Skip to main content
. 2020 Sep 24;11:4829. doi: 10.1038/s41467-020-18606-2

Fig. 9. Bone and vessel segmentation framework of the CerebralDoc software.

Fig. 9

a Schema and structure of the 3D ResU-net to automatically segment the skull and head and neck vessel from CTA slices. b The whole software pipeline is present. Skull segmentation and subtraction are automatically completed after two cascaded ResU-Nets firstly and the vessel is fully segmented by a physiological anatomical-based ResU-Net. After that, a revised model CGPM is proposed to fix any ruptured vessel by learning from the input data of the original CTA image without labeling, the current segmentation result generated by ResNet1, ResNet2, and ResNet3, and the labeled CTA images. c An ablation study was performed to justify the contribution from each component in the pipeline.