Image Reconstruction for Spinal Cord PET/MRI

Despite some neurological conditions also affecting the spinal cord, research using
PET/MRI has focused on brain imaging. Spinal cord MRI is common due to its
high resolution and soft tissue contrast. PET, however, presents challenges due
to its low resolution relative to the spinal cord’s diameter. In PET/MR, spinal
cord PET may also be affected by incorrect attenuation correction due to a lack of
vertebral bone in MR-derived attenuation maps.
The impact of these effects and possible solutions were determined through a
series of studies. A literature review established whether currently available physical
phantoms were suitable for assessing the performance of PET/MR scanners.
It was found that replicating tissue properties for both modalities within the same
phantom was challenging, so typically one modality is prioritised.
Phantom experiments were performed using a PET performance phantom and
an anatomical ‘head and neck’ phantom to assess the performance of the SIGNA
PET/MR Scanner. It was found that for features less than 20mm, the size and
tracer uptake in these regions was underestimated in PET images.
A simulation study is presented exploring the impact of MR-derived attenuation
maps and PET detector resolution on measured tracer uptake in the spinal
cord. This was compared to patient data. The results indicate that whilst tracer
uptake is underestimated when vertebral bone is not accounted for in attenuation
maps, partial volume effects due to PET detector resolution have the greatest
impact.
MR-guided reconstruction of PET data was investigated as a solution to resolving
partial volume effects during PET image reconstruction, using MR images
simultaneously acquired with PET data. Images reconstructed with MR-guided
algorithms showed an increase in measured tracer uptake in the spinal cord compared
to existing reconstruction algorithms. This suggests that MR-guided reconstruction
of PET data can contribute to resolving partial volume effects.