Fully automated convolutional neural network-based affine algorithm improves liver registration and lesion co-localization on hepatobiliary phase T1-weighted MR images

The authors of this study aimed to assess the performance of a convolutional neural network (CNN) algorithm to register cross-sectional liver imaging series and its performance to manual image registration. The study included three hundred fourteen patients who underwent gadoxetate disodium-enhanced magnetic resonance imaging (MRI) and were retrospectively selected.

Key points

• Image registration across series can improve lesion co-localization and reader confidence
• Combining convolutional neural network-based segmentation with affine transformations created a fully automated three-dimensional registration method for magnetic resonance images of the liver
• This algorithm improved liver overlap and focal liver observation co-localization over standard manual registration.

Article: Fully automated convolutional neural network-based affine algorithm improves liver registration and lesion co-localization on hepatobiliary phase T1-weighted MR images

Authors: Kyle A. Hasenstab, Guilherme Moura Cunha, Atsushi Higaki, Shintaro Ichikawa, Kang Wang, Timo Delgado, Ryan L. Brunsing, Alexandra Schlein, Leornado Kayat Bittencourt, Armin Schwartzman, Katie J. Fowler, Albert Hsiao & Claude B. Sirlin

Written by

European Radiology Experimental