Which combination of MR imaging modalities is best for predicting recurrent glioblastoma? Study of diagnostic accuracy and reproducibility.

PURPOSE To compare the added value of dynamic contrast material-enhanced ( CE contrast enhanced ) ( DCE dynamic CE ) magnetic resonance (MR) imaging with that of dynamic susceptibility CE contrast enhanced ( DSC dynamic susceptibility CE ) MR imaging with the combination of CE contrast enhanced T1-weighted imaging and diffusion-weighted ( DW diffusion weighted ) imaging for predicting recurrent glioblastoma. MATERIALS AND METHODS This retrospective study was approved by the institutional review board, with the requirement for informed patient consent waived. CE contrast enhanced T1-weighted images, DW diffusion weighted images, DSC dynamic susceptibility CE MR images, and DCE dynamic CE MR images in 169 patients with pathologically or clinicoradiologically diagnosed recurrent glioblastoma (n = 87) or radiation necrosis (n = 82) were retrospectively reviewed. Histogram cutoffs of quantitative parametric values were calculated from DW diffusion weighted images, DSC dynamic susceptibility CE MR images, and DCE dynamic CE MR images. Area under the receiver operating characteristic curve ( Az area under the ROC curve ) and interreader agreement were assessed. RESULTS For predicting recurrent glioblastoma, adding DCE dynamic CE MR imaging to the combination of CE contrast enhanced T1-weighted imaging and DW diffusion weighted imaging significantly improved Az area under the ROC curve from 0.84 to 0.96 for reader 1 and from 0.81 to 0.97 for reader 2, respectively. Adding DSC dynamic susceptibility CE MR imaging also significantly improved Az area under the ROC curve (0.95 for reader 1 and 0.93 for reader 2). However, there was no significant difference in Az between the combination of CE contrast enhanced T1-weighted imaging, DW diffusion weighted imaging, and DSC dynamic susceptibility CE MR imaging and the combination of CE contrast enhanced T1-weighted imaging, DW diffusion weighted imaging, and DCE dynamic CE MR imaging for both readers. The interreader agreement was highest for the combination of CE contrast enhanced T1-weighted imaging, DW diffusion weighted imaging, and DCE dynamic CE MR imaging (κ = 0.78) and lowest for CE contrast enhanced T1-weighted imaging and DW diffusion weighted imaging (κ = 0.65). CONCLUSION Adding perfusion MR imaging to the combination of CE contrast enhanced T1-weighted imaging and DW diffusion weighted imaging significantly improves the prediction of recurrent glioblastoma; however, selection of perfusion MR method does not affect the diagnostic performance.