Developing reliable quantitative biomarkers for improving the utility of Low-Field MR in the study of disease
Supervisors: Dr N Dowell, Dr Samira Bouyagoub, Dr I Simpson
Application deadline: Friday 16 June 2023
Applicants for this 3-year Brighton and Sussex Medical School-funded PhD starting in September 2023 should possess or expect to be awarded a minimum of a First or Upper Second Class Honours degree (or equivalent) in a relevant discipline.
This project will develop novel low-field magnetic resonance imaging (LF-MRI) acquisition and analysis methods to improve both the signal-to-noise ratio (SNR) and contrast within a feasible acquisition timescale. These developments will enable low-cost and low-power quantification of common imaging biomarkers including morphometry and detection of white matter hyperintensities. We will explore the efficacy of these developments in healthy controls and subjects with HIV/AIDS, comparing accuracy with paired images of the same subject on a 3T scanner in the imaging centre.
MRI is an indispensable technology, enabling substantial advances in science and medicine. Commercial MRI machines that operate at high magnetic fields (1.5-3 Tesla) are expensive to install (£2-3 million) and maintain (>£100,000 p.a). Consequently, the number of scanners is limited, especially in less developed countries.
Recently, interest has grown in LF-MRI using field strengths 60x lower than a typical clinical MR scanner. Here, the dangers posed by the magnetic field are effectively eliminated while the scanner itself is small enough to be portable. However, clinical utility of LF-MRI is currently limited by poor (SNR) and low image contrast. In the brain, grey/white matter boundaries are indistinct leading to difficulties in differentiating brain areas through image segmentation.
The successful applicant will work with the supervisory team to tackle these challenges. For example, to improve SNR, we will implement a high SNR acquisition technique: steady-state free precession (SSFP). This will afford higher spatial resolution and permit the detection of white matter hyperintensities and improve reliability of volume measurements. However, SSFP suffers from image artefact on LFMR scanners due to poor magnetic field homogeneity. This can be addressed through phase cycling, where repeated images are acquired with different parameters. Such approaches offer the potential to employ data-driven compressed sensing approaches, which will further reduce the scan-time. Moreover, we can take advantage of the multiple images to reduce the effects of image noise and distortions.
We will improve image contrast by implementing MR fingerprinting, which generates unique signal evolutions (“fingerprints”) at every location in the image; matching these with computer simulations reveal parameters such as T1 and T2. With this information, contrast may be augmented so that the images are more familiar to radiologists and more conducive to segmentation.
This project would suit a student with an interest in medical imaging and magnetic resonance imaging physics and the computation aspects will appeal to those with good mathematical and programming skills. Students will be expected to present their work at top-tier MR physics, medical image analysis, computer vision or machine learning venues such as ISMRM, ENC, MICCAI, IPMI, CVPR, ICCV/ECCV, NeurIPS etc.
This project is jointly supervised by Dr Ivor Simpson, a lecturer in the Predictive Analytics Lab within the AI research group at the University of Sussex and Dr Samira Bouyagoub, MR Physicist at the Clinical Imaging Sciences Centre at Brighton and Sussex Medical School. Both institutions are situated within the fantastic south coast city of Brighton and Hove, adjacent to the beautiful South Downs national park.
Home fees will be paid for UK citizens; non-UK citizens will be liable for the difference in fees between the rate for home students and the overseas student rate. Applicants whose first language is not English are expected to meet the minimum requirements (7.0 IELTS). Informal enquiries should be directed to Dr Nicholas Dowell (N.G.Dowell@bsms.ac.uk).
APPLY FOR YOUR PHD HERE >