MR görüntüleri kullanarak MS bölütlemesi

Multiple Sclerosis (MS) is a disease of the central nervous system, brain and spinal cord. The inherent heterogeneity of MS lesions, which is reflected on MR images, causes some of the difficulties encountered while attempting to separate lesions from healthy brain tissue and cerebro-spinal fluid. Automated segmentation of the lesions is useful for trace and diagnosis of MS. In this study, different tissue classification techniques are compared for segmentation of the MS lesions. Artificial neural networks are considered as supervised segmentation techniques and clustering algorithms are considered as unsupervised techniques. Fuzzy C-Means, K-Means and K-medoid clustering techniques with Principle Component Analysis, Independent Component Analysis and Multi Layer Back Propagation Neural Network algorithms are compared and results are evaluated. Keywords: Multiple Sclerosis, Magnetic Resonance Imaging, Segmentation, Artificial Neural Network, Fuzzy C-Means, K-Means, Clustering, Principle Component Analysis (PCA), Independent Component Analysis (ICA).