An intelligent interactive segmentation method for the joint space in osteoarthritic ankles

Clinical reality is full of complex images that cannot be segmented automatically with current computer vision technology, requiring intensive user intervention. In [1] and [2] we proposed a framework for the systematic development of intelligent interactive segmentation techniques that aim at repeatable and predictable results obtained via efficient interaction. In this paper we apply this framework to segment the joint space boundary of osteoarthritic ankles. The solution is based on a heterogeneous boundary representation implemented with a new piecewise deformable model. User intervention is necessary only when this model fails, being performed via specialized interactive tools. Results obtained by a non-medical user are presented, indicating improvement over the manual practice in terms of accuracy and repeatability.