Parallelization of filtered back-projection algorithm for computed tomography

Processing time has become increasingly a major factor in computed tomography, hence the need for reconstruction and real-time diagnostics. Since the filtered back-projection algorithm (FBP) requires significantly intensive computational time when the amount of data becomes increasingly large; which is the case of computed tomography, parallel computing technique is used in Mat-Lab, as it enables the implementation of multi-tasking. Exploiting this advantage by partitioning the input data and allowing each core of cluster to work on its own sub-image. Therefore, FBP algorithm was applied simultaneously on all cores. All the reconstructed sub-images will be sent to client in order to be gathered. The comparison results between sequential FBP algorithm and parallel algorithm, show that the parallel computing process is up to seven times more efficient than the sequential algorithm on the voluminous images size, otherwise the small images results were poor because of the workers intercommunications that require times comparing with the sequential algorithm that takes a short computational time.