A Fully Automated Method for Segmentation and Thickness Map Estimation of Femoral and Acetabular Cartilages in 3D CT Images of the Hip

Abstract

In this research, we propose a multi-step method for automatic segmentation of the hip femoral and acetabular cartilages from CT images. Differentiation of the acetabular and femoral cartilages in conventional CT images due to very narrow space and low intensity of the articular space is very difficult. For this reason, MRI of the hip under continuous leg traction is the choice for assessment of the hip articular cartilage. However, this procedure is time consuming, painful, and hence inconvenient. By injecting the contrast media in the hip during data acquisition, we separated the femoral and acetabular cartilage from each other. The multi-step approach for segmentation of the femoral and acetabular cartilages is as follows. We first enhance the dynamic range and contrast of the data set by a conventional procedure. We then estimate the center of the femoral head utilizing a Hough transform. The estimated center is used as a pivotal point for beginning the process. Next, based on anatomical knowledge about the femoral head shape and size, we derive a region of interest (ROI) for further operations. In this case, by employing an Image Bottom Hat technique and anatomical knowledge about the curved shaped of the cartilages, we extract the valleys between contrast media and hip bones. The valley between the contrast media and femoral head and pelvis bone is associated with the femoral and acetabular cartilages. The anatomical assumptions we apply in this research is valid in most hip joint images. After segmentation, we present a 3D visualization and thickness map estimation of the hip femoral and acetabular cartilages. The proposed model is successfully applied to fourteen sets (3584 images) of actual in vivo hip CT data.