This image analysis project is constructing a tool for performing computer assisted detection of defects in large CT data sets. We address two primary challenges: proving an algorithm that can reliably highlight the objects of interest, and coupling this algorithm to efficient and effective data management and interface routines. The goal of this project is to create an end to end tool that will efficiently extract from large CT data sets a manageable set of candidate defects that highlights 100% of the unambiguous (to a trained analyst) true defects and as many as possible of the ambiguous objects. Objective metrics for defect ambiguity are elusive given the ill defined nature of both the defects and their CT images (e.g., variations in size and contrast, defect/object density gradients). Therefore the true metric of success will be gaining the confidence of the Weapons Program NDE analysts in this tool, which will be a qualitative function of efficiency and accuracy. In consultation with Weapons Program NDE analysts voids it was decided to focus this initial effort on searching for small voids (order of tenths to tens of millimeters in extent). The program has produced a Tungsten ring with surrogate defects drilled into it …