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

Abstract

An automated method for the segmentation of thrombus in abdominal aortic aneurysms (AAA) from CTA data is presented. Three segmentation schemes, inspired by Active Shape Model (ASM) segmentation, were investigated. (1) The original ASM scheme as proposed by Cootes and Taylor [1], applied to sequential slices, using the contour obtained in one slice as the initial contour in the adjacent slice. (2) A similar approach, steered by profile greyvalue correlation with adjacent slices rather than by correlation with profiles from the training data and (3) as in (2), with additional attraction to nearby edges. A leave-one-out experiment was performed, using five datasets containing 239 slices to segment. Both adapted ASM schemes yield considerably better results than the original scheme. Scheme 3 showed best overall performance. Using one manually delineated image slice as a reference, on average a number of 21 slices could be automatically segmented with an accuracy within the bounds of manual inter-observer variability.