Computed Tomographic Angiography (CTA) has become a popular image modality for the evaluation of arteries and the detection of narrowings. For an objective and reproducible assessment of objects in CTA images, automated segmentation is very important. However, because of the complexity of CTA images it is not possible to find a single parameter setting that results in an optimal segmentation for each possible image of each possible patient. Therefore, we want to find optimal parameter settings for different CTA images. In this paper we investigate the use of Fitness Based Partitioning to find groups of images that require a similar parameter setting for the segmentation algorithm while at the same time evolving optimal parameter settings for these groups. The results show that Fitness Based Partitioning results in better image segmentation than the original default parameter solutions or a single parameter solution evolved for all images.

doi.org/10.1007/978-3-540-78761-7_28, hdl.handle.net/1765/29716
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Erasmus MC: University Medical Center Rotterdam

Eggermont, A., Li, R., Bovenkamp, E., Marquering, H., Emmerich, M., van der Lugt, A., … Reiber, J. (2008). Optimizing computed tomographic angiography image segmentation using Fitness Based Partitioning. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4974 LNCS, pp. 275–284). doi:10.1007/978-3-540-78761-7_28