We build the core computational methods that power the lab's work, from fractal geometry and deep learning to signal processing and the infrastructure that turns data into discovery.

Neuromethods is the methodological engine of the lab, the place where the algorithms, models and infrastructure that power everything else are developed and refined. Rather than being tied to a single clinical question, this domain builds the general-purpose computational tools that our imaging, pathology and surgical work depend on. A signature strand is fractal geometry, a long-standing line of the lab's work that quantifies the self-similar complexity of the brain and its structures, from vasculature to tumours. Alongside this we develop deep-learning architectures, signal-processing methods for neural data such as MEG, and the frameworks, including transfer learning, generative and foundation-model approaches, that let models learn effectively from the limited, heterogeneous data typical of clinical neuroscience. This domain also encompasses the computational resources and pipelines that make the science reproducible and scalable, from local high-performance compute to national supercomputing infrastructure.