Affinely Registered Multi-object Atlases as Shape Prior for Grid Cut Segmentation of Lumbar Vertebrae from CT Images

Abstract

In this paper, we present a method for automatic segmentation of lumbar vertebrae from a given lumbar spinal CT image. More specifically, our automatic lumbar vertebrae segmentation method consists of two steps: affine atlas-target registration-based label fusion and bone-sheetness assisted multi-label grid cut which has the inherent advantage of automatic separation of the five lumbar vertebrae from each other. We evaluate our method on 21 clinical lumbar spinal CT images with the associated manual segmentation and conduct a leave-one-out study. Our method achieved an average Dice coefficient of 93.9 ± 1.0% and an average symmetric surface distance of 0.41 ± 0.08 mm.