Regularize Network Skip Connections by Gating Mechanisms for Electron Microscopy Image Segmentation

Recently, one earliest skip connected networks named Lmser was revisited and its convolutional layer based version named CLmser was proposed. This paper studies CLmser for segmentation (shortly CLmser-S) of Electron Microscopy (EM) images and also one further development. First, we experimentally show that CLmser-S outperforms the popular U-Net and save many free parameters. Second, we combine one newest formulation named Flexible Lmser (F-Lmser) and CLmser-S into a version called F-CLmser-S, together with learned masks replacing the similarity based one used in F-Lmser for implementing fast-lane skip connections. Experimental results on the ISBI 2012 EM dataset show that F-CLmser-S improves CLmser and achieves competitive performance with state-of-the-art results.

1. Introduction

High-resolution Electron Microscopy (EM) image has been used in biomedical research to investigate the detailed structure of tissues, cells, organelles and so on. For example, EM images were used to study Drosophila brain structure [1] which required segmentation of neural structures from the images. Manual labeling of each element in the image requires by an expert human neuroanatomist. However, due to the visual complexity of the EM images, it can be time-consuming for human experts to interpret them one-by-one, which drives the demand for automated approaches.