Automatic segmentation of neonatal brain MRI

This paper describes an automatic tissue segmentation method for neonatal MRI. The analysis and study of neonatal brain MRI is of great interest due to its potential for studying early growth patterns and morphologic change in neurodevelopmental disorders. Automatic segmentation of these images is a challenging task mainly due to the low intensity contrast and the non-uniformity of white matter intensities, where white matter can be divided into early myelination regions and non-myelinated regions. The degree of myelination is a fractional voxel property that represents regional changes of white matter as a function of age. Our method makes use of a registered probabilistic brain atlas to select training samples and to be used as a spatial prior. The method first uses graph clustering and robust estimation to estimate the initial intensity distributions. The estimates are then used together with the spatial priors to perform bias correction. Finally, the method refines the segmentation using sample pruning and non-parametric density estimation. Preliminary results show that the method is able to segment the major brain structures, identifying early myelination regions and non-myelinated regions.