Image Artifact Influence on Motion Compensated Tomographic Reconstruction in Cardiac C-arm CT

In C-arm CT, electrocardiogram (ECG)-gating of data from a single C-arm rotation provides only a few projections per heart phase for image reconstruction. This view starvation leads to prominent streak artifacts and a poor signal to noise ratio. Motion compensation techniques allow for the use of all acquired data for image reconstruction. Cardiac motion can be estimated by deformable 3-D/3-D registration processed on initial 3-D images of different heart phases. The initial 3-D images are computed from the few, ECG-gated data. In this paper, the sensitivity of the 3-D/3-D registration step to the image quality of the initial images is studied. Different reconstruction algorithms are evaluated for a recently proposed cardiac C-arm CT acquisition protocol. An iterative few-view reconstruction, and a filtered backprojection method (FDK) with and without a bilateral filter are investigated with respect to the final motion compensated reconstruction quality. The algorithms were tested on a phantom and on a porcine model using qualitative and quantitative measures. The phantom projection data and geometry is publicly available and can be downloaded from conrad.stanford.edu/data/heart. The results show minor differences between the three motion compensated reconstructions. For two heart phases a relative root mean square error (rRMSE) of ≈ 0.09 and 0.06 and an universal image quality index (UQI) of ≈ 0.98 and 0.99 was achieved. The motion compensated reconstructions that use all of the projection images show a clear improvement compared to the initial reconstructions. Given the relatively small differences in final image quality, the algorithm of choice is likely to be the one with smallest computational complexity.