The scenarios seem borrowed from science fiction: Doctors use a digital model of the human heart to try new cardiac surgery techniques. Drug manufacturers test innovative cancer treatments on virtual systems that mimic patient response. Hospitals execute “digital stress tests” on their computerized doppelgangers to anticipate how a global pandemic might affect capacity and resources.
Welcome to the world of “digital twin”–assisted healthcare, where virtual modeling of medical systems and procedures, as well as organs, tumours and even entire human bodies, is transforming medicine — and paving the way for promising new personalized treatments and faster, more efficient delivery of services.
The global market for digital twins is surging, fueled by the rise of artificial intelligence and cloud-based computing, technologies that make these models more efficient and affordable. By 2025, 25 per cent of healthcare delivery organizations will incorporate digital twins in new tech initiatives, according to a study by Gartner Peer Insights.
Virtual simulations can help medical experts develop customized treatments that are individually tailored to specific patients. And by enabling administrators and clinicians to run “what if” scenarios — what if this drug were used instead of that, or what if 500 patients suddenly showed up in the emergency room? — digital twins can predict risk and help decision-makers avoid costly mistakes.
Opening the ‘digital front door’
Toronto-based Verto Health has created digital twins for the Fraser Health Authority in B.C. at a scale that’s unprecedented in the province’s hospital sector. The Verto project, which involves virtual replicas of two million patients, will make navigating the hospital system faster, more effective and more convenient, says Sheazin Premji, executive director of FHA’s Centre for Advanced Analytics Data Science and Innovation.
The first phase is a “digital front door” that allows patients to search for services, book appointments and seek information by way of a virtual assistant called (aptly) Fraser. Longer term, the system will perform a wider range of functions — optimizing acute-care bed capacity by streamlining the discharge process, for instance — that will enable clinicians to make better use of their time and expertise. “From a patient perspective, it’s a real game-changer,” she says.
When files are long or complex, medical staff often spend an inordinate amount of time reviewing notes and related scans and documents. Verto’s AI-assisted system is designed to help them focus on people rather than paperwork, says founder and CEO Michael Millar. Once deployed, it will be able to instantly analyze a patient’s digital-twin file and highlight critical information — a family history of lung cancer, for instance, when someone presents complaining of a cough.
The Verto tool “can read 10 years of records and give a 30-second summary,” Millar says, with obvious enthusiasm. “So physicians can focus on [determining] what is the best care to deliver given the information that’s provided to them by the digital twin.”
It can also perform “light prescriptive” functions, he says, such as referring patients to screening programs or specialists. Finding ways for healthcare providers to work more efficiently is critical, he says. “Canada has huge sustainability concerns over the next decade or so. We don’t have enough doctors and we don’t have enough nurses and the problem is global. We’re going to have to get smarter about the types of things we want our limited doctors and nurses doing.”
AI-assisted trials
Altis Labs, another Toronto startup, uses digital twin technology to help cancer drug developers “get the best new drugs to patients as quickly as possible,” says founder and CEO Felix Baldauf-Lenschen. The cost of developing a new drug can be prohibitive, Baldauf-Lenschen says. For a novel treatment to move through clinical trials and secure regulatory approval, he says, typically requires more than US$2 billion and can take at least 10 years. The exorbitant cost has much to do with the high failure rate of phase two and phase three trials.
Altis aims to minimize late-stage setbacks by using digital twins to troubleshoot issues early on in the process, essentially helping them “fail faster, fail sooner.” In the case of oncology drugs, trials generally use changes in tumour size as an indicator of efficacy — if the tumours are shrinking, the treatment is considered to be working. But in fact, tumour size isn’t necessarily the best predictor of overall survival rates. It’s been used primarily because better metrics haven’t been available.
Altis has developed a system that generates richer, more textured metrics by training AI models to predict outcomes for each patient; the tech compares CT scans of a diseased area with AI-generated images that show what the same individual’s tissue would look like without cancer. It can then analyze a variety of other factors — bone density, for example, and the overall spread of the disease — to forecast a person’s chances of survival.
The result, says Baldauf Lenschen, is “a better way to measure disease burden and anticipate differences in survival as patients undergo treatment”; one recent study using historical patient data found Altis’s digital twin technology to be 40 per cent more accurate at predicting survival rates than conventional radiology scan interpretation..
For now, the technique is used only to collect better, faster evidence as research in clinical trials, but Baldauf Lenschen says Altis hopes to eventually establish a range of applications for digital twin technology to help the healthcare system move away from one-size-fits-all care. He envisions a future where a clinician might have the ability to simulate the trajectory and outcomes a given patient might expect with one treatment versus another. “Those are some of the blue-sky things that are exciting,” he says.
Getting a sense of the big picture
Digital twin technology can be used to test what-if scenarios in broader social contexts too. RUNWITHIT Synthetics, another Canadian venture, designs replicas of geographic areas — typically cities and linked regions — to assess emergency preparedness and other issues that can affect the well-being of a population.
In Tennessee, for instance, RUNWITHIT developed a model to stress-test extreme heat scenarios, exploring which areas would be most affected and where to best deploy cooling centres and other health-related resources. Simulated scenarios helped planners and administrators make better informed decisions, and allowed them to identify vulnerable populations to ensure services are distributed equitably.
As healthcare organizations roll out digital twin initiatives, says Ibukun Abejirinde, an assistant professor at the University of Toronto’s Dalla Lana School of Public Health who also works as a scientist with Trillium Health Partners, equity must be top of mind. “It’s important to ask: Who is this going to serve, and is it at the expense of another group?”
Digital twins need massive amounts of data, she points out, and historically, marginalized groups have typically been left out or not explicitly identified when information is collected. “We don’t have their data, or if we do, it isn’t complete or accurate,” she says. “When you’re replicating an imperfect system, even if it’s a virtual replication, you’re just building on an imperfect reality, and this limitation needs to be acknowledged.”
Abejirinde also worries that systems that are overly reliant on technology can neglect the human values that are critical to healthcare. “There’s no algorithm for compassion,” she says. For instance, a patient may well experience anxiety if a virtual consultant suggests they seek cancer screening. “How does a bot handle the panic they might have in that moment: ‘What? You’re saying I have cancer?’””
Healthcare organizations must also prioritize privacy and data security when implementing DT initiatives, as noted in a recent report by the U.S. National Academy of Sciences, Engineering and Medicine.
“Protecting individual privacy requires proactive ethical consideration at every phase of development, and within each element of the digital twin ecosystem,” the authors concluded.
Verto, Altis and RWI all take steps to ensure that data is either anonymized or can only be accessed behind secure firewalls. Premji says FHA’s digital twins are all “guardrailed,” and the digital front door tool doesn’t collect private data. The technology has nevertheless already prompted the hospital to reassess existing services, based on patients’ questions and searches.
Verto’s Millar acknowledges the crucial importance of security issues and equity concerns. Still, he sees digital twins as a key element of a bright — and more equitable —future.
“If we don’t have tools that solve the healthcare system’s problems intelligently, we’re all going to have less access to services and providers,” he says. “The important fight is to ensure that every minute a clinician is working is meaningful and is about connection with people. If we don’t change that paradigm, then the future will just be about not getting adequate service.”
CTVNews.ca has partnered with MaRS to highlight Canadian innovations in health care
