Perfusion analysis concluded differently on angiogenesis' presence on both software.
Twelve exams out of 54 do not have the same diagnosis according to the software used.
Results obtained by two different operators on one single software are reproducible.
In clinical routine, follow-up should preferably be performed with the same software.
Background and purpose
The main purpose of this retrospective study was to evaluate inter-software variability in patients affected of a high-grade glioma for the post-processing of dynamic susceptibility contrast (DSC1) perfusion imaging in MRI.2
Materials and methods
The included patients were either anaplastic astrocytoma (WHO3 grade III) or glioblastoma (WHO grade IV) located in the cerebral parenchyma. The postprocessing of 54 MRI-DSC imaging from 46 patients using both Intellispace© (Philips) and Olea© (Olea Medical) software was performed. The hemodynamic parameter studied was the normalised relative cerebral blood volume corrected for the T1 leakage effect (nrCBVc4). The inter-operator variabilities were also evaluated.
Regarding inter-software reproducibility, Cohen’s Kappa from therapeutic follow-up obtained were 0.61, close to the recommended limit (0.60). Subgroups were created to complete the analysis and to evaluate the partial volume effect. Even if necrosis or vascular structures from regions of interest (ROI5) were avoided, results did not improve. ROI of a minimum area of 250 mm2 yielded a Cohen’s Kappa of 0.65. The inter-operator reproducibility on Intellispace and Olea were 0.90 and 0.73 respectively, which is satisfactory.
The reproducibility between Intellispace and Olea was below recommended threshold in a clinical context. This discrepancy can be explained by the partial volume effect and the models used. ROI with an area of at least 250 mm2 improves this reproducibility and becomes acceptable.Le texte complet de cet article est disponible en PDF.
Keywords : MRI, Dynamic susceptibility contrast perfusion, High-grade glioma, Post-processing software, Reproducibility