Preterm birth affecting approximately 10% of live births worldwide, disrupting brain development and significantly increasing the risk of neurological complications and disabilities. Accurate, repeated monitoring of brain growth is therefore essential in the neonatal intensive care unit (NICU). Although cranial ultrasound (cUS) is safe and portable, its clinical use remains limited to lesion detection. Measuring total brain volume as a biomarker is restricted because automated 3D cUS brain segmentation remains an unsolved challenge. This thesis which complements the GALATEA project, funded by the Horizon Europe pro-gramme under the Marie Skłodowska-Curie action presents an automated deep learning pipeline for brain segmentation in preterm neonates from 3D cUS volumes.