Parametric 3D hip joint segmentation for the diagnosis of developmental dysplasia.

In this paper, we present a segmentation method for the hip joint from 3D ultrasound data. The proposed approach starts from a well-known energy formulation of the segmentation problem, and employs the extended local structure tensor as image feature in order to incorporate gray level and texture information in a common framework. Using the Kullback-Leibler distance as an intrinsic dissimilarity measure, the energy minimization is performed by estimating the optimal parameters of the sphere and paraboloid that best approximate the femoral head and acetabulum, respectively. A 2D level set segmentation of the iliac bone and an easy user interaction step allow for the introduction of the necessary constraints to make the energy minimization feasible. Experimental results over several data volumes show this approach to be capable of successfully approximating the anatomy of the hip joint by simple geometrical surfaces.