Paul Aljabar1, Mellisa S. Damodaram2, Mary A. Rutherford3, Daniel Rueckert1, Joseph V. Hajnal3
1Visual Information Processing group,
Department of Computing, Imperial College London,
Automatic segmentation of the cortex from fetal brain MRI has potential as a significant tool in developmental neuroscience. We developed an accurate and robust method for extracting the cortex based on creating subject specific cortical priors using label propagation from automatically produced neonatal atlases. The method was tested on 12 fetal subjects with gestational age range from 20 -30 weeks, imaged using single shot Fast Spin Echo sequences with Slice to Volume (SVR) 3D reconstruction. The method was validated against manual segmentation and found to yield a mean error of 1.151.03mm.