From Brain Models to Executable Digital Twins: Execution Semantics and Neuro-Neuromorphic Systems
A new taxonomy proposes moving from isolated models to real-time, physically coupled neuro-neuromorphic systems.
In a new arXiv survey, researcher Alexandre Muzy from ILLS tackles a critical roadblock in neuroscience and computational medicine: the fragmented state of brain digital twin development. Current approaches are siloed across different data pipelines, model types, temporal scales, and computing platforms, preventing a cohesive, end-to-end workflow. Muzy argues this fragmentation breaks the 'execution semantics'—the consistent rules governing how a model's state evolves over time in response to events like simulation steps or real-world data inputs.
To solve this, Muzy introduces 'physically constrained executability' as a core unifying principle. This perspective shifts the comparison of different digital twin approaches from their internal model structure to their external execution behavior. The paper proposes a detailed taxonomy of execution regimes, ranging from basic isolated offline models all the way to advanced 'neuro-neuromorphic physical systems.' In these futuristic systems, biological brain dynamics and computational model dynamics would be 'co-executed' under shared physical constraints, enabling unprecedented real-time interaction and validation.
This framework clarifies why simply building a more accurate static model is insufficient for clinical utility. Instead, Muzy's agenda prioritizes semantic interoperability between systems, correctness across hybrid timescales, new evaluation protocols, and, crucially, scalable and safe workflows for closed-loop validation. By adopting this systems- and runtime-oriented view, the survey provides a common language for researchers and engineers to bridge disciplines and build digital twins that are not just theoretically sound but are executable, reliable, and clinically actionable tools for understanding and intervening in brain function.
- Proposes 'physically constrained executability' as a core principle to unify fragmented brain digital twin development across data, models, and platforms.
- Introduces a taxonomy of execution regimes, from offline models to real-time 'neuro-neuromorphic' systems co-executing with biological dynamics.
- Shifts focus from model accuracy alone to semantic interoperability, hybrid-time correctness, and safe closed-loop validation for clinical prediction.
Why It Matters
Provides a crucial framework to move brain simulations from academic models to executable, clinically reliable tools for personalized medicine and intervention.