A MODEL FOR GOOD HEALTH
Although plenty of connected wireless medical devices exist today, the future of the Internet of Medical Things will likely include digital twins of human health and disease.
Written by Kayt Sukel
Digital twins of the biomechanical function of the uterus and cervix, mapping the tissue stretch for five pregnant patients at different anatomical resolutions. Image: Erin Louwagie, Columbia University
A PREGNANT PATIENT IN HER 38th WEEK feels a deep contraction in her abdomen and wonders if labor has finally begun. Thanks to sensors placed strategically on her smart watch and within her maternity clothes—and a unique biological sensor placed on her cervix—her OB-GYN is getting real-time data about both her and her baby’s health status. It was those very sensors that allowed the doctor to diagnose her with Braxton-Hicks contractions, sometimes called “false labor,” just a few weeks earlier, saving the patient an unnecessary trip to the hospital.
As the patient’s contractions strengthen and true labor is confirmed, the obstetrician messages the patient and tells her it’s time to make her way to the hospital. Once they meet in the labor and delivery ward, the doctor pulls up a detailed digital twin, or virtual model simulating the organ function of both mother and baby built from past and current data, and uses it to plan the best approach for a healthy delivery.
This vignette may sound like something directly out of a science fiction novel or your favorite Star Trek franchise. But scientists and engineers are already hard at work on the next phase of the Internet of Medical Things (IoMT), and many believe it will be anchored by digital twins of the human body. Engineers envision these models to help scientists test new drugs and treatments, monitor patients remotely, provide early diagnoses, detail disease prognoses, and help doctors make more informed clinical decisions about patient care.
“Connected medical devices, ranging from pacemakers to insulin pumps, are already fairly commonplace in the medical field. The real-time data from these devices allows healthcare providers to monitor patients and make early interventions to improve healthcare outcomes,” said Jeffrey Karp, a professor of biomedical engineering at Harvard Medical School and the Massachusetts Institute of Technology. “Digital twins, or virtual models that simulate a patient’s unique physiology, is a fascinating next step to help doctors test new treatments and provide the kind of insights to offer safer, more personalized healthcare.”
While the development of such models is still in its infancy, their potential is great, said Kristin Myers, an assistant professor of mechanical engineering at Columbia University and director of the Soft Tissue Lab who had been working on digital twin models of both the uterus and cervix.
“Marrying the physics to the biology of these organs provides a powerful three-dimensional representation of the human body,” she said. “And it can give physicians and caretakers an evidence-based picture of a person’s current and future health to help them make smarter decisions.”
Taurus Vascular's overflow procedure. Video: Taurus Vascular
“Every day, millions of patients have different [heart] measurements taken. There’s a massive amount of information being collected, yet no one has ever put it all together to create a model that demonstrates this is how a heart is supposed to work.”
—Steve Levine, senior director of Virtual Human Modeling at Dassault Systèmes
HAVE A HEART
Steve Levine, senior director of Virtual Human Modeling at Dassault Systèmes, grew interested in the potential of a digital twin for the human heart after his daughter was diagnosed with a challenging congenital heart condition as a young child.
“Her doctors were doing their best to try to figure out how to treat her, and as her parent, I was being asked what I thought about the different interventions they were proposing,” he explained. “It occurred to me that cardiologists could use a functional model where they could get data points from every interaction over time to see what’s happening and then use that as a prototype to test different ideas to see if they would work.”
Levine, with the help of his employer, then embarked upon the Living Heart Project, which brought together engineers and cardiologists to develop and validate a digital twin of the human heart.
“Billions of dollars have been spent on research into the cardiovascular system. Every day, millions of patients have different measurements taken,” he said. “There’s a massive amount of information being collected, yet no one has ever put it all together to create a model that demonstrates this is how a heart is supposed to work.”
The Living Heart Project currently has a healthy heart model for researchers to test against. Scientists and clinicians can use the simulation to look at different disease conditions and see the impact of specific problems on the entire system.
“You can disturb the electrical system or reduce the strength of the tissue in a heart valve. You can drop the blood flow and see what happens,” Levine said. “We let physics do the job so we can better understand what different diseases do to the heart.”
