3D MRI brain segmentation based on MRF and hybrid of SA and IGA

This paper proposes a novel combinational approach for statistical de-noising and segmentation of 3D magneticresonance images (MRIs) of the brain. The proposed method is based on Markov Random Field (MRF), conjunction with simulated annealing (SA) and improved genetic algorithm (IGA). MRF methods have been widely studied for segmentation. Despite the Markovianity which depicts the local characteristic, which allows a global optimization problem to be solved locally, MRF still has a heavy computation burden, especially when it isused with stochastic relaxation schemes such as SA. Although, search procedure of SA is fairly localized and prevents from exploring the same diversity of solutions, it suffers from several limitations. In comparison, GA has a good capability of global researching but it is weak in hill climbing. Therefore, the combination of these two methods may have the advantages of both procedures while alleviating their individual shortcomings and high computation complexity. Evaluation of proposed approach shows that our algorithm outperforms the traditionalMRF in both convergence speed and solution quality.

Keywords:

Magnetic Resonance Imaging , Markov Random Field (MRF) , Simulated Anealing , Improved Genetic Algorithm