Validating data-driven population-reference models (Braincharts, Bethlehem et al., Nature, 2022) with neuroimaging and clinical records for disease stratification and prognosis in neurodegenerative cohorts.
This work involves analyzing neuroimaging and clinical data from memory clinics at Addenbrooke’s Hospital and other NHS trusts around the UK, using artificial intelligence (AI) and normative models to develop individualized tools for patient stratification and prognostication in dementia.
The PASSIAN Project focuses on implementing federated learning in the NHS to create a secure, scalable clinical data-sharing solution. This work addresses a critical barrier to developing AI implementations for real-world biomedical data.
Using artificial intelligence (AI) tools in combination with clinical and biological data for individualized dementia diagnosis. This work involves expertise with data-driven approaches to neuroimaging in memory clinic cohorts (QMIN-MC and NACC), including data standardization, preprocessing, and harmonization.
Constructing and investigating Morphometric Similarity Networks from structural and diffusion weighted imaging data in a large, longitudinal cohort of patients with Parkinson’s Disease Psychosis.
Using multi-level meta-analytic tools to investigate if a specific profile of impaired cognition and visual function is linked to vulnerability to visual hallucinations in Parkinson’s Disease. The overall aim is to better understand the complex relationship between psychosis and cognitive decline in Parkinson’s patients.
Using graph theoretical approaches and resting fMRI data to better characterise the neural fingerprints of visual hallucinations in Parkinson’s Disease. This includes (i) evaluating group differences in FC in terms of both Von Economo cytoarchitectonic principles and well-established functional connectivity networks, (ii) NBS analyses, and (iii) machine learning approaches to identify patterns of covariance between rsfMRI networks and cognitive and clinical biomarkers of interest (cognitive tasks, MCI tests, cerebrospinal fluid biomarkers such as β-Amyloid, T-Tau and α-Synuclein)
Developing targets for understanding current treatments and developing novel treatments The aim of this pharmachological intervention is to implement and enhance a neuroimaging protocol to test for whole brain impairment in PD patients with and without psychosis to (i) enhance understanding of the neural basis of PD psychosis, (ii) estimate the magnitude of impairment both in predefined brain regions and across brain networks and (iii) test for drug effects (5-HT2a inverse agonism) in these networks.
This work focuses on the application of computational models of reinforcement learning (RL) in patients with schizophrenia, in addition to investigating the link between RL task performance and molecular genetic risk for the disorder.
Hallucinations can occur in different sensory modalities, both simultaneously and serially in time. Hallucinatory experiences occurring in multiple sensory systems—multimodal hallucinations (MMHs)—are more prevalent than previously thought and may have greater adverse impact than unimodal ones, but they remain relatively underresearched.