Automatic segmentation of neonatal brain magnetic resonance images

Abstract

This paper provides an overview of magnetic resonance imaging of the neonatal brain, presents the challenges involved in segmenting the neonatal brain images and reviews the existing techniques for automatic segmentation, including atlas-based probabilistic segmentations and morphology based brain segmentation. It compares the various methods in practice and highlights their limitations, particularly the inadequacies in segmenting the myelinated portions of the brain. It also proposes a new approach to overcome these shortcomings.