微型电脑应用

计算机层析成像技术,在医学和工业等诸多领域中有着广泛应用。在三维锥束CT图像重建算法中,基于圆形轨道和二维平板探测器的FDK算法最为著名。传统CPU上实现的FDK算法,计算复杂性主要集中在所谓的反投影阶段,占据了整个重建时间的99%。给出了基于CUDA统一计算架构的FDK算法的GPU实现,对于整个重建过程获得了超过百倍的加速。

Computerized Tomography (CT) is widely used in medicine, industry and elsewhere. Different 3D cone-beam CT image reconstructionalgorithms have been developed including the well-known Feldkamp-Davis-Kress (FDK) algorithm applicable to cone-beam geometry with a circulartrajectory and a flat two-dimensional detector. The most computationally intensive step in the FDK algorithm is the so-called back-projection operationthat takes up to 99% of the reconstruction time in a CPU-based implementation. This paper presents a GPU-based implementation of FDK algorithm oncompute unified device architecture (CUDA), which speeds up the reconstruction process by more than 100 times.