So far, the [18F]FDG uptake in normal and diseased spinal cord was evaluated in PET/CT studies [2, 3, 4, 10, 11, 12, 13, 14, 15, 16]. It is recognized that clinical PET/MRI could be advantageous with respect to PET/CT for assessment of spinal cord diseases . PET/CT studies have shown inconsistent results about the impact of gender, age, body mass index and blood glucose on SUVmax of the spinal cord [3, 10, 11, 14, 17]. However, in pediatric population [18F]FDG uptake of the spinal cord increased with body weight .
We are aware of several technical and methodological differences when the average SUV values we measured in PET/MRI are compared with those measured in PET/CT by us (and others). We administered an activity of 3.5 MBq/kg body weight that is lower to those used in prior [18F]FDG/PET studies (4.8 – 5.2 MBq/kg of body weight ). However, while a lower signal to noise ratio can result from lower activity, no effect on SUV is expected . In our protocol we subsequently acquired PET/CT and PET/MRI. While this may imply lower PET values in PET/MRI, the duration of acquisition was reported not to affect SUVmax values in case of spinal cord tumors  and healthy tissue considering that the time activity curve of normal organs (liver and spinal cord) has a slow and linear decay over time [20, 21].
Our method for ROI computation of the spinal cord [18F]FDG uptake values in PET/CT was different from that employed in prior PET/CT studies. In fact, we coregistered in PET/CT the ROIs that we drew on MRI taking advantage of the spinal cord direct visualization. In prior PET/CT studies the spinal cord uptake refers to the SUVmax of a ROI that is drawn along the contour of the spinal canal with care not to include the cortical vertebral bone. This notwithstanding, Patel et al.  emphasized the need of correcting for bone marrow contamination when reporting [18F]FDG uptake of the spinal cord. Therefore, our spinal cord ROI annotation procedure could be more accurate with respect to prior studies.
All the above differences considered, it is noteworthy that the SUVmax values of the normal spinal cord we measured in PET/CT as at each spinal level or on average over the entire spinal cord were similar to those measured in other PET/CT studies [3, 4, 10, 11, 14, 17]. This notwithstanding, we generally measured higher average [18F]FDG uptake values (computed as values at levels C2, C5, T6 and T12 /4) of the spinal cord in PET/MRI than in PET/CT. Moreover, it is conceivable that if a direct PET/MRI acquisition could have been performed without prior PET/CT, [18F]FDG uptake values of the spinal cord in PET/MRI in our patients could even be higher than those we measured.
Remarkably, the longitudinal distribution of SUVmax and NSUVmax of the spinal cord we measured in PET/MRI is in line with that previously reported in PET/CT [2, 3, 10, 14, 17]. In particular, using 3 mm ROI we observed the highest SUVmax values at vertebral level C5 and the lowest values at vertebral level T6. This feature has been attributed to the variable amount of nervous tissue and in particular of the metabolically more active central gray matter in the spinal cord [2, 17]. In fact the cross-sectional area of the spinal cord decreases from C1 to the conus medullaris with the exception of cervical (from C3 to T2 vertebral body) and lumbar (from the T9 to T12) enlargements, where the transverse diameter increases due to the relative expansion of the gray and white matter that is associated with a greater number of neurons correlated to upper and lower limbs sensory and motor functions. In particular, the C4 to T1 spinal cord neural metamers, corresponding to the vertebral body level from C3 to T2, show a mean cross-sectional area of gray matter from 7.8 to 10.7 mm2 and the L3 to S1 spinal cord neural metamers corresponding to vertebral body level from T9-T10 to L1-L2 show a mean cross sectional of gray matter of 13.2 to 16.7 mm. Differently, the remainder thoracic spinal cord neural metamers show a range from 3.7 to 5.6 mm2 of mean cross section of the gray matter [22, 23]. Accordingly, the SUVmax of the [18F]FDG uptake of the spinal cord would reflect different segmental (metameric) levels of specialization and demands with higher values in segments involved in sensory and motion functions of the limbs and lower values in thoracic segments mainly involved in the functions of thoracoabdominal visceral organs and sensory and motion functions of the trunk.
