Differences in dynamic susceptibility contrast MR perfusion maps generated by different methods implemented in commercial software

There are several potential sources of difference that can influence the reproducibility of magnetic resonance (MR) perfusion values. We aimed to investigate the reproducibility and variability of dynamic susceptibility contrast (DSC) MR imaging (MRI) parameters obtained from identical source data by using 2 commercially available software applications with different postprocessing algorithms.

We retrospectively evaluated DSC-MRI data sets of 24 consecutive patients with glioblastoma multiforme. Perfusion data were postprocessed with 2 commercial software packages, NordicICE (NordicNeuroLab, Bergen, Norway) and GE Brainstat (GE Healthcare, Milwaukee, Wis), each of which offers the possibility of different algorithms. We focused the comparison on their main analysis issues, that is, the gamma-variate fitting function (GVF) and the arterial input function (AIF). Two regions of interest were placed on maps of perfusion parameters (cerebral blood volume [CBV], cerebral blood flow [CBF], mean transit time [MTT]): one around tumor hot spot and one in the contralateral normal brain. A one-way repeated-measures analysis of variance was conducted to determine whether there was a significant difference in the calculated MTT, CBV, and CBF values.

As regards NordicICE software application, the use of AIF is significant (P = 0.048) but not the use of GVF (P = 0.803) for CBV values. Additionally, in GE, the calculation method discloses a statistical effect on data. Comparing similar GE-NordicICE algorithms, both method (P = 0.005) and software (P < 0.0001) have a statistical effect in the difference. Leakage-corrected and uncorrected normalized CBV (nCBV) values are statistically equal. No statistical differences have been found in nMTT values when directly calculated. Values of nCBF are affected by the use of GVF.

The use of a different software application determines different results, even if the algorithms seem to be the same. The introduction of AIF in the data postprocessing determines a higher estimates variability that can make interhospital and intrahospital examinations not completely comparable. A simpler approach based on raw curve analysis produces more stable results.