Swarm Intelligence for Medical Volume Segmentation: The Contribution of Self-reproduction

For special applications in diagnostics for oncology the analysis of imaging data from Positron Emission Tomography (PET) is obfuscated by low contrast and high noise. To deal with this issue we propose a segmentation algorithm based on Ant Colony Optimization (ACO) and evolutionary selection of ants for self reproduction. The self reproduction approach is no standard for ACO, but appears to be crucial for volume segmentation. This investigation was focused on two different ways for reproduction control and their contribution to quantity and quality of segmentation results. One of the evaluated methods appears to be able to replace the explicit ant movement through transition rules by implicit movement through reproduction. Finally the combination of transition rules and self reproduction generates best reproducible segmentation results.