Spinal cord FDG PET uptake in healthy adults: an AI-driven segmentation study

¹⁸F-fluorodeoxyglucose (FDG) Positron Emission Tomography (PET)/Computerized Tomography (CT) is an important imaging modality in oncology, but its use in spinal cord assessment is limited due to a lack of baseline metabolic reference values. Regional spinal cord FDG uptake, especially in healthy populations, remains poorly defined, limiting accurate delineation between physiological and pathological uptake. The study aimed to quantify regional FDG uptake in the cervical, thoracic, and lumbar spinal cord in healthy adults, and evaluate associations between FDG uptake and demographic factors, including age, sex, body mass index (BMI), and inflammatory/metabolic biomarkers. A secondary aim was to apply artificial intelligence (AI)-driven segmentation to facilitate standardized region-of-interest extraction across spinal cord levels.

This was a retrospective, cross-sectional observational study conducted using data from the CAMONA clinical trial at Odense University Hospital, Denmark. 76 healthy adult participants (mean age 44.66 ± 14.04 years; 53.9% male) were selected from a total of 139 subjects in the CAMONA trial. All participants were free from chronic medical conditions. Primary outcomes include mean standardized uptake values (SUVmean) of FDG in the cervical, thoracic, and lumbar spinal cord regions. Secondary outcomes include correlations of SUVmean with age, sex, BMI, HbA1c, and inflammatory biomarkers (CRP, WBC, homocysteine, fibrinogen). Participants underwent FDG PET/CT imaging 180 min post-injection of 4.0 MBq/kg FDG. Using TotalSegmentator, an AI-driven segmentation tool, the spinal cord was delineated into cervical, thoracic, and lumbar volumes of interest (VOIs). Regional SUVmeans were calculated and analyzed. Statistical tests included one-way ANOVA with Tukey's post-hoc, Spearman/Pearson correlations and t-tests to assess associations with age, sex, BMI, and inflammatory/metabolic biomarkers.

FDG uptake was highest in the cervical spinal cord (SUVbw 1.36 ± 0.23), followed by the lumbar (0.85 ± 0.18) and thoracic (0.76 ± 0.17) regions (p < 0.001); this regional ordering was preserved after lean-body-mass-normalized SUV (SUL) recomputation. No significant associations were found with age or sex. BMI correlated positively with SUVbw across all regions (ρ = 0.34-0.47, all p ≤ 0.003); under SUL, the cervical and thoracic associations attenuated to non-significance (cervical ρ = +0.13, p = 0.255; thoracic ρ = +0.17, p = 0.151), whereas a residual lumbar association persisted (ρ = +0.32, 95% CI + 0.09 to + 0.52, p = 0.004). A nominal association between HbA1c and cervical uptake (ρ = 0.24, p = 0.040) did not survive Bonferroni correction for multiple comparisons (α = 0.0033) and is reported as exploratory. Inflammatory markers showed no significant correlation in any spinal cord region.

FDG PET/CT reveals consistent regional variation in spinal cord metabolic activity, with highest uptake in the cervical region. Normative uptake was stable across age and sex but showed a positive association with BMI when SUV was normalized to total body weight that was largely abolished after lean-body-mass normalization (SUL) in the cervical and thoracic regions but persisted in the lumbar region (lumbar SUL ρ = +0.32, 95% CI + 0.09 to + 0.52, p = 0.004). These findings offer foundational data for clinical interpretation and differentiation between physiology and pathology of spinal FDG PET/CT uptake and demonstrate the feasibility of AI-driven segmentation for standardized region-of-interest extraction in spinal imaging analysis.