Real-time region-based segmentation of 3D inhomogeneous objects in medical images

Active geometric functions were recently introduced as a tool to perform real time 3-D segmentation. In the present paper we propose a B-spline formulation of the problem, which further improves the computational efficiency of the algorithm. We also introduce local region based energies, overcoming the limitations of the original method. The feasibility of real-time 3D segmentation in challenging images using the proposed algorithm is demonstrated in liver tumor data. The segmentation accuracy is comparable to a recently proposed state-of-the-art algorithm, but with CPU times below 0.1s.