Brain Segmentation is a standard preprocessing step for neuroimaging applications, but can however be subject to differences in MR acquisition that can lead to added noise, bias field and / or partial volume effect. To address those protocol differences, we therefore present a generic supervised framework, using consecutively two deep learning networks, to produce a fast and accurate brain extraction aimed at being robust across MR protocol variations. While we only trained our network on Human Connectome Project 3T dataset, we can still achieve state-of-the-art results on1.5T cases from LPBA dataset.
This abstract and the presentation materials are available to members only; a login is required.