Automated knowledge-based segmentation and analysis of the hip bones and cartilages using multi-slice CT data

摘要

In this research, two independent multi-step methods for automatic segmentation of the hip femoral and acetabular cartilages, femur and pelvis bones from CT images are presented. In data acquisition, by injecting the contrast media in the hip joint, the hip articular space is enhanced in CT images. The hip bones and cartilages are then extracted based on available anatomical assumptions, employing quantitative measures and techniques such as radial differentiation and image bottom hat (IBH) as well as proposing several heuristic techniques. After segmentation, applying a marching cube surface rendering technique, three-dimensional visualisation of segmented cartilages and bones followed by thickness map estimation of the hip cartilages is performed. Manual segmentations of experts were employed as gold standard for evaluating the results. The proposed techniques were effective in the presence of 20 sets (5120 images) of actual in vivo hip CT data.