eduzhai > Applied Sciences > Engineering >

HookNet multi-resolution convolutional neural networks for semantic segmentation in histopathology whole-slide images

  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: We propose HookNet, a semantic segmentation model for histopathologywhole-slide images, which combines context and details via multiple branches ofencoder-decoder convolutional neural networks. Concentricpatches at multipleresolutions with different fields of view are used to feed different branchesof HookNet, and intermediate representations are combined via a hookingmechanism. We describe a framework to design and train HookNet for achievinghigh-resolution semantic segmentation and introduce constraints to guaranteepixel-wise alignment in feature maps during hooking. We show the advantages ofusing HookNet in two histopathology image segmentation tasks where tissue typeprediction accuracy strongly depends on contextual information, namely (1)multi-class tissue segmentation in breast cancer and, (2) segmentation oftertiary lymphoid structures and germinal centers in lung cancer. Weshow thesuperiority of HookNet when compared with single-resolution U-Net modelsworking at different resolutions as well as with a recently publishedmulti-resolution model for histopathology image segmentation

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×