Efficient reconstruction from truncated data in circular cone-beam CT

Abstract

In circular cone-beam computed tomography (CT), to solve the 3D image reconstruction from truncated projection data which has no truncation along PI-line, backprojection-filtration (BPF) algorithm is a preferred choice. However, in its performance the integral interval of backprojection is variable for different PI-line, rendering the parallelism performance of backprojection low. So it cannot satisfy the requirement of fast image reconstruction in practical CT system. In this paper, a tent BPF (T-BPF) algorithm is developed based on the data rebinning method, which was performed by first rearranging the cone-beam data to tent-like parallel-beam data, and then applying the proposed BPF-type algorithm to reconstruct images from the rearranged data. T-BPF turns the variable view-angle integral interval of backprojection into a fixed integral interval, and there are no relations in the loops of backprojection calculation, which means the parallelism performance of T-BPF is an improvement over that of the original BPF algorithm. The results of experiments show that compared with the conventional CPU implementation, the GPU accelerated method provides images of the same quality with a speedup factor 1036 for the reconstruction of 2563 Shepp-Logan model. The speedup factor is an improvement in the original BPF algorithm. T-BPF provides a solution for the 3D fast reconstruction from truncated data.