The paper is addressed to detect the parameters of a sphere-center coordinates and radius based on a stack of CT slices. It is proposing a new hierarchical Hough transform approach. In the first step, all slices are taken into consideration sequentially and a 2D accumulator array is used to obtain the coordinates (x"0,y"0), the projecting value of the sphere center into every X-Y-plane. In this step, also a new type of 2D Hough transform for circle or circular detection is proposed based on an effective point filtering. In the second step, the radii of the circles in the different slices are obtained using 1D accumulator arrays. In the last step, the coordinate z"0 and the radius R of the sphere are acquired using a 2D planar Hough transform based on the correlation between the radii of circles, the coordinates z of the slice and the sphere radius. The hierarchical Hough transform is applied to analyze the structure of femoral head of human hip joints. Compared to the established Hough transform techniques for 3D object detection, the hierarchical Hough transform reduces storage space and calculation time significantly and it has a good robustness to noise in the images.