This paper presents a novel approach to fetal magnetic resonance image segmentation and biometric analysis of the posterior fossa's midline structures. We developed a semi-automatic segmentation method (based on a region growing technique) and tested the algorithm on images of 104 normal fetuses. Using the segmented regions of interest (posterior fossa, vermis, and brainstem), we computed four relative area ratios. Statistical and clinical analysis of our results showed that the relative development of these structures appears to be independent of pregnancy term. In an additional study of 23 pathological cases, one of the four measurements was always significantly different from the corresponding value observed in normal cases.