The Importance of Skip Connections in Biomedical Image Segmentation

Abstract

In this paper, we study the influence of both long and short skip connections on Fully Convolutional Networks (FCN) for biomedical image segmentation. In standard FCNs, only long skip connections are used to skip features from the contracting path to the expanding path in order to recover spatial information lost during downsampling. We extend FCNs by adding short skip connections, that are similar to the ones introduced in residual networks, in order to build very deep FCNs (of hundreds of layers). A review of the gradient flow confirms that for a very deep FCN it is beneficial to have both long and short skip connections. Finally, we show that a very deep FCN can achieve near-to-state-of-the-art results on the EM dataset without any further post-processing.

arXiv commentAccepted to 2nd Workshop on Deep Learning in Medical Image Analysis (DLMIA 2016); Added references

References in corpus (5)

• Deep Residual Learning for Image Recognition
• Brain Tumor Segmentation with Deep Neural Networks
• FitNets: Hints for Thin Deep Nets
• Going Deeper with Convolutions
• An Iterative Convolutional Neural Network Algorithm Improves Electron Microscopy Image Segmentation

Cited by in corpus (27)

• MultiResUNet : Rethinking the U-Net Architecture for Multimodal Biomedical Image Segmentation
• ResUNet-a: a deep learning framework for semantic segmentation of remotely sensed data
• Recent advances and clinical applications of deep learning in medical image analysis
• Sharpness-aware Low dose CT denoising using conditional generative adversarial network
• UNet++: A Nested U-Net Architecture for Medical Image Segmentation
• Rotation equivariant vector field networks
• Machine Learning for Precipitation Nowcasting from Radar Images
• RedNet: Residual Encoder-Decoder Network for indoor RGB-D Semantic Segmentation
• CNN-based Segmentation of Medical Imaging Data
• The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation
• Revisiting Single Image Depth Estimation: Toward Higher Resolution Maps with Accurate Object Boundaries
• CascadePSP: Toward Class-Agnostic and Very High-Resolution Segmentation via Global and Local Refinement
• A Multi-Stage Multi-Task Neural Network for Aerial Scene Interpretation and Geolocalization
• Recent Advances in the Applications of Convolutional Neural Networks to Medical Image Contour Detection
• Learning Steerable Filters for Rotation Equivariant CNNs
• Image Segmentation by Iterative Inference from Conditional Score Estimation
• IVUS-Net: An Intravascular Ultrasound Segmentation Network
• HarDNet: A Low Memory Traffic Network
• Learning Normalized Inputs for Iterative Estimation in Medical Image Segmentation
• Beyond Forward Shortcuts: Fully Convolutional Master-Slave Networks (MSNets) with Backward Skip Connections for Semantic Segmentation
• Deeply Cascaded U-Net for Multi-Task Image Processing
• DRUNET: A Dilated-Residual U-Net Deep Learning Network to Digitally Stain Optic Nerve Head Tissues in Optical Coherence Tomography Images
• Towards Label-Free 3D Segmentation of Optical Coherence Tomography Images of the Optic Nerve Head Using Deep Learning
• Mixed context networks for semantic segmentation
• SynNet: Structure-Preserving Fully Convolutional Networks for Medical Image Synthesis
• Focusing on Shadows for Predicting Heightmaps from Single Remotely Sensed RGB Images with Deep Learning
• Ensemble of Multi-sized FCNs to Improve White Matter Lesion Segmentation