ABSTRACT
Image-based atlas selection has been used for more than two decades as an approach to improve atlas-based auto-contouring. Most published investigations have shown some level of improvement level in the performance of auto-contouring over random selection of atlases when using image-based selection. In this chapter, published research into atlas-selection is reviewed, asking the question, “How close are we to optimal atlas selection?” An experiment is presented that assesses the most common approach to atlas selection – ranking similarity based on normalized mutual information between the atlas and the test case – using the 2017 AAPM Thoracic Auto-segmentation Challenge data. In phrasing the evaluation with respect to optimality, it is seen that while there is some improvement in auto-contouring performance, atlas selection is far from optimal. The chapter concludes with an optimistic outlook that there is still room for improvement in atlas selection, but also with a recommendation to consider selection ranking in future studies.
