Brain surface segmentation of Magnetic Resonance images of the fetus

In this work we present a method for the image analysis of Magnetic Resonance Imaging (MRI) of fetuses. Our goal is to segment the brain surface from multiple volumes (axial, coronal and sagittal acquisitions) of a fetus. To this end we propose a two-step approach: first, a Finite Gaussian Mixture Model (FGMM) will segment the image into 3 classes: brain, non-brain and mixture voxels. Second, a Markov Random Field scheme will be applied to re-distribute mixture voxels into either brain or non-brain tissue. Our main contributions are an adapted energy computation and an extended neighborhood from multiple volumes in the MRF step. Preliminary results on four fetuses of different gestational ages will be shown.