Quantification of FDG in the spinal cord using PET/MRI.

Background: In this study, we investigate the impact of MR-derived attenuation maps and limited detector resolution on the quantification of positron emission tomography (PET) activity uptake in the spinal cord during PET/MRI. This was performed by simulating [ ${}^{18}$ F]FDG PET data in the neck and thorax and then modifying the attenuation map to remove bone features. We then compared Ordered Subset Expectation Maximisation-reconstructed images to those with full attenuation correction. This simulation was performed at two detector resolutions of 2.1 and 4.4 mm. Acquisitions from a clinical study were then used to assess the ability of point spread function (PSF) modelling and time-of-flight (TOF) corrections, as implemented on the SIGNA PET/MR scanner (GE HealthCare), to correct for these quantification errors. For comparison, mean uptake was measured in regions of interest at each vertebral position along the spinal cord.

Results: Simulation results showed a decreasing pattern of uptake from the cervical to the thoracic spinal cord. When bone was not included in attenuation correction, the mean uptake decreased by 3%-10.4%. This difference in measured uptake was 6.4%-23.9% in images simulated at a detector resolution representative of a clinical PET/MRI scanner. At a detector resolution of 4.4 mm, a 32.2% decrease in uptake was measured compared to the 2.1 mm simulation. In patient data, introducing vertebral bone to the attenuation correction pseudo-CT led to a 1.8%-18.3% difference in ${\mathrm{SUV}}_{\mathrm{mean}}$ in the spinal cord. Applying PSF modelling did not lead to any statistically significant changes. TOF correction reduces the difference in ${\mathrm{SUV}}_{\mathrm{mean}}$ between data attenuation corrected with and without vertebral bone to 4.3%-7%. TOF Q.Clear images with beta = 100 showed the smallest difference between attenuation correction approaches at 0.6%-5.2%.

Conclusion: Ignoring bone during image reconstruction in PET/MRI reduces the activity measured during quantification of the spinal cord; however, the partial volume effect has a greater impact on reducing measured uptake in lower-resolution data. While time-of-flight correction goes somewhat resolves these quantification errors, further research is needed into partial volume correction.