Science Friday

Could a ‘digital twin’ help you get better health care?

March 17, 2026

Key Takeaways Copied to clipboard!

  • A true 'digital twin' in medicine is defined as an ongoing, interactive simulation that journeys with a patient over time, constantly updating predictions based on new data and patient actions, rather than just a static digital avatar. 
  • The concept of digital twins originated in aerospace engineering to pressure-test complex physical devices in a digital space before physical construction, but applying it to biology is more complex due to ongoing discoveries in molecular mechanisms. 
  • In oncology, digital twins are being developed to personalize radiation treatment by anticipating aggressive tumor regions and optimizing dosing to spare normal tissues, while other applications include personalizing cancer screening follow-ups and optimizing chemotherapy schedules. 

Segments

Defining Digital Twins
Copied to clipboard!
(00:01:26)
  • Key Takeaway: Digital twins are more than visual avatars; they require ongoing interaction between prediction, patient action, and continuous data updates.
  • Summary: The concept of digital twins involves compiling personal health data into a simulation to predict treatment efficacy, supercharged by AI. However, the definition is often blurry, differing from the original intent derived from aerospace engineering. A true digital twin must involve continuous interaction and updating over time, not just a static model.
Aerospace Origin and Complexity
Copied to clipboard!
(00:03:33)
  • Key Takeaway: Digital twins originated in aerospace to pressure-test costly physical devices digitally before real-world testing, bridging the physical and digital worlds.
  • Summary: The concept arose because physically testing every possibility for spacecraft or aircraft is too costly and difficult. Digital replicas allowed engineers to pressure-test designs virtually before moving promising ones to physical testing. Applying this to biology presents greater complexity because many molecular mechanisms are still being discovered.
Data Availability and Model Types
Copied to clipboard!
(00:05:22)
  • Key Takeaway: Sufficient data exists to model systems governed by known physical laws, like fluid dynamics in the heart, but gaps remain for less understood biological areas.
  • Summary: Confidence exists in modeling certain systems, such as heart mechanics, due to established physical laws. Digital twins may utilize physics-informed (mechanistic) models, AI models, or hybrids combining both. Mechanistic models help constrain results, preventing completely inaccurate predictions when data is input.
Oncology Application: Radiation Dosing
Copied to clipboard!
(00:06:54)
  • Key Takeaway: Digital twins can personalize radiation therapy by anticipating resistant tumor cells to adapt dosing and minimize toxicity to normal tissues.
  • Summary: A key application for radiation oncologists is defining sub-regions of a tumor needing higher radiation doses while sparing healthy tissue. Digital twins can anticipate where aggressive or resistant tumor cells are located. This allows for personalization of radiation treatment early in the process to improve outcomes by accounting for toxicity.
Broader Medical Use Cases
Copied to clipboard!
(00:08:36)
  • Key Takeaway: Digital twins show promise in cardiology for predicting catheterization needs and in cancer care for personalizing screening frequency and chemotherapy schedules.
  • Summary: In cardiology, twins have mimicked blood flow to anticipate necessary interventions before a heart attack occurs. For cancer follow-up, twins can personalize screening tests, potentially starting them earlier for high-risk individuals or spacing them out for others. Simulations also revealed that the same total dose of chemotherapy can yield different outcomes based on personalized delivery schedules.
Scope and Privacy Concerns
Copied to clipboard!
(00:10:32)
  • Key Takeaway: While a full ‘virtual human’ is an aspiration, digital twins should be fit for purpose, as compiling all systems escalates severe privacy and identifiability risks.
  • Summary: There are global efforts aiming to build a digital twin of an entire human being. However, the investment must be justified by the specific clinical question being asked, as connecting all pieces may not always be necessary. Compiling all data into one twin makes the individual highly identifiable, significantly escalating privacy concerns.
Culpability and Human Oversight
Copied to clipboard!
(00:15:06)
  • Key Takeaway: Clinical decisions remain the collective responsibility of the physician and patient, requiring critical evaluation of model outputs to maintain the humanity of medicine.
  • Summary: Ownership of a digital twin is complex, requiring system supports to keep the model alive, making individual access difficult. If a twin’s prediction leads to a bad outcome, the practicing clinician considers the output as one frame of information alongside personal and social factors. Physicians must balance enthusiasm for technology with skepticism, ensuring critical thinking is facilitated by how information is presented.
Future Inevitability and Adoption Hurdles
Copied to clipboard!
(00:19:38)
  • Key Takeaway: Emerging versions of digital twins are likely to enter mainstream medicine, contingent upon resolving operational, legal, and regulatory hurdles regarding data flow and accessibility.
  • Summary: It is likely that some forms of digital twins will eventually be adopted across medical practices. Systemic adoption requires addressing how data is generated and flows within healthcare systems. Operational, legal, and regulatory practices must be addressed to ensure accessibility for everyone.