Towards a 4D Spatio-Temporal Atlas of the Embryonic and Fetal Brain using a Deep Learning approach for Groupwise Image Registration

Wietske Bastiaansen, Melek Rousian, Régine Steegers-Theunissen, Wiro Niessen, Anton Koning, Stefan Klein

Keywords: Embryonic and Fetal Brain Atlas, Groupwise Image Registration, First Trimester Ultrasound, Deep Learning

TL;DR: The paper describes preliminary results from the development of a 4D spatio-temporal atlas of the embryonic and fetal brain based on 3D ultrasound.

Abstract:

Brain development during the first trimester is of crucial importance for current and future health of the fetus, and therefore the
availability of a spatio-temporal atlas would lead to more in-depth insight into the growth and development during this period. Here, we propose a deep learning approach for creation of a 4D spatio-temporal atlas of the embryonic and fetal brain using groupwise image registration. We build on top of the extension of Voxelmorph for the creation of learned conditional atlases, which consists of an atlas generation and registration network. As a preliminary experiment we trained only the registration network and iteratively updated the atlas. Three-dimensional ultrasound data acquired between the 8th and 12th week of pregnancy were used. We found that in the atlas several relevant brain structures were visible. In future work the atlas generation network will be incorporated and we will further explore, using the atlas, correlations between maternal periconceptional health and brain growth and development.