Variability and accuracy of different software packages for dynamic susceptibility contrast magnetic resonance imaging for distinguishing glioblastoma progression from pseudoprogression

Abstract

Determining whether glioblastoma multiforme (GBM) is progressing despite treatment is challenging due to the pseudoprogression phenomenon seen on conventional MRIs, but relative cerebral blood volume (CBV) has been shown to be helpful. As CBV's calculation from perfusion-weighted images is not standardized, we investigated whether there were differences between three FDA-cleared software packages in their CBV output values and subsequent performance regarding predicting survival/progression. Forty-five postradiation therapy GBM cases were retrospectively identified as having indeterminate MRI findings of progression versus pseudoprogression. The dynamic susceptibility contrast MR images were processed with different software and three different relative CBV metrics based on the abnormally enhancing regions were computed. The intersoftware intraclass correlation coefficients were 0.8 and below, depending on the metric used. No statistically significant difference in progression determination performance was found between the software packages, but performance was better for the cohort imaged at 3.0 T versus those imaged at 1.5 T for many relative CBV metric and classification criteria combinations. The results revealed clinically significant variation in relative CBV measures based on the software used, but minimal interoperator variation. We recommend against using specific relative CBV measurement thresholds for GBM progression determination unless the same software or processing algorithm is used.

Keywords: cerebral blood volume; dynamic-susceptibility contrast; glioblastoma; magnetic resonance imaging; perfusion.

Fig. 1

Example of the change in relaxivity versus time curve for an individual tumor voxel. The change in relaxivity reflects the concentration of gadolinium-based contrast present within the voxel. The shaded region represents the basis of cerebral blood volume (CBV) calculation.

Fig. 2

Tumor segmentation method: (a) example enhancing region with surrounding lasso drawn manually, (b) histogram of voxel intensities, with the red line specifying the calculated Otsu threshold, and (c) final segmentation result, with the enhancing tissue shaded in red.

Fig. 3

Relative CBV (rCBV) values for each sampled voxel for a selected case. Each marker represents the rCBV value for two separate software packages for the same voxel. (a) CBV values of contrast-enhancing pixels for FuncTool IB Neuro. (b) CBV values of normal-appearing white matter pixels for FuncTool versus IB Neuro. (c) CBV values of contrast-enhancing pixels for nordicICE versus IB Neuro. (d) CBV values of normal-appearing white matter pixels for nordicICE versus IB Neuro. (e) CBV values of contrast-enhancing pixels for nordicICE versus Functool. (f) CBV values of normal-appearing white matter pixels for nordicICE versus Functool.

Fig. 4

Percent of subjects above rCBV metric threshold. The lines represent the software-specific averages across the three operators. The shaded area on either side represents the interoperator range for that software. (a) Shows the percentage of cases that are above threshold using the mean rCBV value as the metric. (b) Shows the percentage of cases that are above threshold using the intensity of the 95th percentile as the metric. (c) Shows the percentage of cases that are above threshold using the percentage of tumor voxels with rCBV above white matter as the metric.

Fig. 5

Percentage of cases where one software package disagreed with the other two (by operator). The $x$ axis range plotted is for all software’s percent of cases above the threshold (as shown in Fig. 4) being between 25% and 75%. The threshold is used to differentiate between pseudoprogression and progression. (a) The percentage of cases with disagreement in assessment of progression across a range of mean rCBV thresholds for different operators. (b) The percentage of cases with disagreement in assessment of progression across a range of 95th percentile thresholds for different operators. (c) The percentage of cases with disagreement in assessment of progression across a range of percentage of voxels above NAWM thresholds for different operators.

Fig. 6

Percentage of cases where one software package disagreed from other two (by software). Operator 1’s data. The $x$ axis range plotted is for all software’s percent of cases above the threshold (as shown in Fig. 4) being between 25% and 75%. The threshold is used to differentiate between pseudoprogression and progression. (a) The percentage of cases where one software package disagreed from the other two across a range of mean rCBV thresholds. (b) The percentage of cases where one software package disagreed from the other two across a range of 95th percentile rCBV thresholds. (c) The percentage of cases where one software package disagreed from the other two across a range of thresholds for percent of voxels above NAWM.