Imagine a future where needing a new heart means you become your own donor. “It’s going to be your cells, your design—an exact replica of your heart, only made healthy,” said Steven Levine, Ph.D., senior director of virtual human modeling at Dassault Systèmes, in a recent interview.

For Levine, the goal of transforming cardiac care is personal. As he recounted in an April Fortune article, his now adult daughter received her first pacemaker at age two owing to a rare condition that reversed the lower chambers of her heart, making it less efficient at circulating blood. Traditional medical imaging offered limited insight, and the condition was so rare that doctors were uncertain how to treat it. “When she was first diagnosed, doctors had little confidence in their options,” Levine wrote. “As an engineer, I knew there had to be a solution,” he wrote in Fortune.

He found inspiration in virtual twins—the software models used to help airplane and automotive makers test designs and performance in digital simulations. “I thought, the heart is designed like a machine. It pumps blood in a regular pattern. It runs on electricity. Maybe I could help,” he wrote.

Unraveling the physics of the heart

This realization led him to consider how similar virtual twin technology could be applied to unravel the complexities of the human heart, ultimately inspiring him to launch the Living Heart Project in 2014. “I started it as an attempt to see if the technology—as you probably know, for building virtual twins—really was borrowed from the aerospace and automotive industries to build machines,” Levine explained. “And I wondered, well, you know, the human body obeys the laws of physics and chemistry and material science just like everything else.”

[All images below from Dassault Systèmes]

Hyperspecialization leads to a sort of Humpty Dumpty problem

Since the Industrial Revolution, the field of medicine has grown progressively more specialized, a trend that has gained momentum in recent years where advances in genomics, AI, and molecular medicine have created ultra-niche fields such as precision oncology or interventional neuroradiology. While this enables deep expertise in specific areas, it creates a fundamental challenge: putting the pieces back together.

“We fragment up the complexity of the human body into small enough pieces, and then we have all these specialists, and nobody brings it back together,” Levine said. “So I thought, what if I bring them together? Do we know enough to actually build a human heart?” The answer, it turned out, was yes—the human race had the knowledge; it just needed to be integrated.

Building a collaborative ecosystem

The Living Heart initiative at Dassault Systèmes brought together cardiovascular researchers, medical device developers, and practicing cardiologists to create highly accurate digital heart models. Their breakthrough came in 2014 with the world’s first 3D realistic simulation of a whole human heart. The centerpiece, the SIMULIA Living Heart Human Model, offers a detailed four-chamber simulation incorporating electrical, structural, and fluid flow physics—enabling unprecedented understanding of cardiac function in a virtual environment.

Levine structured the project deliberately to bridge professional divides while also enabling experts to dig into the details of their respective specialities. “We divided it up into teams focused on different aspects—treatment creation, application, device development, clinical practice, and statistics,” he explained. “Statisticians could talk to statisticians, not doctors and statisticians who don’t talk the same language.”

Toward superhuman cardiac understanding

The project takes a novel approach to AI, using it to generate virtual populations for testing. “We trained it on virtual twin data instead of human data,” Levine explained. “You’re actually feeding it massive amounts of fundamental human physiology. That training set, which is highly curated—we know exactly what we put into it—is incredibly rich and accurate when it comes to predicting how the human being behaves.”

This approach enables an understanding of the human heart that beyond human comprehension. “It’s a multi-dimensional surface that we as humans can’t comprehend, but the computer can keep track of all of that,” Levine explains. “So you can ask it, ‘Generate me a mitral valve that behaves this way,’ and it can do that because it knows all the pieces necessary. The average doctor couldn’t possibly—in fact, our best engineers couldn’t do that. It might take a week after training, but once trained, the AI can do it in seconds.”

Yet Levine acknowledges current limitations: “What we don’t have is the long-term longitudinal data—what happens three, five, ten years from now? That requires the next level of understanding of human physiology and data collection.”

Transforming regulatory decision-making

In November 2023, the FDA issued the final guidance document titled “Assessing the Credibility of Computational Modeling and Simulation in Medical Device Submissions.” The FDA’s integration of virtual twins represents a shift in medical device evaluation. The agency has been progressively incorporating digital technologies into its regulatory framework, recognizing the potential of virtual twins in medical device evaluations.

This evolution culminated in October 2024 with the release of the Enrichment Playbook, a collaborative effort between Dassault Systèmes and the regulatory agency. This 44-page peer-reviewed guide details how virtual twins can be effectively deployed in clinical trials, establishing credibility standards for in silico (computer-simulated) trials while bolstering patient safety and speeding innovation.

The traditional approach, where companies spend years developing products before presenting end results to regulators, is evolving. “A clinical trial is very much at the end of the process,” Levine explained. “After the company has spent five or 10 years convincing themselves that their technology works, they present that to the FDA as a ‘Yeah, and we believe it works,’ at which point the FDA starts asking, ‘Well, why do you think that?'”

Virtual twins offer a more dynamic alternative. “These twins show it to you in very clear terms,” Levine noted. “You don’t have to poke around inside an animal; you have a nice model. You can say, ‘Here’s the problem; here’s why it works. In the case of the mitral valve, here’s where the valve leaks; here’s why it leaks when we repair it; here’s our device; here’s the evidence; here’s what happens when it goes wrong; here’s how we mitigate that.'”

This clarity from virtual twins promises to transform how the FDA balances speed with scientific rigor. Traditional regulatory processes often hit roadblocks when uncertainties arise, leading to repeated cycles of “collect more data” requests that cost time and money. But virtual twins enable a more dynamic approach—creating an iterative feedback loop where each simulation builds on previous insights.

Cross-species applications of the virtual heart

The Living Heart Project could also shed light on cross-species applications. A Stanford team, collaborating with the project, recently received an ARPA-H grant to design and print a functioning human heart using human cells, with plans to test it in a pig. “We’re five years away, potentially, from creating an actually fully functioning human heart based on human tissue—not a mechanical one—because we now understand it well enough,” Levine said.

The work has revealed insights about transplant compatibility, particularly relevant to recent pig-to-human heart transplant attempts. “They’ve tried to transplant pig hearts into humans, thinking that it’s just a rejection mechanism,” he explained. “Both attempts failed, and I can tell you why: the operating conditions are sufficiently different.”

“We recognize that we have to design it differently to work in a pig because we’ve built pig hearts too,” he added. “If it works—which is obviously still a big ‘if’—it means we understand the heart well enough to build one and adapt it to an environment as unique as an animal. This suggests we could make it work inside any human because we now understand what it takes.”

The virtual twin technology also holds promise in helping individual cardiologists understand difficult cases. Practicing cardiologists have approached Levine asking, “‘What data do you need to help me model my patients? Because I’m tired of guessing,'” Levine noted.