Skip to Main Content
There are a variety of sensor systems deployed at border crossings and ports of entry throughout the world that scan for illicit nuclear material. These systems employ detection algorithms that interpret the output of the scans and determine whether additional investigation is warranted. In this work, we demonstrate an approach for comparing the performance of such detection algorithms. We optimize each algorithm by minimizing risk, which considers the probability distribution of threat sources and the consequence of detection errors. Our method is flexible and is easily adapted to many different assumptions regarding the probability of a conveyance containing illicit material and the relative consequences of false positive and false negative errors. This approach can help developers and decision makers identify optimal settings for these algorithms. We illustrate the method by comparing the risk from two families of detection algorithms and discuss the generalizability of the method.