The Living Heart Project is only one of several large-scale efforts to virtually model aspects of human health. Amanda Randles, Alfred Winborne, and Victoria Stover Mordecai, Associate Professor of Biomedical Sciences at Duke University, are also focused on building a digital twin of the heart, but with the goal of providing doctors real-time information to monitor patients with cardiovascular disease from afar.
Meanwhile, Myers and her lab are working on digital twins of the uterus and cervix to promote better women’s health outcomes. And Jun Deng, a professor of therapeutic radiology at Yale Medical Center and a leader of the Digital Twins for Health Consortium, is developing digital twins of tumors to help physicians determine the best course of treatment for cancer patients.
But even with such cutting-edge developments already in play, there are several big questions that remain. As engineers and clinicians pursue a future with more sensors and connected devices to support virtual twins for human health and disease, just what kind of data, exactly, are needed to make the models as robust and accurate as possible? How will developers manage the many potential privacy and security vulnerabilities involved with collecting and sharing that data? And, as digital twins of different bodily systems come to fruition, how can they be integrated so doctors can use them to support holistic clinical decision-making?
An illustration envisioning digital twins for health (DT4H). Image: Yale University
The Dassault Systèmes Living Heart Project model. Image: Dassault Systèmes
A PET linear accelerator platform like this one at Yale-New Haven Hospital enables biology-guided radiation treatments of lung cancer. Studies show that digital twin dyads could help determine treatment choice pathways, from chemotherapy to radiation. Photo: Yale University
ALL MODELS ARE WRONG IN SOME WAY
Cameron Kim, a professor of biomedical engineering at Duke University and a biomedical ethicist, reminds his students of the famous line by British statistician, George Box: “All models are wrong, but some are useful.”
“When we start talking about digital twins in health, and how we synthesize all of the potential data points we have, we need to understand that these models, by their very nature, are going to be wrong in some fairly fundamental ways,” he said. “That’s important to think about because the very phrase ‘digital twin’ has a certain connotation—it gives off the idea that the model is identical to the organ or human it represents—and no matter how much data we collect and integrate, it will never be that.”
Once engineers understand (and can effectively communicate) that their digital twin is not a duplicate of a patient, however, they can work on building a model that is useful. But that requires carefully considering which data points are most pertinent to achieving their goals with the model, Deng said.
“We are trying to model cancer. And there are so many types of data involved even in just a single type of cancer,” he continued. “There are now massive data sets available. But we need to find ways to identify the core data we need—the basic requirements needed to develop a digital twin for a particular medical phenomenon. And that will take a consensus of experts in that field.”
For the Living Heart Project, Levine and Dassault Systèmes brought together experts in different aspects of cardiovascular function, much in the same way engineering organizations “bring together many, many different organizations to build a commercial jet,” Levine explained.
“We found all the different heart experts and then we acted as a hub, using [Dassault’s] platform to bring everyone and their data together,” he continued. “That’s how we created this phenomenological model.”
The goal of the Living Heart model is to model a healthy heart and how it may be affected by disease, not the individual heart function of a particular patient, although there are researchers working to create models that can do that. Others would like to employ digital twins to support remote patient monitoring or clinical decision making and that, Randles said, is going to require a “huge discovery phase.” To predict a potential atrial fibrillation event with accuracy, for example, will likely require the right combination of several different data points, she explained.
“We’ve never been able to collect all this data before from patients. We don’t know, for certain, what’s most important,” Randles said. “We need to do the studies to determine what data points are most relevant, and how we connect all this multimodal data in the right way. We need to do more research work so we can not only identify the data that will help us build these models and validate them, but calibrate them once they are in use.”
Today, it’s estimated that 30 percent of the world’s data is generated by the healthcare industry, and trying to make proper use of it is a serious challenge for engineers and digital twin developers, Kim said.
“It’s hard for us to fathom just how much data is at our fingertips,” he added. “The first conversation anyone who is developing a digital twin should be having is, what does the data I need actually look like? How will I use it? And how do I deal with the uncertainty that is going to come as I incorporate it into a model like this that so many people are going to assume is a carbon copy of the real thing? Then you also have to think long and hard about how the model will benefit the patient.”
