CBR Seminar Series: Ajay Halai
CBR Seminar Series: Ajay Halai, June 29, 2026
Title: Leveraging AI to improve diagnosis and prognosis in post- stroke aphasia
Abstract:
Aphasia affects at least one third of stroke survivors, yet the field has historically characterised it through syndrome-based taxonomies that poorly capture the graded, multidimensional nature of language impairment. My programme of research reconceptualises aphasia using a transdiagnostic approach — identifying latent dimensions of language and cognitive variation that cut across traditional diagnostic categories and map more faithfully onto underlying neural architecture.
Using large-cohort chronic post-stroke aphasia datasets, data-driven dimensionality reduction consistently reveals language and executive principal components, with convergent neural substrates identified through univariate and multivariate lesion-symptom mapping. This dimensional structure provides a framework for brain-behaviour prediction modelling. We evaluated how imaging modality, brain partition, and algorithm choice influence predictive performance, finding that structural lesion information largely captures the network-level disruption indexed by functional and diffusion connectivity.
A further challenge to translational application is lesion measurement itself. I present work benchmarking automated stroke lesion segmentation using a U-Net framework across heterogeneous, publicly available MRI datasets spanning acute and chronic stroke. Across DWI, FLAIR, and T1 modalities, models achieve accuracy approaching inter-rater human reliability, with lesion volume and image quality emerging as key determinants of generalisability.
Together, this work advances a cohesive framework: precise automated lesion quantification feeds transdiagnostic behavioural decomposition, which in turn enables machine learning prediction of individual patient profiles — offering a scalable pipeline for prognosis and stratification in clinical stroke research.
Bio:
I studied at University of York (BSc, MSc) and University of Manchester (PhD). I am now at the MRC Cognition and Brain Sciences Unit, University of Cambridge, originally as a Postdoc and continuing as a PI, running my own lab (Halai Aphasia Lab). I was awarded the MRC Career Development Award to build my own research group. I was also award the prestigious Elizabeth Warrington Prize by the British Neuropsychological Society for outstanding contribution to the field of Neuropsychology as an early career researcher. My research interests are related to brain mechanisms related to recover in post-stroke aphasia and utilising advanced neuroimaging and machine learning approaches to improve diagnosis and prognosis.
Watch the Recording