Digital Twins in Medicine: Personalizing Care, AI, and Privacy

Overview

Science Friday explores digital twins in medicine, a concept that combines patient data such as genetic information, blood tests, imaging, and family history into a dynamic digital model. These models aim to predict how treatments will affect an individual, moving medicine toward personalized care enhanced by AI and physics-based reasoning.

Dr. Caroline Chung, a radiation oncologist at MD Anderson, explains that a true digital twin is more than a static model or a visual avatar; it is an evolving system that updates predictions as new data arrive, effectively journeying with the patient through time. The conversation also delves into data gaps, the mix of mechanistic and AI-driven approaches, and the practical path toward clinical adoption and responsible use.

Introduction: What is a Digital Twin in Healthcare?

In this episode, Flora Lichtman speaks with Dr. Caroline Chung, a radiation oncologist at the forefront of digital twin research, to unpack how digital twins could transform medicine. The central idea is straightforward in concept: assemble a patient’s genetic information, lab results, imaging data, health history, and family background into a computational model that can simulate disease progression and response to therapies. But Chung emphasizes a crucial distinction: a digital twin is not merely a static visualization of a patient’s data. It is an ongoing interaction between a predictive model and continuous data collection that updates the model’s predictions over time, effectively traveling with the patient through their health journey.

Origins and Core Concepts: From Aerospace to Biology

Lichtman and Chung discuss the origin of digital twins in aerospace, where engineers create digital replicas of complex machines to anticipate failures and test designs in a risk-free environment. The goal is to identify the most promising designs to build physically. Translating this to biology is more complicated because living systems obey complex and not fully understood biological laws. Chung explains that while some physical laws (like those governing fluid dynamics in heart flow) provide a solid foundation for digital twins in medicine, many molecular mechanisms in humans remain under active discovery, leaving gaps that researchers must navigate as they build these models.

Data and Modeling Approaches: AI, Physics, and Hybrids

Not all digital twins rely on AI; Chung notes that many use physics-informed or mechanistic models, and others employ hybrid approaches that blend AI with established physical constraints to avoid implausible predictions. The emphasis is on smarter data integration rather than blind data fitting. In other words, digital twins leverage known physical principles to guide predictions while AI helps interpret complex data patterns when the underlying biology is not fully understood.

"A digital twin is more than just a model that can actually predict what is going to happen to you. It really is an ongoing interaction between what the model will predict your actions based on what information you receive, and continued data collection to update the information and predictions using that model. So it really is something that journeys with you through time." - Dr. Caroline Chung, radiation oncologist

Clinical Applications: Toward Personalization in Radiation Oncology and Beyond

Chung describes a concrete application in radiation oncology: digital twins could help identify which tumor subregions require higher radiation doses while sparing surrounding normal tissue. By simulating tumor heterogeneity and potential resistance, clinicians could adapt treatment plans early in the course of therapy, potentially improving outcomes and reducing toxicity. Beyond oncology, digital twins have been explored in cardiology to predict the need for catheterization before a heart attack, guide follow-up imaging schedules, and personalize screening intervals. The technology could also be used to test different chemotherapy schedules to optimize effectiveness while considering patient-specific physiology and tumor biology.

"we simulated giving the same chemotherapy with different schedules to get better results" - Dr. Caroline Chung, radiation oncologist

Practical Considerations: Data Flow, Accessibility, and Trials

The conversation turns to practical milestones for bringing digital twins into routine care. Realizing the potential requires robust data infrastructure, clear data governance, and regulatory pathways. MD Anderson and collaborators have begun designing prospective clinical trials to evaluate differential dosing and other twin-informed strategies, illustrating a path from concept to evidence-based practice. Chung also highlights the need for accessible data generation and sharing practices so digital twins can be used across diverse healthcare settings.

Privacy, Ownership, and the Human Element

A central concern is how to protect privacy as digital twins combine multiple data streams into a single, highly identifiable profile. The host and guest discuss the ethical dimensions: data ownership, access, and the risk that a digital twin could be misused if it becomes synonymous with a person. The dialogue stresses that clinical decision-making remains a human process, with the digital twin serving as a decision-support tool rather than an arbiter of care. The human-centered nature of medicine and thoughtful user interface design are emphasized as essential to preserving patient humanity even as technology advances.

"The final piece around this is that once you start to build together all of the different pieces into one entire human being, the privacy issues do escalate" - Flora Lichtman

Outlook: Inevitable Steps Toward Adoption

Looking ahead, the episode acknowledges that digital twins are already emerging in various medical disciplines and are likely to become more common over time. Realizing widespread adoption will depend on how medicine generates and flows data, how operational and regulatory frameworks evolve, and whether digital twins can deliver measurable clinical value. The discussion concludes with a balanced view: digital twins offer powerful potential for discovery and patient care, but their deployment must be guided by rigorous science, robust privacy protections, and a commitment to maintaining the human touch in medicine.