My research integrates data science, signal processing, machine learning, control theory, and computational neuroscience to study and engineer brain systems. I focus on developing methods that combine principled mathematical models with large-scale data, enabling neural dynamics to be interpreted, predicted, and acted upon in clinically meaningful ways. The primary application domains of this work include epilepsy, anaesthesia, neuroimaging, and neural engineering, with an emphasis on reproducibility, scalability, and real-world clinical impact.
Research Pillars
This research program is organized around five interconnected pillars:
• Model-based depth of anaesthesia monitoring
• Seizure prediction and forecasting
• Deep brain stimulation and seizure control
• Estimation and control of neural systems
• Brain networks and consciousness under anaesthesia
Model-Based Depth of Anaesthesia Monitoring
Focus
Understanding and monitoring brain state during general anaesthesia using EEG.
Methods
This work combines classical time-series models, such as autoregressive moving average (ARMA) representations, with physiologically motivated neural mass and neural field models. These models are integrated with real-time state estimation and filtering techniques to infer latent brain dynamics from noisy EEG measurements.
Outcomes and Impact
The primary clinical goal is to improve the reliability of anaesthesia monitoring, reducing the risk of intraoperative awareness while also avoiding excessive anaesthetic dosing that may contribute to postoperative cognitive impairment. This research aims to bridge the gap between theoretical models of brain dynamics and practical monitoring tools used in operating theatres.
Collaborations
This work is conducted in collaboration with clinicians, biomedical engineers, and anaesthesia researchers, ensuring strong alignment between methodological development and clinical relevance.
Seizure Prediction and Forecasting
Focus
Forecasting seizure risk over time using ultra-long-term intracranial EEG recordings.
Methods
Rather than treating seizures as isolated events, this research models epilepsy as a dynamic disease process with slowly varying susceptibility. It combines large-scale feature extraction from continuous EEG data with physiologically grounded models that capture long-term rhythms and state transitions. Particular attention is given to robust biomarkers such as critical slowing and circadian or multiday modulations.
Outcomes and Impact
The goal is to provide clinically useful seizure forecasts that support patient decision-making and enable adaptive therapeutic interventions. A key outcome of this work is a strong commitment to reproducibility, embodied by the development of Epilepsyecosystem.org, an open platform for collaborative seizure prediction research.
Collaborations
This research involves international collaborations across neuroscience, engineering, and clinical epilepsy communities.
Deep Brain Stimulation and Seizure Control
Focus
Control-oriented approaches for epilepsy using implantable neurostimulation devices.
Methods
This work explores how estimation and forecasting methods can be integrated with implantable systems, such as the Medtronic PC+S device, to enable adaptive and potentially closed-loop deep brain stimulation. Control theory and signal processing methods are used to design stimulation strategies that respond dynamically to changes in brain state.
Outcomes and Impact
The long-term objective is to move beyond fixed stimulation protocols toward intelligent neuromodulation systems that personalize therapy in real time, improving efficacy while minimizing side effects.
Collaborations
This research is carried out in close collaboration with clinical partners and industry stakeholders involved in neuromodulation technologies.
Estimation and Control of Neural Systems
Focus
Fundamental limits of inference and control in complex neural systems.
Methods
This line of work develops observer designs, filtering techniques, and observability analyses for nonlinear, high-dimensional neural models. A key question is what aspects of brain dynamics can be reliably inferred from limited and noisy measurements, and under what conditions effective control is possible.
Outcomes and Impact
The results provide theoretical foundations that support applied work in brain monitoring, stimulation, and neural engineering, while also contributing to broader understanding of inference in biological systems.
Collaborations
This research is primarily methodological, with links to applied projects in epilepsy, anaesthesia, and neural control.
Brain Networks and Consciousness under Anaesthesia
Focus
Network-level mechanisms of loss and recovery of consciousness.
Methods
Using EEG and magnetoencephalography (MEG), this work applies network analysis and dynamical systems approaches to study how functional connectivity patterns change under different anaesthetic agents.
Outcomes and Impact
A central question is whether a common “backbone” network structure underlies changes in consciousness, independent of the specific anaesthetic used. This research contributes to fundamental neuroscience as well as the development of interpretable markers of brain state during anaesthesia.
Collaborations
This work involves collaborations with neuroimaging researchers and neuroscientists studying consciousness and brain networks.
Selected Grants Supporting This Research
• NHMRC Ideas Grant — Data-driven and model-based methods for seizure forecasting and neural state estimation
• ARC Discovery Project — Modelling, observability, and control of large-scale neural systems
• NIH Collaborative Funding — Reproducible seizure prediction using long-term intracranial EEG
• Clinical and Industry Partnerships — Translational research in anaesthesia monitoring and neuroengineering