Lukas Havla, Presenter: Nothing to Disclose

Kolja Thierfelder MD, MSc, Abstract Co-Author: Nothing to Disclose

Sebastian Ekkehard Beyer, Abstract Co-Author: Nothing to Disclose

Maximilian F. Reiser MD, Abstract Co-Author: Nothing to Disclose

Wieland H. Sommer MD, Abstract Co-Author: Nothing to Disclose

Olaf Dietrich PhD, Abstract Co-Author: Nothing to Disclose

To evaluate a new approach for the reconstruction of high-quality 3D angiographic datasets based on the pixel-by-pixel application of wavelet transforms on CT perfusion data in the time domain.

Cerebral CT perfusion data of 14 consecutive patients with suspected stroke but no ischemia on follow-up MRI and without any other pathology that could alter the cerebral hemodynamics were included in this study. All patients were examined at multi-detector CT systems acquiring 32 dynamic phases (temporal resolution: 1.5s) of 99 slices (total slab thickness 99mm) at 80kV/350mAs. Typically, 35 mL of iomeprol-350 were injected at flow rate of 4.5 mL/s. Angiographic datasets were calculated after initial rigid-body motion correction using (a) temporal maximum intensity projections (tMIP) (E.J. Smit et al. Radiology 2012) and (b) the proposed wavelet method performed with the commonly used Paul-wavelet (order 4). In the latter approach, we calculated the wavelet power spectrum of the time-attenuation curves for each pixel and displayed the maximum of this spectrum as angiographic signal intensity. Both approaches were compared with respect to the contrast-to-noise ratio (CNR) relative to surrounding tissue of 16 different vessel segments, and qualitatively using a 5pt Likert scale (0 non diagnostic, 4 excellent) with respect to image quality by two blinded and experienced readers.

The CNR for the wavelet reconstruction (580.2±474.5) was significantly higher than for the tMIP approach (60.7±31.0, Wilcoxon test p < 0.00001). Qualitatively, our new method performed significantly better than the tMIP approach with mean scores of 3.7/3.7 (reader 1/reader 2), inter-observer Cohen’s κ=1 vs. tMIP scores of 2.8/2.9, κ=0.594 (p<0.001/p=0.001).

A ten-fold increase of contrast-to-noise ratio can be achieved for intracranial vessels by using wavelet transforms of intracranial CT perfusion datasets compared to currently used tMIP methods. The higher CNR and the resulting increase in image quality (plus method-inherent bone subtraction) are especially important for the assessment of small peripheral branches as well as for leptomeningeal collateral vessels.

Using wavelet transforms, angiographic data with excellent image quality can be obtained from dynamic CT perfusion data, potentially allowing to omit a separate CT angiography examination.

Havla, L, Thierfelder, K, Beyer, S, Reiser, M, Sommer, W, Dietrich, O, Reconstruction of Cerebral Angiographic Data from Time-resolved CT Perfusion Acquisitions Using Wavelet Transforms.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14012703.html