3D Detection of Spheric and Aspheric Femoral Heads in Coronal MR Images of Patients with Legg-Calve-Perthes Disease Using the Spherical Hough Transform

Abstract

This paper presents a study of 3D femoral head detection in coronal magnetic resonance (MR) imaging sequences to detect the spheric and aspheric femoral heads using the Spherical Hough Transform (SHT). Shape structure and functionality of the femoral head can be affected by a wide range of hip disorders. In most of the computerized analysis and assessment cases of the hip joint, segmentation of the proximal femur and femoral head is performed on the image sequences that are provided by the several medical imaging modalities. Determination of initial points, detection of the tissue boundaries and segmentation of the tissues are important challenges in automatic or semi-automatic proximal femur segmentation methods. In this paper, we aimed to perform a preliminary study based on 3D femoral head detection in bilateral MR hip images in coronal plane of patients with Legg-Calve-Perthes disease (LCPD). Firstly, MR sections are divided vertically into two equal halves automatically to separate the healthy and pathological hip joints and then Canny's edge detection method is performed to obtain edge images. Finally, 3D SHT is performed to detect the spherical structure of femoral heads. Automatic detection of the femoral head with SHT provides a starting point for the segmentation of the femoral head ball and proximal femur. Furthermore, it also allows for 3D segmentation of the femoral head sphere with the radius information that it provides with the 3D sphere center coordinates. Performance evaluations and experimental results observed on 8 bilateral MR imaging sequences belong to 6 subjects show that proposed method has promising results.