This paper investigates the impact of both long and short skip connections on Fully Convolutional Networks (FCNs) for biomedical image segmentation. Standard FCNs use only long skip connections to recover spatial information lost during downsampling. The authors extend FCNs by adding short skip connections, similar to those in residual networks, to build very deep FCNs (hundreds of layers). They find that both types of skip connections are beneficial for convergence in very deep networks. The study demonstrates that a very deep FCN can achieve near-state-of-the-art results on the EM dataset without any post-processing. The paper also explores the learning behavior of the model with and without skip connections, showing that short skip connections stabilize updates and improve convergence speed. The results highlight the importance of both long and short skip connections in achieving high performance in deep FCNs for biomedical image segmentation.This paper investigates the impact of both long and short skip connections on Fully Convolutional Networks (FCNs) for biomedical image segmentation. Standard FCNs use only long skip connections to recover spatial information lost during downsampling. The authors extend FCNs by adding short skip connections, similar to those in residual networks, to build very deep FCNs (hundreds of layers). They find that both types of skip connections are beneficial for convergence in very deep networks. The study demonstrates that a very deep FCN can achieve near-state-of-the-art results on the EM dataset without any post-processing. The paper also explores the learning behavior of the model with and without skip connections, showing that short skip connections stabilize updates and improve convergence speed. The results highlight the importance of both long and short skip connections in achieving high performance in deep FCNs for biomedical image segmentation.