My work is in data science, with a particular background in health-related applications. I have an MSc in Data Science & Business Analytics and a PhD in Health Data Science, and I have experience applying Python, R, machine learning, statistical modelling, and reproducible workflows to complex data problems.
Much of my recent work has focused on medical imaging and explainable AI, but the underlying skills are broader: data preparation, model development, evaluation, documentation, and communication with technical and non-technical stakeholders.
I am especially interested in work that uses data carefully and usefully, whether that is in healthcare or in other settings where clear analysis and sound judgement matter.
Selected outputs
Three Minute Thesis
I won the University of Plymouth’s Three Minute Thesis competition with a short talk on my doctoral research. It was a good opportunity to explain a technical project to a general audience, and it remains one of the clearest summaries of my work:
Doctoral thesis
My doctoral thesis, Explainable Deep Learning for Medical Imaging Classification, explores explainable deep learning in several medical imaging tasks. It covers MRI safety, confounding pathology detection in radiology report text, and Parkinson’s-related MRI applications, with SHAP used to help explain model predictions.
Parkinson’s disease
A major strand of my PhD work focused on Parkinson’s disease and brain imaging. I presented this work at the International Congress of Parkinson’s Disease and Movement Disorders, and it also fed into my thesis and wider research communication.
Stroke risk modelling
More recently, I have worked on stroke-related prediction problems, including machine learning to predict stroke risk from routine hospital data and an AI-based stroke risk factor classification and treatment study. This work reflects my interest in applying machine learning to clinically meaningful questions using real-world data.
Radiology report classification
I also worked on explainable transformer-based classification of radiology reports, which was published in BJR|Artificial Intelligence. This project combined natural language processing with interpretability, and it fits closely with my interest in transparent model development.
MRI safety
Another example is my work on deep learning detection of aneurysm clips for MRI safety, published in the Journal of Digital Imaging. This was a useful example of applying machine learning to a practical clinical safety problem.
How I work
Across these projects, I have aimed to work carefully and transparently: using Python and reproducible workflows, writing documentation, evaluating models critically, and paying attention to governance and sensitive data. I have also worked with NHS and university stakeholders, which has helped me keep the work grounded in real-world use rather than treating it as a purely technical exercise.
Publications and links
For a fuller record of my work, see my publications, ORCID profile, and CV. If you are mainly interested in a quick overview, the Three Minute Thesis video and the selected outputs above are probably the best starting points.
Additional roles
I also hold a number of voluntary and committee roles, including membership of Parkinson’s UK’s Involvement Steering Group, service as Churchwarden at St Petroc’s, Harford, and committee/publicity work with Brent Singers. These roles reflect my interest in public involvement, practical responsibility, and contributing to the organisations and communities I care about.
Cryptic Meg 