Vortrag 03

Bookmark(0)

 

Semantically Guided 3D Abdominal Image Registration with Deep Pyramid Feature Learning

Schumacher, Mona; Frey, Daniela; Ha, In Young; Genz, Andreas; Bade, Ragnar; Heinrich, Mattias
Universität zu Lübeck, Institut für Medizinische Informatik

Abstract

Deformable image registration of images with large deformations is still a challenging task. Currently available deep learning methods exceed classical non-learning-based methods primarily in terms of lower computational time. However, these convolutional networks face difficulties when applied to scans with large deformations. We present a semantically guided registration network with deep pyramid feature learning that enables large deformations by transferring features from the images to be registered to the registration networks. Both network parts have U-Net architectures. The networks are trained end-to-end and evaluated with two datasets, both containing contrast enhanced liver CT images and ground truth liver segmentations. We compared our method against one classical and two deep learning methods. Our experimental validation shows that our proposed method enables large deformation and achieves the highest Dice score and the smallest surface distance of the liver in constrast to other deep learning methods.

Schlüsselwörter: 

Paper:

Persönlicher Kontakt:

Abstract

Deformable image registration of images with large deformations is still a challenging task. Currently available deep learning methods exceed classical non-learning-based methods primarily in terms of lower computational time. However, these convolutional networks face difficulties when applied to scans with large deformations. We present a semantically guided registration network with deep pyramid feature learning that enables large deformations by transferring features from the images to be registered to the registration networks. Both network parts have U-Net architectures. The networks are trained end-to-end and evaluated with two datasets, both containing contrast enhanced liver CT images and ground truth liver segmentations. We compared our method against one classical and two deep learning methods. Our experimental validation shows that our proposed method enables large deformation and achieves the highest Dice score and the smallest surface distance of the liver in constrast to other deep learning methods.

Schlüsselwörter: 

Paper:

Persönlicher Kontakt: