Skip to main content
Fig. 2 | Insights into Imaging

Fig. 2

From: Automatic differentiation of ruptured and unruptured intracranial aneurysms on computed tomography angiography based on deep learning and radiomics

Fig. 2

Aneurysm Segmentation network overview. a The complete pipeline of AS network. 3D patches were cropped uniformly from original CTA images and corresponding vessel segmentation images. Patches were balanced between positive (containing aneurysm) and negative types. Both original CTA and vessel segmentation patches were sent to ResUNet1 to firstly detect aneurysm with two channels of output, a probability map and corresponding aneurysm size map. Combined with original CTA images, two output channels, two layers of ResUNet1 were resized and input to ResUNet2 for segmentation. Emerging predicted cubes from ResUNet2 obtained the predicted whole volume. b ResUNet1 architecture illustration. c ResUNet2 architecture illustration

Back to article page