Supervisor: Professor Jian Zhang, School of Mathematics, Statistics and Actuarial Science, University of Kent, Canterbury
Brain injury often occurs in our community such as sport clubs and veteran clubs, which may have wide-ranging physical and mental effects on victims. This project, considering manifold structures of brain, aims to develop manifold learning tools for diagnosis of brain conditions by integrating scan data generated from non-invasive imaging devices such as structural magnetic resonance imaging (MRI), diffusion weighted imaging (DWI), and diffusion tensor imaging (DTI).
Most of the existing clinical usage of these devices in hospitals and health centres is based on T1-weighted MRI and DTI. The current diagnosis suffers from important limitations: for T1-weighted MRI and DTI data, the existing feature extracting approaches rarely explore structural constraints. Scan data often lie on or near an intrinsic manifold, taking values in an ambient space which is not necessarily Euclidean. There is a growing need for using geometric structures in data analysis. Identifying these underlying manifold structures can help improve the accuracy of denoising and classifying observations.
This project will address the above challenges through the development of novel machine learning approaches such as manifold learning, reproducing Hilbert kernel-based prediction and nonparametric inference of neural differential equations in an infinite dimensional space. The overall aim of the project is to produce signatures (or biomarkers) that extract robust and clinically useful information from these image scans. Imaging and other related data from the Human Connectome Project and the Parkinson’s Progression Markers Initiative will be employed to evaluate the proposed manifold approach.
The project provides unique opportunities for cross-disciplinary training in innovative methodologies at the interface of neuroscience and statistical machine learning. This is a computational project, which would suit a student with good mathematical and programming skills in using Python and Matlab, and a keen interest in medical imaging.
Alongside completing your PhD programme of research and development, as a GTA you will normally be expected to work for 200 hours per annum in years 1 to 3, including teaching (maximum 96 contact hours per year) or demonstrating (maximum 130 contact hours per year) and related duties such as marking, preparation and examination. Further details of GTA terms and conditions are here: https://www.kent.ac.uk/scholarships/postgraduate/terms-and-conditions-gtas
Entry requirements: Eligible applicants should have recently received a good MSc qualification and a first or high 2:1 undergraduate in statistics/computing/applied mathematics or have substantial recent experience working in a large scale data analysis or imaging analysis.
Scholarship value: The scholarship includes home/overseas tuition fees plus a combined maintenance grant and salary equivalent to the Research Councils UK National Minimum Doctoral Stipend (£18,662 for 2023/24) for the first three years followed by fees and a maintenance grant for a further six months. Scholars also receive fee-paid teacher in Higher Education training through the Associate Fellowship Scheme.
Closing date: 15 March 2024
Start date: September 2024
Eligibility: UK/Overseas
Application procedure: Apply for a Statistics PhD at IPP login screen (kent.ac.uk) and specify the research topic “Manifold Models for Neuroimaging Data”