Segmenting 3D X-Ray microtomography using two different approaches: Morphological filters and Artificial Ant Colony

Abstract

Three-dimensional X-Ray Micromotomography (3D µCT) has become an important tool to investigate bone morphology. Several investigators have searched a standard method for determining the optimal threshold value (optimal TH) to segment microtomographic images and quantify the bone morphology. The Conventional methods are based on subjective methods, and it is possible to obtain under or overestimated TH values. In this work, two automatic methods for segmenting 3D µCT are presented. The first one is based on Morphological Filters and Otsu's method (MFO), and the other one is based Artificial Ant Colony (AAC). Both method performances are evaluated quantitatively using a numerical phantom which mimics a trabecular bone. Besides, the proposed methods are applied to segment a microtomographic image from a rat vertebral bone. The segmented image of phantom based on AAC method showed results closer to the binary reference image of the numerical phantom and the histomorphometric parameters values presented smaller error (11.3%) than the one based on MFO (22.4%). When the methods were applied to segment a microtomographic image, the performances were compatible, and the histomorphometric parameter values present differences inferior to 9%.

Artificial Ant Colony: Features and applications on medical image segmentation

“…One can observe that the segmentation depicted by the proposed procedure is comparable to the one achieved by [12]. A quantitative evaluation comparing both methods' performance is presented in [13]. …”