The doctoral dissertation focuses on the segmentation of magnetic resonance images with deep learning. The automatic segmentations were of higher quality than those produced with other methods, and more accurate than those annotated by an expert, which is important for reproducibility.