Deep Learning Based Automated Brain Segmentation from Computed Tomography Scans

Motivation: While computed tomography (CT) imaging has been actively employed in clinical practice, its limited contrast for brain tissues makes it challenging to achieve precise brain segmentation. Goal(s): In this study, we developed a deep learning (DL)-based method enabling brain tissue segmentation from CT image. Approach: MRI-derived tissue labels were provided as ground truth to a DL network, where U-Net and VGG16 interact to each other for model optimization by means of a perceptual loss. Results: Results demonstrate the effectiveness of incorporating the perceptual loss to the model in preserving image details, and in terms of evaluation scores. Impact: The presented method, upon further validation and optimization, is expected to be a valuable means to a range of brain imaging studies where MRI is somehow not available.