The significant correlation between spinal cord and bone marrow [18F]FDG uptake values we observed in both PET/MRI and PET/CT is in line with the PET/CT data by Patel et al. . A distinct advantage provided by PET/MRI as compared to PET/CT is the possibility of reliably measuring the SUVmean, namely the mean [18F]FDG uptake of the cross-sectional area of the spinal cord that is not visible in PET/CT. As expected, inclusion of the cord white matter with its lower metabolism/glucose consumption in the ROI implies lower SUVmean values as compared to the SUVmax values which essentially reflects the higher metabolism/glucose consumption of the central gray matter. However, SUVmean and NSUVmean may be of potential interest for spinal cord diseases affecting predominantly the cord white matter as multiple sclerosis, Friedreich’s ataxia and VitB2 deficiency.
The longitudinal distribution of SUVmean an NSUVmean we observed closely matches that of SUVmax and NSUVmax. So far, the [18F]FDG uptake in PET/CT was measured in spinal cord tumors that are rare conditions in which a pattern of variably increased [18F]FDG uptake was reported [12, 17]. [18F]FDG uptake in PET/CT of the spinal cord have also been applied to the evaluation of inflammatory myelopathies, including multiple sclerosis, neurosarcoidosis, and of amyotrophic lateral sclerosis [12, 13]. While a pattern of decreased or increased [18F]FDG uptake was observed in inflammatory myelopathies and neurosarcoidosis, presumably reflecting varying delay between disease onset and time of PET and possible interference with steroid therapy [12, 13], in case of amyotrophic lateral sclerosis elevated [18F]FDG uptake values were reported . Additional potential diseases suitable to be evaluated with PET of the spinal cord include trauma, post-radiation myelopathy, Vitamin B12 deficiency myelopathy and some neurodegenerative diseases which primarily affect the spinal cord as Friedreich’s ataxia and hereditary spastic paraplegia. We anticipate that in these conditions both SUVmax and SUVmean could provide valuable information because the cord white matter is frequently more affected than the central gray matter.
An issue addressed in the present study was the possible effect of normalizing the spinal cord SUV max and mean values to the liver [18F]FDG uptake. This type of normalization was originally proposed by Marini et al.  for spinal cord SUVmax in order to pool data provided by two PET/CT scanners. In our instance the normalization was justified by the need to overcome possible differences in [18F]FDG uptake between PET/CT and PET/MRI due to the time interval between the two acquisitions and the tracer decay. Remarkably, all the results we obtained by analyzing SUVmax and SUVmean values in PET/MRI were consistent when the data were normalized to the liver, considering the different decay rate between spinal cord and liver, with a quicker one for the liver. This suggests that liver normalization might improve stability of results and might be useful for multi-center studies.
We recognize the following limitations of our study.
We evaluated the spinal cord [18F]FDG uptake in a population of adult subjects with lymphoma rather than of healthy subjects. However, the patients we enrolled had neither neurologic symptoms and signs nor MRI evidence of any spinal cord or marrow abnormality. Moreover, they were examined for initial staging purposes and had not received yet chemo or radiation therapies. The sequential acquisition of PET/CT and PET/MRI could have affected spinal cord uptake values in the PET/MRI. Future PET/MRI studies of neurological conditions could avoid preliminary PET/CT acquisition that was requested by Health Authorities in Italy at the time of the present study. In addition, it is worth mentioning here that current MR-based AC techniques, such as the one used in the present work, produce slightly biased estimates relative to CT-based AC [6, 25]. Therefore, the specific impact on SUV quantification of the spine and spinal cord should be evaluated and taken into account in further studies.
We did not control for factors that in some prior studies were reported to potentially modify the [18F]FDG uptake such as lean body mass, body surface area, serum glucose, insulin, free fatty acid, and exact time to acquisition. However, these factors are generally considered crucial in [18F]FDG repeat studies and not in cross-sectional studies like ours.
Finally, we used a whole-body protocol instead than a dedicated spinal MRI-PET. The latter could allow better morphologic definition of the spinal cord with improved co-registration and especially correlation with sequences for volumetric and microstructural assessment of the spinal cord. Of course, the combination of metabolic information from [18F]FDG PET with macro and microstructural information with high spatial resolution from MRI in the same examination will represent a valuable tool for improved understanding of the physiopathology of spinal cord diseases and might constitute a valuable surrogate marker for future trials. Also, exploration of other radiotracers than [18F]FDG may expand the role of spinal cord PET.