A Novel Segmentation Method of MR Brain Images Based on Genetic Algorithm

In the quantitative analysis of brain tissues (white matter, gray matter and cerebrospinal fluid) in magnetic resonance (MR) brain images, segmentation is the preliminary step. This paper proposes a novel segmentation method based on k-means objective function combined genetic algorithm, which is known for its global optimum searching ability. The method operates slice by slice via three main steps: (1) the non-brain tissues are removed from the original images using level set method, (2) the bias in the images which is caused by the inhomogeneity in the magnetic field is corrected by statistic method, and (3) the brain tissues are classified by k-means objective function combined genetic algorithm. The performance of the segmentation method was evaluated by the comparison with the fuzzy c-means (FCM) algorithm which is commonly used in segmentation of MR brain images. The accuracy of the proposed method is 3.21% higher than that of FCM algorithm. The time consumed in the proposed method is 0.596 second per image.