Brain tissue segmentation of fetal MR images

Abstract. We present a segmentation method for fetal brain tissues of T2w MR images, based on the well known Expectation Maximiza-tion Markov Random Field (EM-MRF) scheme. Our main contribution is an intensity model composed of 7 Gaussian distribution designed to deal with the large intensity variability of fetal brain tissues. The second main contribution is a 3-steps MRF model that introduces both local spatial and anatomical priors given by a cortical distance map. Prelim-inary results on 4 subjects are presented and evaluated in comparison to manual segmentations showing that our methodology can successfully be applied to such data, dealing with large intensity variability within brain tissues and partial volume (PV).