Catheter Artifact Reduction (CAR) in Dynamic Cardiac Chamber Imaging with Interventional C-arm CT

A C-arm CT system offers the possibility to acquire 2-D high-resolution X-ray images of the patient from different views. These projections can be used for 3-D imaging. Anatomical three-dimensional imaging holds great potential to improve cardiac interventions. However, the image acquisition using a C-arm CT system takes several seconds. A standard FDK reconstruction using all acquired projection images results in a motion blurred image. In order to improve the temporal resolution, an electrocardiogram (ECG) is acquired synchronously with the acquisition and reconstruction is performed only with subsets of projection images, each belonging to a certain heart phase. This retrospective ECG-gating of data from a single C-arm rotation provides only a few projections per heart phase for image reconstruction. This view sparsity leads to prominent streak artifacts and a poor signal-to-noise ratio. Therefore, motion estimation and correction are required for the reconstruction of the cardiac chambers. We recently presented a deformable image registration approach which allows for motion estimation on the initial ECG-gated volumes using a specifically designed imaging protocol. In this paper, an additional step to improve the initial image quality is presented which removes dense objects, i.e. pacing electrodes or catheters, before the deformable 3-D/3D registration step. The algorithm is evaluated quantitatively and qualitatively on a simulated phantom dataset. The relative root mean square error (rRMSE) was reduced by 27 % and the universal image quality index (UQI) improved by 13 % compared to the algorithm without removing the dense objects from the reconstructions. Finally, the presented algorithmic framework was applied to a first clinical patient dataset and the preliminary results are presented in this paper.