And that’s before the sensors and devices that feed one digital twin system need to be successfully integrated to support a virtual model of the entire body. After all, what affects the heart might also affect the brain or uterus. Systems engineers are well aware of the design problems that come when they try to put together data that generally exists in silos.
“Each of these models are complex in their own right,” said Matthew Kuhn, founder and CEO at Taurus Vascular, a medical device company. “How do you put those pieces together? Because you are just adding more complexity, and you need to make sure that anything you are using to collect data isn’t interfering with something else that is helping a patient to maintain a different model.”
It is a complex undertaking, agreed Myers. But she sees great value in the pursuit. She likens the development of digital twins of the human body to self-driving automobiles. Even if it takes decades to get to a fully functional end point, she said, we are still going to discover incredible information that will help advance medicine along the way.
“There are a lot of steps we need to take between now and the creation of a reliable model of a human organ system,” Myers said. “But getting there is going to inform science so much. There’s very little understood about the uterus and what factors and forces work upon this muscular pressure vessel that affect a woman’s health. What we learn as we go is going to help us design better models, better devices, and better interventions. Just as self-driving cars have now given us a range of new safety technologies, our work on these models is going to feed discoveries with the power to change the field of women’s healthcare.”
“We have an important role, as engineers, to not be ignorant of what kind of data these models need, how we get it, how it’s used, and how we can protect the people who are trusting us with it so it isn’t used in ways that were not intended.”
—Cameron Kim, professor of biomedical engineering at Duke University
HEALTH DATA PROTECTED
As engineers access colossal data sets to build different types of digital twins, developers also face challenging data privacy and security concerns, Kim said. With so much vital data in play, it is imperative that engineers consider these factors as they design, Karp added.
“My sense is there will be a parallel evolution of sensor and device cybersecurity that will occur along with advances in the models themselves,” Karp said. “The algorithms will keep getting better, new data streams will come in, and we will have to constantly think about security measures and finding ways to plug any holes or gaps in more sophisticated ways.”
That parallel evolution should not just focus on nefarious actors, Kim said. While many hear the word cybersecurity and think about hacking for profit, he is more concerned about the way ignorant actors may misuse health digital twins in the future.
“Robert Merton, a founding father of sociology, said back in the 1930s that you need to think about what might cause unintended, negative consequences in science—and his number one thing was ignorance,” he said. “We have an important role, as engineers, to not be ignorant of what kind of data these models need, how we get it, how it’s used, and how we can protect the people who are trusting us with it so it isn’t used in ways that were not intended.”
A CONNECTED WORLD OF HEALTHCARE
As Kim stated, every decision an engineer makes is an ethical one—from material choice to data security capabilities. And as scientists, clinicians, and engineers work toward a future where virtual models of health and disease help to enhance care delivery, he emphasized that they must be reminded of the importance of being able to articulate how such models work, what data is needed to power them, and what they really offer patients above the status quo. It’s not an easy task, he said, but it is a necessary one.
“It’s important for us as engineers to ensure that any device or model works the way it is supposed to. That they are improving a patient’s ability to engage with doctors and access care,” Kim said. “But it’s also our job to think about who is going to use what we build and what we need to do and communicate to both patients and physicians in order to make sure such models are used properly.”
We may not be able to offer that pregnant patient a robust virtual model of her pregnancy yet, but Kuhn, for his part, believes that the future IoMT, including digital twins, holds great promise in one day getting us there. But it’s going to take a lot of work—and a fair amount of introspection—to succeed.
“There are going to be great challenges along the way. There will be hiccups. There may even be some tragedies,” he said. “But the ultimate potential to use these models to help underserved patients and improve access to healthcare, to improve the quality and durability of treatment, and to improve healthcare outcomes is super exciting.”
Kayt Sukel is a technology writer and author in Houston.
Jun Deng, professor of therapeutic radiology at Yale Medical Center. Photo: Yale University
© 2025 The American Society of Mechanical Engineers. All rights reserved.