Wavelet-based calculation of cerebral angiographic data from time-resolved CT perfusion acquisitions

To evaluate a new approach for reconstructing angiographic images by application of wavelet transforms on CT perfusion data.

Fifteen consecutive patients with suspected stroke were examined with a multi-detector CT acquiring 32 dynamic phases (∆t = 1.5s) of 99 slices (total slab thickness 99mm) at 80kV/200mAs. Thirty-five mL of iomeprol-350 was injected (flow rate = 4.5mL/s). Angiographic datasets were calculated after initial rigid-body motion correction using (a) temporally filtered maximum intensity projections (tMIP) and (b) the wavelet transform (Paul wavelet, order 1) of each voxel time course. The maximum of the wavelet-power-spectrum was defined as the angiographic signal intensity. The contrast-to-noise ratio (CNR) of 18 different vessel segments was quantified and two blinded readers rated the images qualitatively using 5pt Likert scales.

The CNR for the wavelet angiography (501.8 ± 433.0) was significantly higher than for the tMIP approach (55.7 ± 29.7, Wilcoxon test p < 0.00001). Image quality was rated to be significantly higher (p < 0.001) for the wavelet angiography with median scores of 4/4 (reader 1/reader 2) than the tMIP (scores of 3/3).

The proposed calculation approach for angiography data using temporal wavelet transforms of intracranial CT perfusion datasets provides higher vascular contrast and intrinsic removal of non-enhancing structures such as bone.

• Angiographic images calculated with the proposed wavelet-based approach show significantly improved contrast-to-noise ratio. • CT perfusion-based wavelet angiography is an alternative method for vessel visualization. • Provides intrinsic removal of non-enhancing structures such as bone.