Segmentation of 3D cellular networks from SR-micro-CT images

Bone fragility involved in diseases such as osteoporosis implicates many mechanisms at the cellular level. It was recently shown that the lacunar-canalicular network interconnecting osteocytes has a major role in mechanosensitivity. So far, this system has only been studied from 2D microscopic images. In a previous work, we demonstrated the feasibility of synchrotron radiation micro-CT with a voxel size of 0.28μm, to image the lacunar-canalicular porosity in 3D. Nevertheless, the segmentation of this dense network of slender channels with average diameters of ~300-600 nanometers, at the limit of the spatial resolution, is difficult. In this work we propose a level set based method to automatically segment this complex system. To this aim, we designed an automatic initialization process and we apply a post-processing filter. Quantitative results on a ground truth prepared image are presented. On real data sets, expert evaluation showed good results.