Dirk Ertel, Presenter: Nothing to Disclose

Ronny Hendrych, Abstract Co-Author: Nothing to Disclose

Tobias Pflederer MD, Abstract Co-Author: Nothing to Disclose

Sylvia Wilharm, Abstract Co-Author: Nothing to Disclose

Stephan Achenbach MD, Abstract Co-Author: Research support, Siemens AG Consultant, Bracco Group Research support, Bayer AG

Willi A. Kalender PhD, Abstract Co-Author: Consultant, Siemens AG Founder, VAMP GmbH Scientific Advisor, VAMP GmbH Shareholder, VAMP GmbH

To provide a tool improving diagnostic performance and clinical workflow in cardiac CT with respect to the trade-off between image noise, temporal resolution and reconstruction phase selection.

We developed a procedure which allows for an interactive adaptation of image noise, temporal resolution and reconstruction phase used in phase-correlated CT imaging. Initially, multiple phase-correlated image reconstructions are performed covering the complete cardiac cycle. A dedicated 4D filtering procedure is applied to the datasets. The procedure is embedded in a CT viewing tool offering MPR views where the reconstruction phase and the temporal resolution can be varied continuously. For validation we used a 3D motion robot (QRM, Moehrendorf, Germany) providing heart specific trajectories. Assessment of the temporal modulation transfer function (t-MTF) and noise measurements were used for performance validation. Measurements were performed on a dual-source CT (SOMATOM Definition, Siemens, Forchheim) with 0.33 s rotation time. 10 patient datasets which were found difficult diagnose were used to evaluate our approach.

The 4D filtering procedure enables interactive image post-processing, i.e. the retrospective trade-off between temporal resolution, image noise and reconstruction phase in real time. Image noise can be interactively adjusted whereas the final signal-to-noise ratio is highly dependent on kernel size and kernel shape used. The temporal resolution is inversely related to the kernel size indicated by a decreasing t-MTF for increasing filter size. Filter kernels with distance weighting (e.g. Gauss, Hann) provide better results compared to constant filter kernels (e.g. rectangular) with respect to continuous adjustment. Automatic pre-selection of a dedicated VOI reduced the computational effort and seemed to be essential. Case studies revealed improved image quality with respect to coronary stenosis detection.

Appropriate adjustment of temporal resolution, image noise and reconstruction phase for cardiac imaging can be made interactively available.

The proposed tool can improve the medical workflow by a fast and intuitive 4D image post-processing.

Ertel, D, Hendrych, R, Pflederer, T, Wilharm, S, Achenbach, S, Kalender, W, An Interactive Temporal Filtering Procedure for Cardiac CT with Real-time Performance.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6007812.html