Reproducibility of volume analysis of dynamic susceptibility contrast perfusion–weighted imaging in untreated glioblastomas

Purpose Despite a high variability, the hotspot method is widely used to calculate the cerebral blood volume (CBV) of glioblastomas on DSC-MRI. Our aim was to investigate inter- and intra-observer reproducibility of parameters calculated with the hotspot or a volume method and that of an original parameter assessing the fraction of pixels in the tumour volume displaying rCBV > 2: %rCBV > 2. Methods Twenty-seven consecutive patients with untreated glioblastoma (age: 63, women: 11) were retrospectively included. Three observers calculated the maximum tumour CBV value (rCBVmax) normalized with a reference ROI in the contralateral white matter (CBVWM) with (i) the hotspot method and (ii) with a volume method following tumour segmentation on 3D contrast–enhanced T1-WI. From this volume method, %rCBV > 2 was also assessed. After 8–12 weeks, one observer repeated all delineations. Intraclass (ICC) and Lin’s (LCC) correlation coefficients were used to determine reproducibility. Results Inter-observer reproducibility of rCBVmax was fair with the hotspot and good with the volume method (ICC = 0.46 vs 0.65, p > 0.05). For CBVWM, it was fair with the hotspot and excellent with the volume method (0.53 vs 0.84, p 2, that could be promising during the follow-up of such heterogeneous tumours.