Solving emission tomography problems on vector machines

Shepp and Vardi’s maximum likelihood reconstruction algorithm for emission tomography has mostly been ignored by the medical community because of its perceived high computational costs. However, the algorithm is very suited to parallel and vector machines, and on these machines can be made economically feasible even on medically reasonable problems of 16000 variables. It is also a good test problem for parallel optimization algorithms for large simply bounded non-linear problems.