SEGMENTATION OF LOW CONTRAST FEATURES IN BONE MICRO-CT IMAGES BY A CONSTRAINED REGION GROWING APPROACH BASED ON WATERSHED

Quantitative micro-CT using monochromatic X-ray beams provided by synchrotron sources opens new perspectives for assessing bone metabolism. It makes it possible to visualize remodeling regions with different mineral concentrations. However, so far their segmentation has only been performed by simple thresholding which is insufficient due to noise and weak contrast of remodeling regions with respect to bone. In addition, automatic methods are desired to process large set of 3D images. In this paper, we propose a new region growing based approach to segment each individual remodeling region. First, a non linear denoising technique based on wavelet transform is applied. Then, we propose a customized region growing algorithm taking into account some biological prior information. Furthermore, in order to include shape constraints, we suggest another strategy, based on distance map and watershed. Finally, results on experimental synchrotron radiation micro-CT images of cortical bone samples are presented