Bone segmentation in human CT images / 生物医学工程学杂志

ABSTRACT

In 3D visualization of human skeleton, distinguishing bones from soft tissue in 2D CT slides is the first and most critical procedure. This article presents the methods for image pre-processing, segmentation and smoothing. 1733 CT images of human body from Visible Human Project provided by the American National Library of Medicine are treated in this paper. We use the technique of Chebyshev uniform approximation filtering for denoising and present a new simple adaptive threshold method in segmentation, which combines the similarity of consecutive slices with the region-growing method. In post-processing, we use the algorithms of mathematical morphology and multi-resolution filtering. The accuracy of segmentation is examined and certified by comparing the segmented images with the original one. The results also demonstrate a wide applicability of the method.