HKBU physicist wins HK$7.8 million to build virtual brain models for neuromodulation

Hong Kong's Research Grants Council has awarded chair professor Zhou Changsong approximately HK$7.8 million (US$993,400) over five years to develop AI-based prediction models that could change how clinicians plan neuromodulation therapy. The grant comes from the council's Senior Research Fellow Scheme during its inaugural 2023-24 term.

Zhou, chair professor in physics and complex systems at Hong Kong Baptist University's department of physics, leads a neuroscience programme that treats the brain as a highly complex interacting network rather than a set of isolated regions. His dynamics-based evaluation methods use AI to model massive, highly non-linear brain activity — the kind of whole-brain electrophysiological data that conventional analysis struggles to capture.

"Mental health and neurological disorders present some of our greatest medical challenges, but current neuromodulation treatments [targeted electrical impulses or tiny pharmaceutical doses] often rely on trial and error," Zhou said. "My research aims to change this by treating the brain as a highly complex interacting network. By building virtual brain models and using systems-analysis methods, we hope to pinpoint precisely which brain regions are sensitive to treatment."

What does this change for clinical practice?

The practical target is straightforward: move clinicians away from iterative dose-and-observe protocols toward simulations run before a patient is ever stimulated. Zhou's team indicators reveal how different brain areas coordinate to support memory, language and decision-making, and how ageing and dementia risk degrade that coordination.

"By virtually mapping patient brains, clinicians would be able to run virtual trials for neuromodulation therapies, predict outcomes with mathematical precision and deliver targeted, highly effective treatments with minimal side effects," he said.

The work remains investigational. Zhou set a concrete timeline: "In five years, we hope to move from personalised brain modulation based on modelling and computer simulations to potential clinical applications." No device in this pipeline is FDA-cleared or clinically deployed; the current output is computational modelling and simulation, not patient-ready hardware.

His long-term goals are twofold:

• Establish sensitive early biomarkers derived from brain activity that flag abnormal ageing and neurodegenerative disease before clinical onset.
• Develop automated tools that monitor the efficacy of prevention and intervention measures over time.

Zhou was explicit about the limits of the computational approach. "In neuroscience, we use AI to model massive, highly non-linear brain activity that is difficult to understand," he said. "The technology is indispensable for making predictions, since the brain is biological and always changing — AI is our lens to understand this complexity. But to make a real clinical impact, it must always be paired with biological knowledge and medical reality."

For reading-room and monitoring-unit audiences, the relevance is in the modelling framework: network-level indicators of inter-regional coordination could eventually inform how treatment response is tracked, provided validation in patient cohorts follows the simulation phase.

A second award targets urban heat and air pollution

The same RGC round recognised a second HKBU researcher. Professor Gao Meng, an ecological engineer at the university's Academy of Geography, Sociology and International Studies, will receive about HK$8.5 million over five years from the 2026-27 Senior Research Fellow Scheme.

Gao applies machine learning to satellite and sensor data, downscaling global climate models into neighbourhood-level heat and flood forecasts for cities across southern China's Greater Bay Area. "We are building the first dynamic model for the GBA that simultaneously predicts how transport and energy changes can combat both extreme urban heat and air pollution," he said. The model integrates climate science with urban planning, and his team examines strategies including urban spatial redesign and the transition to new-energy vehicles.

Gao echoed Zhou's caution about pure data-driven prediction. "AI is a massive game-changer, but without incorporating physical laws, it can make physically impossible predictions," he said. "Our breakthrough comes from fusing AI with atmospheric physics, ensuring our predictions are both smart and physically accurate."

He expects cross-disciplinary work to define the coming decade: "Bringing together expertise from atmospheric science, physics, computer science, engineering and social sciences will allow us to develop more comprehensive solutions for sustainable cities and society."

If Zhou's five-year timeline holds, virtual-trial planning for neuromodulation — predicted outcomes, targeted delivery, minimal side effects — would move from simulation to first clinical applications by the end of the decade.