Abstract Background Amino acid PET with 18F-fluoroethyltyrosine (18F-FET) complements MRI to improve specificity of tumor detection. We evaluated whether threshold-based volumetric metrics (TBRthres) provide complementary diagnostic and prognostic information to single-voxel tumor-to-brain ratio maximum (TBRmax). Methods This retrospective single-institution study included 100 previously treated glioma patients (WHO Grade 2-4) who underwent 18F-FET PET/MRI (N = 96) or PET/CT (N = 4) following equivocal findings of tumor progression versus treatment-related changes. Patients received 693.8±79.6 MBq of 18F-FET, with PET analysis based on the 20-40 minute averaged static sequence acquired post-injection. Diagnostic performance, reproducibility, and survival associations were assessed using ROC analysis, intraclass correlation coefficients (ICC), and Cox proportional hazards models. Results TBRthres map generation was highly reproducible (ICC0.86) and significantly associated with overall survival in univariate analyses (HR range: 3.07-4.45). Volumetric metrics demonstrated diagnostic performance comparable to TBRmax (AUC up to 0.91). TBRmax showed moderate correlation with tumor volume (R2 ≈ 0.35-0.40), indicating partially overlapping but non-redundant information. In multivariate models adjusting for clinical covariates, neither volumetric metrics nor TBRmax remained statistically significant. However, in models adjusting for TBRmax, volumetric TBRthres metrics remained significantly associated with survival. Combined models improved predictive performance at selected thresholds, particularly TBRthres ≥ 1.6 (AUC 0.62 to 0.70, p = 0.009) and TBRthres ≥ 2.0 (AUC 0.62 to 0.72, p = 0.019), with smaller improvement at TBRthres ≥ 2.5. Conclusion Volumetric 18F-FET PET metrics provide reproducible measures of tumor burden with diagnostic and prognostic performance comparable to TBRmax. They offer complementary information and, at selected thresholds, modest incremental predictive value beyond TBRmax.
Oo et al. (Fri,) studied this question.