Automated Segmentation of Acetabular Cartilage in MR images of the Hip

摘要

In the orthopedic field, quantitative analysis of the hip joint cartilages is of clinical importance. In this regard, segmentation of acetabular cartilage is required for further corresponding quantifications such as cartilage thickness map estimation and/or generic modeling of a cartilage shape. To date, MRI is considered as one of the most effective choices for diagnosis of cartilage diseases. In this research, we have developed a highly automated technique for segmentation of acetabular cartilages in volumetric MR images acquired by fat suppressed 3D fast spoiled gradient echo (SPGR) sequence. The multi-step approach consists of the following steps: (1) Automatic selection of the volume of interest by estimating the center of a sphere that approximates the femoral head using a Hough transform and anatomical knowledge of a femoral head size; (2) Automatic segmentation of the acetabulum and femoral head by adaptive thresholding and 3D morphological operations; (3) Automatic segmentation of a rough hip joint space utilizing anatomical constraint of the hip joint shape with respect to acetabulum and femoral head; (4) Accurate segmentation of the hip joint space by estimated femoral head center and enhanced cartilage regions utilizing first order 3D directional derivatives along radial directions originating from the sphere center; (5) Automatic segmentation of the acetabular cartilage by localization of cartilage boundaries along the normal lines associated with cartilage initial boundaries and employing a B-Spline curve fitting snake. The developed techniques require no operator intervention from the input of the original MR data to the final segmentation of the acetabular cartilage. The proposed method was successfully applied to 20 sets (1200 images) of actual in vivo hip MR data of ten cases.