A Modified FCM-Based Brain Lesion Segmentation Scheme for Medical Images

Abstract

Segmentation of brain lesion from medical images is a critical problem in the present day. In this work, we have proposed a new distance metric for fuzzy clustering based classification of different brain regions via acquiring accurate lesion structures. The modified distance metric segments the images into different regions by calculating the distances between the cluster centers and object elements, and subsequently classify them via fuzzy clustering. The proposed method can effectively remove noise from the images, which results in a better homogeneous classification of the image. Our method can also accurately segment stroke lesion where the results are near to the ground truth of the stroke lesion. The performance of our method is evaluated on both magnetic resonance images (MRI) and computed tomography (CT) images of brain. The obtained results indicate that our method performs better than the standard fuzzy c-means (FCM), spatial FCM (SFCM), kernelized FCM methods (KFCM), and adaptively regularized kernel-based FCM (ARKFCM) schemes. © 2020, Springer Nature Singapore Pte Ltd.