Automatic segmentation in MR brain images is important for quantitative analysis in large-scale studies with images acquired at all ages. This paper presents a method for the automatic segmentation of MR brain images into a number of tissue classes using a convolutional neural network. To ensure that the method obtains accurate segmentation details as well as spatial consistency, the network uses multiple patch sizes and multiple convolution kernel sizes to acquire multi-scale information about each voxel. The method is not dependent on explicit features, but learns to recognise the information that is important for the classification based on training data. The method requires a single anatomical MR image only. The segmentation method is applied to five different data sets: coronal T-weighted images of preterm infants acquired at 30 weeks postmenstrual age (PMA) and 40 weeks PMA, axial T-weighted images of preterm infants acquired at 40 weeks PMA, axial T-weighted images of ageing adults acquired at an average age of 70 years, and T-weighted images of young adults acquired at an average age of 23 years. The method obtained the following average Dice coefficients over all segmented tissue classes for each data set, respectively: 0.87, 0.82, 0.84, 0.86, and 0.91. The results demonstrate that the method obtains accurate segmentations in all five sets, and hence demonstrates its robustness to differences in age and acquisition protocol.

In the developing human brain, the cortical sulci formation is a complex process starting from 14 weeks of gestation onward. The potential influence of underlying mechanisms (genetic, epigenetic, mechanical or environmental) is still poorly understood, because reliable quantification in vivo of the early folding is lacking. In this study, we investigate the sulcal emergence noninvasively in 35 preterm newborns, by applying dedicated postprocessing tools to magnetic resonance images acquired shortly after birth over a developmental period critical for the human cortex maturation (26-36 weeks of age). Through the original three-dimensional reconstruction of the interface between developing cortex and white matter and correlation with volumetric measurements, we document early sulcation in vivo, and quantify changes with age, gender, and the presence of small white matter lesions. We observe a trend towards lower cortical surface, smaller cortex, and white matter volumes, but equivalent sulcation in females compared with males. By precisely mapping the sulci, we highlight interindividual variability in time appearance and interhemispherical asymmetries, with a larger right superior temporal sulcus than the left. Thus, such an approach, included in a longitudinal follow-up, may provide early indicators on the structural basis of cortical functional specialization and abnormalities induced by genetic and environmental factors.

The human connectome is the result of an elaborate developmental trajectory. Acquiring diffusion-weighted imaging and resting-state fMRI, we studied connectome formation during the preterm phase of macroscopic connectome genesis. In total, 27 neonates were scanned at week 30 and/or week 40 gestational age (GA). Examining the architecture of the neonatal anatomical brain network revealed a clear presence of a small-world modular organization before term birth. Analysis of neonatal functional connectivity (FC) showed the early formation of resting-state networks, suggesting that functional networks are present in the preterm brain, albeit being in an immature state. Moreover, structural and FC patterns of the neonatal brain network showed strong overlap with connectome architecture of the adult brain (85 and 81%, respectively). Analysis of brain development between week 30 and week 40 GA revealed clear developmental effects in neonatal connectome architecture, including a significant increase in white matter microstructure (P < 0.01), small-world topology (P < 0.01) and interhemispheric FC (P < 0.01). Computational analysis further showed that developmental changes involved an increase in integration capacity of the connectivity network as a whole. Taken together, we conclude that hallmark organizational structures of the human connectome are present before term birth and subject to early development.