Recording of the 2025 ECTS–IFMRS Joint Session, Vienna
Speaker: Prof. Liesbet Geris (University of Liège & KU Leuven, Belgium)
Data and model always go together
Digital twins rely on four layers working in concert:
sensors/measurements → middleware → computing hardware → software models.
Twins can represent cells, tissues, organs, whole systems, medical devices, bioprocesses or entire facilities, whichever is sufficient to answer the clinical or research question.
Modelling spans a spectrum, not just AI
From black-box, data-only approaches to white-box, physics-only ones, most useful twins sit in the grey zone – hybrid models that combine mechanistic knowledge with machine-learning components to curb ‘hallucinations’.
Lab examples show the breadth of scales
- Single-cell cartilage atlas & executable GRN model – maps chondrogenic differentiation states, then predicts interventions to steer cells; predictions validated in vitro.
- Multiscale osteoarthritis twin – links gait-derived joint loading → cartilage stress → integrin signalling → intracellular response.
- Curvature-driven scaffold growth model – converts the biological rule ‘cells fill concave corners first’ into maths; guided design of gyroid bone implants that outperformed lattices in mouse and large-animal studies.
- Agent-based fracture-healing model – simulates macrophage M0 → M1 → M2 dynamics and cytokine fields during the inflammatory phase; calibrated with in-vivo immunofluorescence, now being extended with mechanical loading data.
Clinical traction is already real (at single-organ scale)
Personalized left-atrial appendage occluder sizing, surgical planning tools, and other device-focused twins are in routine use and even reimbursed; one physics-only twin has entirely replaced a medical-device clinical trial.
Regulatory and credibility science are maturing
EMA/FDA guidance, emerging “Good Simulation Practice” and ~150 existing standards give clear (if complex) pathways for validation, documentation, and risk management.
Europe’s next step: the Virtual Human Twin initiative
The EU aims to knit dispersed efforts into a shared platform that includes:
- catalogue of models, data sets, and workflows;
- tooling for automated credibility tests;
- common metadata so models can ‘crawl’ data spaces;
- alignment with the European Health Data Space, AI Act, MDR, etc.
Community-built roadmap & manifesto
800+ stakeholders co-authored a 30-recommendation roadmap (plus a short policy brief and 2-page manifesto) emphasising unmet-need-driven use cases, full clinical-workflow integration, and sustainable infrastructure and business models.
Validation remains context-dependent
High-TRL scaffold work moves quickly from mouse to large-animal to spin-off products; the inflammatory fracture-healing twin is lower-TRL and still in fundamental calibration, illustrating that digital first, then bench & animal is iterative, not linear.
Bottom line
Digital twins are already delivering niche clinical value, and hybrid, multiscale approaches are unlocking far richer questions – from single-cell fate decisions to hospital logistics. Europe’s Virtual Human Twin push seeks to make those scattered successes reproducible, interoperable, and trustworthy at continent scale.
