Machine learning for hyperspectral computed tomography

Spectral/hyperspectral tomography aims to improve standard Computed Tomography by taking advantage of additional information in the form of energy-resolved measurements. The objective is to use spectral information to reduce artifacts and improve the image reconstruction in cases of limited-angle projections. To this end, this project will explore the application of machine/deep learning methods in combination with classical and Bayesian inversion methods. Applications will include non-destructive testing of materials and security screening. This project will run in collaboration with Harris Corporation and colleagues at the Alan Turing Institute.

Your responsibilities:
• explore ideas, methods and algorithms in the use of machine learning techniques for hyperspectral CT
• implement algorithms in programming languages (Python, MATLAB)
• present results at international conferences and in scientific publications
• work in a team with other PhD students, postdocs, and staff.

Your qualifications:
• Master’s degree (or equivalent) in computational science or scientific computing or similar degree with a focus on applied mathematics and computing
• knowledge/experience in inverse problems is an advantage
• interest/experience in machine/deep learning
• programming skills in Python or MATLAB
• fluent in spoken and written English

Our group’s research interests are in computational modelling of physical and engineering systems and inverse problems for imaging, process monitoring, and non-destructive testing. This area entails mathematical modelling, statistical inference and optimisation algorithms. For more info see http://www.homepages.ed.ac.uk/npolydor/

Start date: September 2019