Not long ago, digital twins seemed more like science fiction than a viable medical pursuit, but as artificial intelligence has shown, applying complex algorithms to vast quantities of data pushes back the boundaries of what's possible.

Combining sensors with real-time data from various sources, digital twins are meant to create virtual replicas of things in the physical world so that simulations can predict the subject’s behavior under diverse conditions. They have applications in countless industries, and the digital twin of a patient has exciting potential to revolutionize healthcare.

Advocates for digital twin research in medicine believe the technology will someday integrate clinical information — including genetics, lab results, vital signs, medical history, imaging exams, wearable sensor data, and more — ushering in more personalized care and a better understanding of how these variables affect responses to different treatments.
In some industries, digital twins are already being used to test technology, (think aviation: stress-testing a simulated airplane engine under simulated conditions). Applications in medicine, however, are still in the early stages.

“One thing that will determine the speed at which it arrives is having hospital data stored with governance, protection, and security, but also having all the data sets that describe the patient be accessible, ideally in a common framework or at least in a set of frameworks,” Steven Niederer, chair of biomedical engineering at Imperial College London and co-director of the Turing Research and Innovation Cluster for Digital Twins at the Alan Turing Institute told HCB News.

Launched in March of 2023, the Turing Research and Innovation Cluster for Digital Twins partners with specialized experts in various fields to help create the computational digital twin infrastructure that will break down barriers currently hindering their use. The group is explicitly focused on three sectors: Environment, Infrastructure, and Health.

When it comes to healthcare, Niederer says one key missing piece is the mathematics for running simulations continuously and robustly. Such capabilities are necessary to determine uncertainties in patient data and model predictions, as well as ensuring the technology is reproducible and trustworthy.

There are also big questions about how much information existing technologies, such as medical imaging modalities, can contribute to a digital twin, and how digital twins might fit into the wider healthcare ecosystem.

The need for quantitative imaging
One company seeking to tap into the potential of digital twins in healthcare is San Carlos, California-based startup, Q Bio. The company unveiled its Bio Gemini digital twin software as a service (SaaS) in 2021, to capture and monitor comprehensive baseline patient health in a scalable virtual model.

Eventually, Q Bio plans to pair its software with a proprietary MR scanner currently under development called Mark I, a quantitative whole-body imaging system that Q Bio founder Jeffrey Kaditz says will be uniquely suited for the demands of digital twin integration.

Quantitative imaging derives numerical and measurable data directly from the 3D pixels of the scan itself. That's different from conventional (qualitative) scanners, according to Kaditz, where interpretations from human radiologists are required.

“[Qualitative imaging] is not information about tissue that is independent of the human operator or hardware used to take this picture. It’s a photograph, it's art,” he said. “Biomedical imaging needs to become more of a scientific tool than an artistic tool for digital twins to become a reality.”

Achieving quantitative imaging presents its own challenges, in part due to the vast amount of data that accompanies it. Machine learning algorithms will have to play a critical role in filtering that data for digital twins. “They can integrate and keep up with the pace of innovation of how much we can measure about our bodies and filter the signal from the noise to tell the doctor what data matters,” said Kaditz.

The amount of information that digital twins can potentially provide also requires the creation of high throughput data processing and standards, according to Niederer, as well as automation of the manual process for creating twins, software structures, and the integration of data management systems. Scanners also need to be cheaper and faster, and computational power higher to capture the state of the patient at different points in time.

“As you have data being collected sequentially, that allows you to update your model or your digital twin through time, and that’s where value is added,” Neiderer said.

Next steps
Niederer predicts digital twins will initially emerge in medical specialties where providers have already collected large amounts of data, such as pulmonary arterial hypertension and oncology, and expand in use cases as data accumulates in other areas. Over time, he expects this will change not just how providers approach and administer care, but the technologies they use.

“I think there will be value added through multimodality imaging, where you might be able to improve the accuracy of repeat lower-cost imaging, such as echo, using more expensive or radiation-based imaging like CT,” he said.

But to reach this point, healthcare providers, manufacturers, regulators, and other stakeholders need to be willing to share more information to build up these data collections, which in turn requires greater coordination around standards for data interoperability and transparency.

Kaditz expects digital twin innovations will likely start as preventive healthcare tools. “I would predict that in the future the entire triage layer of primary care will become based on digital twins and risk assessment before a doctor even sees a person,” he said.

Because of this, countries where preventive healthcare is not a top priority can expect to take longer to adopt and benefit from the technology clinically and financially. One is the U.S. where insurers do not incentivize preventive care as much as other services, according to Kaditz.

“We need to set up positive feedback to the healthcare system where the system gets cheaper and better over time,” he said. “I think digital twins provide a path forward to doing that if you're collecting this baseline information on people. You get better at forecasting and predicting the cost of care over time and you get better and better at delivering care.”

More Magazine Features

Health IT Homepage