This paper describes an automatic technique to segment the ventricles from brain MRI of young children. The segmentation approach involves skull stripping, extraction of the total cerebrospinal fluid (CSF), isolation of extra-cerebral CSF, removal of CSF in the sulci and fissures and final ventricle segmentation. The segmentation achieves a mean Dice ratio of 0.9246, sensitivity of 0.9363 and specificity of 0.9988. The novelty of the method is as follows: (1) it is completely devoid of brain atlases, templates and spatial priors; (2) it requires no prior training or human involvement. © 2016 IEEE.