Automatic segmentation of newborn brain MRI using mathematical morphology

We propose a novel algorithm for the segmentation of newborn brain MRI, based on mathematical morphology. The algorithm combines image information with high-level anatomical knowledge to deal with the difficulties of newborn brain MRI segmentation (lower signal-to-noise ratio, reduced contrast and different brain structure compared to the adult brain). It robustly segments the brain globally (intracranial cavity, cerebellum, brainstem and the two hemispheres) and at tissue level (cortical gray matter, central gray matter, myelinated white matter, unmyelinated white matter and cerebrospinal fluid). Important advantages compared to existing methods are that the proposed algorithm does not require any manual interaction and that it does not require an atlas, whose construction would be tedious and time-consuming. Experimental results show good agreement with expert manual segmentations and qualitative superiority to state-of-the-art methods in the literature.