The size of the synthetic set (namely 1000 high-resolution volumes) is a significant peril, due to scale (complexity), as well as runtime (duration). There are ways of addressing these issues. One is to distribute the work among several workstations. The other is to run assessments of different models in parallel, accumulating results and refining them progressively, rather than handle the assessments one at the time, i.e. serially. This has the advantage that results emerge rather early and give insight into whether the experiments are worthwhile or not.
The computer clusters will be running Fedora Core 5 with MATLAB 6.5. The time estimated for one assessment to be completed depends on the number of workstations used, but it most probably lies between the range of weeks and one month.