Parallel Iterative Ct Image Reconstruction On A Linux Cluster Of Legacy Computers

The expectation maximization (EM) algorithm is one of the iterative reconstruction (IR) algorithms that enable to reconstruct superior CT images, compared with the conventional filtered back-projection (FBP) method. The EM-IR algorithm can also be used when the data is incomplete. The major disadvantage of the EM-IR is its high demand on computation and slow reconstruction. To improve the performance, we developed a parallel EM on a Linux cluster composed of legacy (recycled) and heterogeneous PCs. The system, speed-up and efficiency from our parallel computations are presented. The study provides basic insight into how to conduct medical image reconstruction using junk PCs to simulate a heterogeneous parallel system.