High-performance medical image registration using improved particle swarm optimization

Optimization of a similarity metric is an essential component in intensity-based medical image registration. In this paper, an improved variable neighborhood selection based particle swarm optimization (VNS-PSO) is proposed. The PSO algorithm is co-operative, population-based global search swarm intelligence mataheuristics. The improved version of PSO algorithm possesses better ability to escape from the local minima to the global optimum, and more adapts for intensity-based medical image registration. The performances of VNS-PSO algorithm and downhill simplex method to medical image registration are compared. Experimental results demonstrate that the improved VNS-PSO method is robust, accurate, efficient and more suitable for medical image registration.