TY - GEN
T1 - Optimizing computed tomographic angiography image segmentation using Fitness Based Partitioning
AU - Eggermont, Jeroen
AU - Li, Rui
AU - Bovenkamp, Ernst G.P.
AU - Marquering, Henk
AU - Emmerich, Michael T.M.
AU - Van Der Lugt, Aad
AU - Bäck, Thomas
AU - Dijkstra, Jouke
AU - Reiber, Johan H.C.
PY - 2008
Y1 - 2008
N2 - 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.
AB - 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.
UR - https://www.scopus.com/pages/publications/47249092264
U2 - 10.1007/978-3-540-78761-7_28
DO - 10.1007/978-3-540-78761-7_28
M3 - Conference contribution
SN - 3540787607
SN - 9783540787600
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 275
EP - 284
BT - Applications of Evolutionary Computing - EvoWorkshops 2008
T2 - European Workshops on the Theory and Applications of Evolutionary Computation, EvoWorkshops 2008: EvoCOMNET, EvoFIN, EvoHOT, EvoIASP, EvoMUSART, EvoNUM, EvoSTOC, and EvoTransLog
Y2 - 26 March 2008 through 28 March 2008
ER -