Fig 1.
Twenty-five patients with pathologically proven primary glioblastoma, with at least 4 MRIs and usable perfusion data acquired on 3T systems, treated by chemoradiotherapy showing progressive lesion enhancement on early follow-up MRI (CRT+1) were included.
Fig 2.
Illustration of the two different methods.
Two methods were used to calculate the maximum relative cerebral blood volume (rCBVmax) and %rCBV>2, in a 60-year-old patient with glioblastoma. a) and c) axial contrast-enhanced T1-weighted images; b) and d) corrected rCBV maps. Top line, the hot-spot method: Two ROIs (red) are placed in hypervascularized areas of the enhancing tumour and one (yellow) is placed in the contralateral white matter for normalization. Bottom line, the volume method: Segmentation of the enhancing tumour (red) slice by slice including areas of high and low neoangiogenesis and segmentation of a volume of interest in the contralateral white matter (yellow) for normalization. e) The Bland-Altmann plot illustrates the concordance between the two methods used to analyse rCBVmax.
Table 1.
Patient characteristics.
Fig 3.
%rCBV>2 is the only parameter that distinguish tumour progression from pseudoprogression on the initial MRI.
Box-and-whiskers graphs of the maximum relative cerebral blood volume (rCBVmax) on the initial MRI in the tumour progression (TP) or pseudoprogression (PsP) groups obtained with a) the hot-spot method (ROI), b) the volume method (VOL), and c) %rCBV>2. %rCBV>2 was the only significantly lower parameter in the PsP group.
Fig 4.
%rCBV>2 differentiate tumour progression from pseudoprogression on the early follow-up MRI with a higher confidence level than rCBVmax.
Box-and-whiskers graphs of the maximum relative cerebral blood volume (rCBVmax) on early follow-up MRI (CRT+1) in the tumour progression (TP) or pseudoprogression (PsP) groups obtained with a) the hot-spot method (ROI), b) the volume method (VOL) and c) %rCBV>2. rCBVmax and %rCBV>2 were significantly lower in the PsP group. d) ROC curves for rCBVmax (ROI), rCBVmax (VOL) and %rCBV>2. Best Youden index for pseudoprogression diagnosis was obtained when %rCBV>2 was less than 41.5%.
Fig 5.
Longitudinal follow-up of two different glioblastomas with similar high rCBVmax (ROI) and different %rCBV>2.
Axial contrast-enhanced T1-weighted images (CE-T1WI) of post-operative MRI (a and e), showing progressive lesions on CE-T1WI of CRT+1 MRI (b and f) with their corrected rCBV maps (c and g) and subsequent evolution on CRT+3 MRI (d and h). Top line (a-d) shows tumour progression (arrows) at CRT+3, calculated rCBVmax was 10.1 and %rCBV>2 was 60%. Bottom line (d-f) demonstrates regression of the lesion (thin arrows) indicative of pseudoprogression, despite high rCBVmax (9.1), %rCBV>2 was 32% and could have predicted pseudoprogression.
Fig 6.
Pseudoprogression diagnosis accuracy of %rCBV>2 after exclusion of the central necrosis.
Axial contrast-enhanced T1-weighted images (CE-T1WI) of early follow-up MRI of two morphologically different false positive (FP) pseudoprogression (PsP): (a) partially necrotic and (b) mostly necrotic. Table (c) presents %rCBV>2 mean measurements of %rCBV>2 in the PsP and tumour progression groups and individual measures for FP PsP patients, before and after exclusion of the central necrotic component from perfusion analysis and ROC curves comparison (d) for the diagnosis of PsP shows no statistical difference.