Active shape models exploiting slice-to-slice correlation in segmentation of 3D CTA AAA images

摘要

An automated method for the segmentation of thrombusudin abdominal aortic aneurysms (AAA) from CTA dataudis presented. Three segmentation schemes, inspired byudActive Shape Model (ASM) segmentation, were investigated.ud(1) The original ASM scheme as proposed byudCootes and Taylor [1], applied to sequential slices, usingudthe contour obtained in one slice as the initial contourudin the adjacent slice. (2) A similar approach,udsteered by profile greyvalue correlation with adjacentudslices rather than by correlation with profiles from theudtraining data and (3) as in (2), with additional attractionudto nearby edges.udA leave-one-out experiment was performed, usingudfive datasets containing 239 slices to segment. Bothudadapted ASM schemes yield considerably better resultsudthan the original scheme. Scheme 3 showed best overalludperformance. Using one manually delineated imageudslice as a reference, on average a number of 21udslices could be automatically segmented with an accuracyudwithin the bounds of manual inter-observer variability.