Comparison of DSC-MRI post-processing techniques in predicting microvascular histopathology in patients newly diagnosed with GBM.

PURPOSE

To evaluate which common post-processing method applied to gradient-echo DSC-MRI data, acquired with a single gadolinium injection and low flip-angle, most accurately reflects microvascular histopathology for patients with de novo, treatment-naive glioblastoma multiforme (GBM).

MATERIALS AND METHODS

Seventy-two tissue samples were collected from 35 patients with treatment-naive GBM. Sample locations were co-registered to preoperative gradient-echo dynamic susceptibility contrast (DSC) MRI acquired with 35° flip-angle and 0.1 mmol/kg gadolinium. Estimates of blood volume and leakiness at each sample location were calculated using four common postprocessing methods (leakage-corrected nonlinear gamma-variate, non-parametric, scaled MR-signal, and unscaled MR-signal). Tissue sample microvascular morphology was characterized using Factor VIII immunohistochemical analysis. A random-effects regression model, adjusted for repeated measures and contrast-enhancement (CE), identified whether MR parameter estimates significantly predicted IHC findings.

RESULTS

Elevated blood volume estimates from nonlinear and non-parametric methods significantly predicted increased microvascular hyperplasia. Abnormal microvasculature existed beyond the CE-lesion and was significantly reflected by increased blood volume from nonlinear, non-parametric, and scaled MR-signal analysis.

CONCLUSION

This study provides histopathological support for both non-parametric and nonlinear post-processing of low flip-angle DSC-MRI for characterizing microvascular hyperplasia within GBM. Non-parametric analysis with a single gadolinium injection may be a particularly useful strategy clinically, as it requires less computational expense and limits gadolinium exposure.