Segmentation of MR Brain Images : A Fuzzy Logic Approach

An approach is developed to MR brain images segmentation, based on pixel classification using Fuzzy Rule Based system and Fuzzy Similarity measures. The cerebral images are segmented into gray matter, white matter, and cerebrospinal fluid (CSF). Image preprocessing was first done to improve the quality of brain MR images and reducing artifacts. The feature vector was selected to be the pixel and its eight neighbours. The two methods implemented are of supervised nature where in the first we build fuzzy rules while in the second we build fuzzy prototypes. The classification in the first method uses fuzzy inference and implication techniques to derive the classes of images. The classification in the second method uses pattern matching and fuzzy similarity measures. These methods are tested using sets of MRI brain images. The results show the efficient and robust performance of these algorithms. In this paper a comparison of these algorithms with Fuzzy C-Means algorithm is presented. Keyword-I: Medical Imaging Keyword-II: Biomedical Application, Cerebrospinal Fluid, Feature Identification and Classification, Fuzzy C-Means, Fuzzy Logic, Fuzzy Rule-Based System, Fuzzy Set Theory, Fuzzy Similarity Measures, Gray Matter, Image Segmentation, Image Processing, Magnetic Resonance Imaging, Pattern Recognition, Rule Extraction, Rule Formulation, Supervised Learning, White Matter.