Research

Current Work

PhD Research - Visual system vulnerability in dementia: from detection to determinants

Supervised by Dr Keir Yong (UCL Dementia Research Centre) & Prof Andre Altmann (UCL Hawkes Institute)

Cortical visual impairments (“brainsight”, not eyesight) are disabling yet under-recognised consequences of dementia. They are common in Alzheimer’s disease (AD) and especially posterior cortical atrophy (PCA) (visual-led dementia), where visual symptoms can precede memory, language and insight loss. Many people first present to eye care and are misdiagnosed with ocular or psychological, leading to unnecessary interventions and years-long delays. Clinic-ready tests of cortical visual function are scarce, and there is limited understanding about why some individuals are more susceptible.

Projects

1. Cortical visual test (GILT): Validate the Graded Incomplete Letter Test across UK Biobank repeat-imaging and clinic cohorts; relate scores to diagnosis/MRI; optimise sensitivity to visual-cortex damage.
2. Latent factor analysis: Combine visual tests + MRI to model a cortical visual factor; test associations with AD risk and functional measures.
3. Discovery genetics: Perform GWAS on the latent factor; compare SNP effects with visual-led AD/PCA.

Broader Research Interests

• Computational neuroscience
• Bayesian and statistical modelling
• Genomics and multi-omics
• Computational antimicrobial resistance (AMR)
• Multimodal and representation learning
• Biomedical imaging
• Interpretability and Explainable AI (XAI)

Academic Background

PhD Training Modules (Year 1 - 2025/2026)

Term 1

• CLNE0029: Clinical Neuroscience of Dementia
• COMP0187: Probabilistic Modelling
• COMP0171: Bayesian Deep Learning
• COMP0172: Artificial Intelligence for Biomedicine and Healthcare
• MPHY0041: Machine Learning in Medical Imaging
• IEHC0036: Genomics, Health and Society

Term 2

• CHME0034: Applied Computational Genomics
• COMP0197: Applied Deep Learning
• COMP0118: Computational Modelling for Biomed Imaging
• MPHY0050: Applied AI in Medical Imaging

Previous Research Projects

(MSc Dissertation) Deep Learning Approaches to Misinformation Detection

Built and evaluated end-to-end NLP systems for social-media misinformation: data collection, feature engineering, and text vectorisation. Compared traditional ML (logistic regression, SVMs, random forests, gradient boosting), deep learning (CNNs, LSTMs), and LLM-based transformers. Assessed both ordinal (graded truthfulness) and binary (true/false) labels to test how models capture factual nuance. Findings: retaining ordinal labels for training improved accuracy by ~10%, and strong feature engineering added a further ~8% gain. LLM-driven augmentation (richer sentiment/subjectivity signals) further lifted a binary baseline from ~80% to ~85%, indicating better nuance capture. The trade-off was increased compute and latency, challenging real-time use without optimisation.

(BSc Dissertation) M-Estimation: Estimation and Inference from Statistical Models in Julia

I developed statistical computing tools for robust inference in Julia, focusing on M-estimators, which generalise maximum likelihood and, with appropriate robust loss functions (e.g., Huber), reduce sensitivity to outliers and modest misspecification. A core part was leveraging dual numbers - numeric types that extend real numbers with an infinitesimal $\varepsilon$ where $\varepsilon^2 = 0$ - to enable forward-mode automatic differentiation, which lets us get derivatives accurate up to floating-point precision (forward-mode AD) and Jacobians of large objective functions efficiently instead of using finite-difference numerical derivatives, improving optimisation speed and stability. I extended an M-estimation package in Julia with profile likelihood: fixing a parameter of interest, re-optimising nuisance parameters, and tracing that profile to get inference and identifiability diagnostics.

Grants & Awards

• Competitive 4-year doctoral studentship funded by the Engineering and Physical Sciences Research Council (EPSRC) and The National Brain Appeal, supporting PhD research on “Visual system vulnerability in dementia: from detection to determinants” within UCL Mechanical Engineering. (Sep 2025 - Sep 2029)
• MSc Project Award. Awarded for best dissertation research among MSc cohort. (Sep 2024)
• Senior Mathematical Challenge - Gold Certificate. Awarded to the top 10% of participants nationally in the UK Mathematics Trust (UKMT) Senior Mathematical Challenge - a prestigious problem-solving competition testing advanced mathematical reasoning, logical thinking, and creativity beyond the school curriculum. (2018 & 2019)

Academic Profiles

• ORCID
• Tech4Health CDT Researcher Web Page

Affilations/ Institutes

• UCL Dementia Research Centre
• UCL Hawkes Institute
• UCL Mechanical Engineering
• Tech4Health CDT
• UKRI Engineering and Physical Sciences Research Council (EPSRC)
• The National Brain Appeal

Academic Events

Table numbers correspond to event markers on the Google Map. Open the side panel to see them.