Medical digital twins represent a transformative approach to healthcare, offering a personalised, data-driven model of care delivery. First developed in engineering, digital twins in medicine dynamically replicate a patient’s health state to support clinical decision-making and treatment optimisation. As conventional treatment frameworks struggle to manage increasingly complex data, digital twins integrate multimodal health inputs and computational models to construct a real-time, evolving patient-in-silico. To fully realise their potential, digital twins must go beyond conventional AI or mechanistic models alone.
Defining Medical Digital Twins and Their Core Components
At the heart of a medical digital twin lies a continuous interaction between a physical patient and their virtual counterpart. The model includes five interconnected components: the patient, data connection, patient-in-silico, interface and twin synchronisation. Each plays a critical role in maintaining the fidelity and clinical relevance of the digital twin.
Must Read: Transforming Healthcare Through Medical Digital Twins
The patient provides the physical reference for the digital twin, encompassing either organ systems or the entire body. Data connections serve as conduits for multimodal health data—ranging from imaging to wearables—requiring harmonisation and fusion to support accurate modelling. The patient-in-silico, the core of the digital twin, replicates biological functions and predicts responses to treatments. Interfaces powered by AI, such as large language models, allow clinical teams to interact with the model effectively. Twin synchronisation ensures that the virtual model evolves in parallel with the patient, updating based on new clinical data or treatment outcomes.
Together, these components ensure that medical digital twins are more than static simulations—they are dynamic, synchronised systems capable of guiding precision medicine.
Technological Foundations and Model Integration
The implementation of medical digital twins relies on a suite of advanced technologies that support data acquisition, interpretation, modelling and application. Continuous data acquisition is facilitated by advances in sequencing and wearable technologies. For instance, liquid biopsies can capture tumour heterogeneity through blood samples, while wearable devices monitor real-time changes in vital signs and environmental exposures.
Artificial intelligence plays a central role in managing the complexity of health data. AI algorithms support feature extraction and data fusion from diverse sources, enabling the creation of structured inputs for the patient-in-silico. These AI tools are particularly useful when mechanistic understanding of a disease is limited. However, standalone AI models can lack interpretability and generalisability, making their fusion with mechanistic modelling essential.
Mechanistic models, rooted in mathematical representations of disease processes, complement AI by offering explainable frameworks for disease prediction. Their use is especially valuable when biological mechanisms are well understood, as seen in drug dosing models. Combining both approaches, including the use of physics-informed neural networks, enhances predictive power while maintaining clinical interpretability. This hybrid strategy offers a more robust and personalised model of care.
Clinical Applications in Oncology and Diabetes
Medical digital twins are already being explored in oncology and diabetes, two fields where personalised treatment can significantly impact outcomes. In oncology, digital twins are being developed to simulate tumour behaviour and treatment responses. Adaptive therapy in prostate cancer, for instance, uses mathematical models to balance treatment efficacy and resistance, extending progression-free survival. These models integrate patient data, tumour markers and evolutionary theory to personalise drug dosing and timing.
In diabetes care, medical digital twins aim to support continuous insulin management. Trials using AI-based decision support systems have demonstrated non-inferiority to physician-led care, especially where follow-up care is inconsistent or difficult to access. These models use inputs such as blood glucose levels and dietary patterns to recommend dosing adjustments. While current systems simplify the interface and synchronisation processes, future iterations are expected to incorporate more complex variables, such as nutrition and exercise.
Retrospective data staging has emerged as a valuable tool for testing medical digital twins without immediate clinical deployment. By simulating sequential data updates, researchers can evaluate predictive accuracy and uncertainty while preserving patient safety. These retrospective validations are especially useful in oncology, where historical data can be used to simulate tumour progression and treatment scenarios.
Medical digital twins offer a compelling vision for the future of personalised medicine. By integrating real-time patient data, AI technologies and mechanistic modelling, they move beyond conventional approaches to deliver adaptive, interpretable and precise care. Their successful implementation requires careful design across five core components, with enabling technologies driving progress in data capture, model fidelity and clinical usability. Use cases in oncology and diabetes underscore their potential to improve outcomes and address systemic gaps in access and effectiveness. In the future, medical digital twins may become central to clinical workflows, reshaping the landscape of healthcare delivery and precision medicine.
Source: The Lancet Digital Health
Image Credit: iStock
