Threshold-based meningioma delineation in SSR-targeted PET

Ziel/Aim Meningiomas are the most frequent primary brain tumors of non-glial origin. Somatostatin receptor (SSR)-targeted PET has gained importance for treatment planning in meningiomas. However, there is no standard method for tumor delineation in PET. Therefore, we investigated different approaches for meningioma segmentation in SSR-targeted PET.

Methodik/Methods 29 treatment-naïve meningiomas visually clearly delineable on MRI and PET were included. MRI tumor volume was set as reference. Tumor delineation in PET consisted in threshold-based segmentations using 30 pre-fixed SUV-cutoffs (SUV1.50-2.95 in steps of 0.05) and 15 pre-fixed isocontours (20 %-90 % of SUV_max, in steps of 5 %). Dice coefficient was used to evaluate the similarity of PET tumor volumes to the reference in order to assess the optimal threshold for every lesion. Multiple regression was used to test the influence of lesion size (i. e. MRI volume) and of SUV_max on the optimal thresholds.

Ergebnisse/Results None of the 45 pre-fixed thresholds was suitable to consistently generate the reference volume. For each single SUV-cutoff/isocontour there was a considerably high range of dice coefficients, even for the previously reported 2.3 SUV-cutoff (range 0 %-90 %). Instead, individual thresholds for each lesion were necessary for optimal tumor delineation in PET. Multiple regression analysis showed that the optimal SUV-based threshold depends on SUV_max (p = 0.003) but not on the lesion size (p = 0.837), whereas vice versa the optimal isocontour depends on the lesion size (p = 0.021) but not on SUV_max (p = 0.688).

Schlussfolgerungen/Conclusions In meningioma patients, individual thresholds per lesion seem to be superior to a single pre-fixed cutoff for tumor delineation in SSR-targeted PET, e. g. superior to the previously reported 2.3 SUV-cutoff. These preliminary results need to be evaluated in larger cohorts including pre-treated meningiomas at anatomically challenging locations.

Publication History

Article published online:
08 April 2021