We explore methods for data augmentation in neuroimaging. Specifically, we investigate the use of 3D Transfontanellar Ultrasound (3D US) for augmenting 2D datasets of neonatal neuroimages, and we also synthesize an artificial dataset of images using Generative Adversarial Networks (GANs). This dataset consists on 2D slices of 3D US of the brain of neonates which has been successfully used to train an unconditional GAN for generating 2D US images as described in the presentation with DOI: 10.5281/zenodo.14917011