PhD Studentship: Efficient Machine Learning Approaches for X-ray Tomographic Image Reconstruction with Limited Training Data

Supervisory Team: Thomas Blumensath (ISVR) and Richard Boardman (Engineering)

Project description

Machine learning has revolutionised many scientific fields over recently years and has become an increasingly useful tools in a wide range of image processing applications; including in X-ray tomography. X-ray tomography is an imaging technique that utilises X-ray radiation to generate three-dimensional image of internal object structures. The technique is thus used routinely in medical diagnostics, security screening as well as scientific investigations.

In many X-ray tomography applications, constraints on the imaging process mean that we are often only able to collect limited X-ray measurements, which can lead to significant image noise and artefacts. Many advanced machine learning methods have thus been proposed to reduce these errors. The problem is however that this typically requires significant amounts of data to train these models, data that is not always available in many applications.

On the other hand, more traditional image estimation methods that have been proposed for X-ray tomographic data, whilst not requiring large amounts of training data, instead typically require careful and laborious tuning of parameters, which, due to the extremely large size of typical X-ray tomography data-sets, leads to very complex and slow fine tuning problems. These approaches  are thus nearly never utilised in real applications.    

In this project you will combine advance machine learning based X-ray tomography image estimation methods with more traditional approaches. Utilising our dedicated high performance computing hardware designed specifically for Machine Learning based X-ray tomographic image estimation, you will work on simulated and real X-ray tomographic data to develop advanced computational methods that can reduce X-ray image noise and artefacts without the need for parameter tuning or large sets of training data.

Whilst the project will be predominately computational, there will also be the chance to work closely with the University of Southampton’s dedicated X-ray Computed Tomography (X-CT) centre “µ-VIS”, which is part of the UK’s National facility for X-CT. The centre houses some of the UK’s largest micro-focus CT scanning systems with the capability to unveil sub-surface information from an extremely wide range of materials, components and structures. With strong links between both research and industry, the centre is used for an extensive list of applications (please see our website for further info: www.muvis.org), which will offer many opportunities to apply your innovations directly to a host of relevant scientific and industrial imaging challenges. 

Potential funding to support this position will be available to the strongest candidates through the Faculty of Engineering and Physical Sciences graduate school studentship programme, which are awarded on a competitive basis.

Entry Requirements

A very good undergraduate degree (at least a UK 2:1 honours degree, or its international equivalent, with internal studentship funding likely to require a first class honours degree or equivalent).

Funding for: Home and International Students

Funding amount:

The studentship is supported for 3.5 years and includes tuition fees plus a stipend of £18,110 per annum (2023/24 rate)

Closing date: 31 March 2024

How To Apply

Apply online: Search for a Postgraduate Programme of Study (soton.ac.uk). Select programme type (Research), 2024/25, Faculty of Engineering and Physical Sciences, next page select “PhD Engineering & Environment (Full time)”. In Section 2 of the application form you should insert the name of the supervisor Thomas Blumensath

Applications should include:

• Curriculum Vitae
• Two reference letters
• Degree Transcripts/Certificates to date

For further information please contact: feps-pgr-apply@soton.ac.uk

Advert information