Medical Imaging

EPSRC Centre for Doctoral Training


106 - Multiparametric low-dose PET-MR for single-scan diagnostic imaging of memory clinic patients

1st Supervisor: Alex Hammers
2nd Supervisor: Andrew Reader

Aim of the PhD Project:

Develop a PET-MR imaging protocol capable of producing diagnostic quality FDG images for the assessment of memory clinic patients in a one-stop PET-MR examination, with massively reduced radiation doses (at least 10-fold, ideally 100-fold). 

Project description / background:

Dementia has been recognised as a major public health problem worldwide, with a prevalence of around 6.5% in the over-65s in England (e.g. Matthews F et al. Lancet 2013). Assessment at memory clinics includes clinical and cognitive assessment and structural imaging with MRI as standard. Among the additional tests (e.g. DAT SPECT, amyloid PET), FDG PET is probably the most versatile, able to provide differential diagnostic information and an indication of the degree of severity of abnormalities, but is often used late in the patient pathway. Its addition only after a patient has been discussed at a multi-disciplinary (team) meeting (MDM) with incomplete data leads to delays of typically 3-4 months before optimised management.

Patient care could be much improved, and patient pathways accelerated, by increased diagnostic confidence earlier in the management pathway. Current barriers to using FDG PET early on are organisational (the need for a second imaging appointment), financial (cost of the tracer, exam and reading), and regulatory (radiation burden ~5 mSv/study). We propose to address all of these points via a single-scan, ~35-min, PET-MR assessment with massively reduced patient doses (at least 10-fold; potentially even 100-fold).

Preliminary work:

Over the past six months and in close discussions with the industrial partner, we have determined that:

  1. Clear patterns (e.g. unilateral temporal lobe hypometabolism) can be easily detected at levels as low as 1 to 5% of the standard dose using MR-assisted PET image reconstruction (Figure 1)
  2. Post-smoothing of up to about 11 mm FWHM is compatible with clinically useable images
  3. Across unseen patients, MR-assisted reconstructions allowed acceptable image quality, acceptable to good/excellent pattern recognition, and acceptable confidence in diagnosis with as little as 1 to 5% of standard dose (Figure 2)

Figure 1: MR-assisted MAPEM reconstruction at 1% of dose. Unilateral temporal hypo-metabolism is clearly visible.

Figure 2: While image quality deteriorates markedly for standard analytical (FBP) and iterative (MLEM) reconstructions, MR-assisted reconstructions allow pattern recognition at massively reduced doses.