A novel cerebrovascular segmentation approach based on Markov random field and particle swarm optimization algorithm

Abstract

In order to solve the complex problems of segmenting cerebral vessels with many branches, small shape, special position and various patterns, a novel approach based on markov random field (MRF) and particle swarm optimization algorithm (PSO) is proposed in this paper to accurately segment cerebral vessels. Firstly, an improved nonlocal means filtering is used to reduce the interference of correlated noise. Then a new finite mixture model (FMM) - two Gaussian distribution and one Rayleigh distribution is used to fit the intensity histogram of brain tissues. Moreover, the MRF is constructed and fused with the PSO to obtain the optimal parameters of FMM. The experimental results verified the high accuracy on cerebrovascular segmentation especially for those small vessels, and relative high robustness and generalization comparing with the classical methods. The method can be widely applied in the clinical prevention and diagnosis of cerebrovascular diseases.