Segmentação de lesões de esclerose múltipla utilizando clusterização iterativa em imagens estruturais de ressonância magnética

Abstract

Multiple Sclerosis (MS) is a neurodegenerative disease that causes vision, mobility and memory problems, which can lead the patient to serious locomotion difficulties. It is caused by damage to the myelin sheath. Analysis of magnetic resonance (MR) images is essential for the diagnosis of MS. Depending on the imaging sequence, lesions appear hyperintense when compared to other tissues. Recent studies use computer vision, image processing and machine learning techniques with the aim of automatically segmenting these lesions and facilitating the diagnosis of MS. This work presents a technique based on the iterative execution of clustering algorithms on FLAIR-type 3D MR images to segment areas containing injured tissue. This work also compares the K-Means and Gaussian Finite Mixture Model clustering algorithms in this task using Dice coefficient between the generated lesion masks and those delineated by a human expert.