Comparative Multiple Sclerosis Lesion Segmentation in Magnetic Resonance Images

Abstract

In this study, the unsupervised clustering method namely K-means algorithm is applied for identifying the multiple sclerosis (MS) lesions in magnetic resonance (MR) images automatically. MS lesion detection is essential for diagnosing the disease and monitoring its progression. The automated method aims to eliminate user-dependent classification errors and to improve computational capacity in detecting more reliable MS segmentation results. K-means algorithm that relies on k cluster number on data is addressed to determine lesions in pathological brain MR images. Comparative segmentation is aimed by generating an in-house developed binary image segmentation routine in MATLAB. Segmented regions are compared to the results of K-means algorithm with respect to the predefined ROIs of lesions. The proposed K-means lesion detection routine is applied on real brain MR images and the results are qualitatively compared, and the method manages to locate the lesions successfully.