Research Of Brain MR Images Segmentation Algorithm Based On Fuzzy Clustering Theory

The brain disease is a common threat to human health. How to use medical imaging to analyze the brain tissue qualitatively and quantitatively, and then analyze the relationship between brain tissue and brain diseases has become a hot topic. Magnetic Resonance Imaging (MRI) is particularly effective for soft tissue such as brain. Therefore, MRI has been widely used in clinical practice. Accurate segmentation of the brain structure is the premise for the subsequent processing of the brain structure analysis and3D visualization. It can improve the reliability of the brain diseases diagnosis and the effectiveness of treatment programs.This paper first gave a overview on image segmentation methods at home and abroad. Among those methods, for the brain MR image segmentation problems, the methods based on fuzzy clustering theory show many advantages and have a great prospect. Therefore, this paper will focus on fuzzy clustering algorithm, the main work and research results are as follows.For the segmentation of brain MR images affected by noise, an improved FCM algorithm named a-MFFCM combined with Markov random field (MRF) has been proposed. The algorithm takes full advantage of the MRF model to introduce the spatial information, which overcomes the poor anti-noise performance of FCM caused by losing consider spatial information. According to different types of pixels, an adaptive weight has been proposed to control the combination weight of FCM and MRF reasonably. The simulation results show that the proposed method is superior to the existing fuzzy clustering algorithm in noise immunity.For the segmentation of brain MR images with intensity inhomogeneity, an improved BCFCM (Bias-Corrected FCM) algorithm named GBCFCM based on global information has been proposed. The introduction of global information makes the segmentation objective function control by both regional information and global information. The algorithm inherits the advantage of the BCFCM algorithm in the intensity inhomogeneity image segmentation, and overcomes it cannot accurately segment the heavy bias field region and edge region for losing consider the global information. The simulation results show that the algorithm is better than some of the existing fuzzy clustering algorithm in the intensity inhomogeneity image segmentation.For the segmentation of brain MR images with both noise and bias field, an improved the CLIC (Coherent Local Intensity Clustering) algorithm named CLICNL based on non-local information (Non Local) has been proposed. The introduction of non-local area information makes the segmentation algorithm objective function control by both local area information and non local area information. The local coherence of the algorithm criterion function ensures the smoothness of the bias field, and the non local term ensures the effective noise-anti ability. So the algorithm has both anti-noise and intensity inhomogeneity regulation ability. The simulation results show that the algorithm has strong anti-noise and bias field estimation ability, and the accuracy of segmentation is higher than some of the existing fuzzy clustering algorithm.