Accelerated generation of digitally reconstructed radiographs using parallel processing

In this paper we present an approach for speeding-up the generation of Digitally Reconstructed Radiographs (DRRs). DRRs are needed to confirm patient setup before preplanned clinical procedures such as robotic surgery or radiation therapy in a process known as 2D/3D medical image registration. Rendering DRR images is a computationally intensive process and is considered a bottleneck in 2D/3D registration and there has been some recent interest in developing fast rendering techniques. This paper explores high speed rendering of DRR images from a CT data volume by parallel processing on multiple CPU cores. We investigate the relation between the execution time of our parallel DRR algorithm, the number of cores, and the number of rays (resolution) which are used to render the DRR image. We also compare the quality of DRR images rendered using an approximate method and compare this with approaches proposed by others. Our experimental results demonstrate a speed-up of better than three times using 4 CPU cores and better than 5 times using 8 cores. Our approximate approach gives a peak signal-to-noise ratio (PSNR) of 37 dB. which is comparable to that produced by other approximate techniques proposed and represents an overall speed-up of 26 times compared with a conventional ray casting approach.