Parameter maps based on wavelet-transform post-processing of dynamic perfusion data offer an innovative way of visualizing blood vessels in a fully automated, user-independent way. The aims of this study were (i) a proof of concept regarding wavelet-based analysis of dynamic susceptibility contrast (DSC) MRI data and (ii) to demonstrate advantages of wavelet-based measures compared to standard cerebral blood volume (CBV) maps in patients with the initial diagnosis of glioblastoma (GBM). Consecutive 3-T DSC MRI datasets of 46 subjects with GBM (mean age 63.0 ± 13.1years, 28 m) were retrospectively included in this feasibility study. Vessel-specific wavelet magnetic resonance perfusion (wavelet-MRP) maps were calculated using the wavelet transform (Paul wavelet, order 1) of each voxel time course. Five different aspects of image quality and tumor delineation were each qualitatively rated on a 5-point Likert scale. Quantitative analysis included image contrast and contrast-to-noise ratio. Vessel-specific wavelet-MRP maps could be calculated within a mean time of 2:27min. Wavelet-MRP achieved higher scores compared to CBV in all qualitative ratings: tumor depiction (4.02 vs. 2.33), contrast enhancement (3.93 vs. 2.23), central necrosis (3.86 vs. 2.40), morphologic correlation (3.87 vs. 2.24), and overall impression (4.00 vs. 2.41); all p < .001. Quantitative image analysis showed a better image contrast and higher contrast-to-noise ratios for wavelet-MRP compared to conventional perfusion maps (all p < .001). wavelet-MRP is a fast and fully automated post-processing technique that yields reproducible perfusion maps with a clearer vascular depiction of GBM compared to standard CBV maps. • Wavelet-MRP offers high-contrast perfusion maps with a clear delineation of focal perfusion alterations. • Both image contrast and visual image quality were beneficial for wavelet-MRP compared to standard perfusion maps like CBV. • Wavelet-MRP can be automatically calculated from existing dynamic susceptibility contrast (DSC) perfusion data.
Read full abstract