Cone Beam Reconstruction Speedup Using Trigonometric Relevancy and GPU Technology

Cone beam computed tomography (CBCT) represents a state-of-the-art CT technique, which is real three-dimensional (3D) imaging modality due to its volume scanning and fully 3D reconstruction. However, the 3D volume reconstruction demands enormous computations, which becomes the bottleneck for CBCT applications. This paper proposes a technique to speed up back projection reconstruction through the use of trigonometric relevant formula associated with the backprojection geometry and graphic process unit (GPU) for implementation. The employment of trigonometric formula avoids repeated trigonometric computations for projections that are 90 degrees apart. The reconstruction process can be further sped up by GPU-based backprojecions. Simulation results show that the new algorithm, through the use of trigonometric formula and GPU technology, can speed up the reconstruction by 3 times, making a great step for real-time volume reconstruction.