by University of Warwick

edited by Gaby Clark, reviewed by Robert Egan

Editors' notes

The GIST

Add as preferred source

Three steps toward a digital twin that doesn't just model your brain, it learns alongside it. First, the AI fine-tunes itself using tricks borrowed from the brain's own playbook for learning. Next, the digital twin and the real brain start responding to the same world in real time, watching, predicting, acting, adjusting, on repeat. Look far enough ahead, and the vision gets bolder still: sustained interaction could let both sides reshape each other, with the brain laying down lasting changes much like it does when learning sticks for good. The line in the diagram marks where science stands today (steps one and two) versus where it's headed (step three), a future where digital and biological intelligence don't just talk to each other, but evolve side by side. Credit: Dr. Ruohan Zhang / University of Warwick

Researchers from the University of Warwick report that how faithfully we can build a "digital twin" of a human brain depends not only on computing power but fundamentally on how much of the living brain we can measure, validate and update over time.

A digital twin brain is not just a detailed simulation. It is a model built from data collected from a specific person's brain and kept updated as new measurements come in: a living, individualized digital copy of your brain.

Now, a major review led by Warwick Manufacturing Group (WMG) researcher Dr. Ruohan Zhang, published in Nature Reviews Electrical Engineering, tackles one of the biggest questions in biology: How close are we to actually building a digital twin of a living human brain?

The bottleneck is measurement

Dr. Ruohan Zhang, assistant professor of applied artificial intelligence at WMG, University of Warwick, said, "A digital twin brain is more than a large-scale simulation of the brain. The key distinction is that it represents a specific individual and remains connected to the biological brain through data. The reality of creating a digital twin brain changes the question from 'How many brain cells can we simulate?' to 'How much of an individual living brain can we actually observe, constrain and update in order to create a true digital twin?'"

The clearest illustration of this challenge is scale: In simple nervous systems, like C. elegans, a roundworm with just 302 neurons, scientists can already map every synapse (brain connection). Human brains have around 86 billion neurons. Full synaptic-level mapping at that scale is currently far beyond technical and economic feasibility.

The researchers formalize this idea in a new framework, the measurement-defined emulation scale, which describes how the achievable fidelity of a digital twin brain is shaped by the resolution, completeness and updatability of the biological measurements available to constrain it, rather than simply by neuron count.

Keeping pace with a changing brain

In practice, that means creating a feasible human digital twin brain will require combining several imperfect tools rather than waiting for a single perfect one: detailed electron microscopy in small local regions of the brain, broader statistical mapping of cell types and circuits, and whole-brain MRI to tie it all together for one individual.

But mapping structure is only half the challenge. Brains change constantly through learning, aging, disease and experience, so a meaningful digital twin must also keep pace over time rather than being frozen at the moment it is built.

The review charts a path from today's largely static models toward ones that update as new data arrives and eventually toward closed-loop systems where a digital twin's behavior is checked against and corrected by the real brain it mirrors.

Professor Jianfeng Feng, University of Warwick and Fudan University, added, "It is important to distinguish what digital twin brains can do today from the longer-term vision. Current models can already reproduce selected structural and functional features using individualized biological data. The next challenge is to make these models increasingly adaptive: able to update as new measurements arrive and eventually interact with the biological system in closed-loop settings."

Three early uses in view

The researchers set out three ways the technology could be used:

• Neuroscience: using digital twin brains as virtual platforms for running experiments that would be impossible or unethical on a living brain, potentially providing one of the earliest mature applications of the technology
• Health care: using virtual patient twins for individualized disease prediction, treatment evaluation and virtual intervention testing
• AI: developing brain-inspired systems for studying perception, action, adaptation and interaction

"The long-term potential of digital twins is not to build larger computational models, but models that are meaningfully connected to individuals and can help us understand how biological systems change over time," concludes Zhang. "If developed responsibly, digital twin brains could become a shared technological and scientific infrastructure spanning health care, neuroscience and brain-inspired artificial intelligence."

Publication details

Ruohan Zhang et al, Building digital twin brains at the limits of measurement, Nature Reviews Electrical Engineering (2026). DOI: 10.1038/s44287-026-00320-8

Journal information: Nature Reviews Electrical Engineering