new, very important technique to make the digital twins scalable.”