Overview
I joined the Cardiff School of Engineering as an EPSRC and Velindre Cancer Centre funded research student in 2015. After completing my PhD, I joined the Cancer Imaging and Data Analytics research group as a research associate. My research focuses on the segmentation of the tumour on PET imaging, as it has been identified as one of the largest sources of error in radiotherapy treatment planning. My research aims to improve the quality of radiotherapy treatment planning by investigating methods that can improve tumour segmentation.
Biography
Work & Education
- Research Associate, Cardiff University 2019 - Current
- PhD, Cardiff University 2015 - 2019
- Data Interface Developer, Travel Places 2014 - 2015
- MSc Computer Science, Bangor University 2013 - 2014
- BSc Computer Science, Bangor University 2010 - 2013
Academic positions
2013 - 2014: Postgraduate Demonstrator, Bangor University
Speaking engagements
- European Association of Nuclear Medicine, Virtual (2020)
- NCRI Virtual showcase (2020)
- European Association of Nuclear Medicine, Dusseldorf (2018)
- All Wales Medical Physics and Clinical Engineering Summer meeting, Cardiff (2018)
- All Wales Medical Physics and Clinical Engineering Summer meeting, Cardiff (2017)
- All Wales Medical Physics and Clinical Engineering Summer meeting, Cardiff (2016)
Committees and reviewing
- Journal Reviewer, Physics in Medicine & Biology
- Journal Reviewer, European Journal of Nuclear Medicine & Medical Imaging Research
Publications
We are currently unable to retrieve the list of publications. Visit our institutional repository.Teaching
Lecturer on the Medical Engineering (MEng) Medical Image Processing module.
Segmentation in radiotherapy treatment planning
In radiotherapy planning, the segmentation of the tumour has been highlighted as one of the greatest sources of error. This is a manual process, which is not only time extensive but also subject to inter and intraobserver variability. Although automated segmentation algorithms have also been proposed for accurate tumour segmentation, they have limited accuracy in PET scans acquired after chemotherapy and have been shown to be not suitable in all cases. Therefore, there is increasing interest in using machine-learning techniques to improve the segmentation of the tumour. ATLAAS is decision-tree based segmentation methodology that was validated in diagnostic H&N cancer and externally validated in intra-treatment PET scans. It uses estimated tumour characteristics to select one of many segmentation algorithms included in the training dataset. In comparison to segmentation performed on a CT image, PET imaging and the subsequent results of segmentation have poor resolution. Therefore, there is interest in improving the quality of the segmentation. Interpolation of PET imaging before segmentation with ATLAAS reduces the aliasing artefacts and improves tumour segmentation accuracy.
Currently, deep-learning techniques are showing promise for accurate tumour segmentation. However, they have not been thoroughly investigated. My current research is looking at deep-learned segmentation of the tumour in comparison to traditional machine-learning techniques.
Projects
PEARL: Adaptive radiotherapy planning September 2020 - Current
ASPIRE: AI Solutions for Personalised Radiotherapy (Efficiency Through Technology programme) April 2019 - September 2020
TITAN: Advanced Automated Segmentation of PET imaging for use in Radiation Therapy October 2015 - April 2019
Collaborations
MAASTRO, Netherlands 2017 - current
Guy's and St Thomas Hospital, London 2017 - current
Velindre Cancer Centre, Cardiff 2015 - current