Abstract
56
Objectives: Studies highlighting the interest of FDG-PET/CT (FDG) in the pre-operative assessment of pheochromocytomas (PHEO) are very sparse. This retrospective study assessed the potential added value of radiomics based on FDG for the characterization of PHEO before surgery and before knowledge of genetic testing results.
Methods: Forty-nine patients (and 52 PHEO) explored with both FDG and metaiodobenzylguanidine (MIBG) scan before surgery, between March 2007 to November 2015, were retrospectively included. A germline mutation was secondarily identified in 13 patients in one of the genes related to cluster 1 (4 patients) or cluster 2 (9 patients). No mutation was identified in 33 patients and 3 did not have genetic testing. Several PET-based biomakers including SUVmax, metabolic tumor volume (MTV), total lesion glycolysis (TLG) and textural features (TF) were analyzed for a potential correlation with biochemical and genetic phenotype. For TF analysis, adrenal lesions were automatically delineated on PET images using a consensus approach (STAPLE) calculated with 4 different segmentations. TF chosen among the most robust and computed using the most intense tumor (minimal size: 64 voxels). Two different methods of quantization (linear using 64 bins and absolute using fixed bins of 0.3 SUV) were assessed for TF computation. All computations followed the framework of the Image Biomarkers Standardization Initiative. The correlation in-between quantitative PET-derived biomarkers was performed using the Spearman rank test adjusted for multiple comparison using the Benjamini-Hochberg approach. Receiver operating characteristic (ROC) was performed for each PET parameter significantly associated to a categorized variable and a model including a subset of selected TF was built to potentially improve the discrimination between patients with or without mutation when compared with the use of a single conventional PET parameter (MTV, TLG or SUVmax).
Results: Sensitivity of FDG-PET/CT alone for the detection of PHEO was 92%, and 98% when combined to MIBG. The SUVmax was significantly higher for tumors classified in Cluster 1 than in Cluster 2 (p = 0.002) or than tumors without identified mutations (p = 0.04). MTV and TLG of the tumors with the most intense uptake discriminated Cluster 2 from sporadic, but not from Cluster 1. TF alone did not outperform conventional PET-based biomarkers but when they were combined with MTV, it led a better differentiation between sporadic and mutated tumors (p<0.05).
Conclusions: FDG-PET/CT was found useful for preoperative detection of PHEO and, with a combination of radiomics biomarkers, for the determination of their genetic component.