A Longitudinal Diffeomorphic Atlas-Based Tissue Labeling Framework for Fetal Brains using Geodesic Regression

Roxane Licandro, Georg Langs, Gregor Kasprian, Robert Sablatnig, Daniela Prayer, Ernst Schwartz

Abstract: The human brain undergoes structural changes in size and in morphology between the second and the third trimester of pregnancy, corresponding to accelerated growth and the progress of cortical folding. To make fetal brains comparable, spatio-temporal atlases are used as a standard space for studying brain development, fetal pathology locations, fetal abnormalities or anatomy. The aim of this work is to provide a continuous model of brain development and to use it as base for an automatic tissue labeling framework. This paper provides a novel longitudinal fetal brain atlas construction concept for geodesic image regression using three different age-ranges which are parametrized according to the developmental stage of the fetus. The dataset used for evaluation contains 45 T2−weighted Magnetic Resonance (MR) volumes between Gestation Week (GW) 18.0 and GW 30 day 2. The automatic tissue labeling framework estimates cortical segmentations with a Dice Coefficient (DC) of up to 0.85 and ventricle segmentations with a DC of up to 0.60