A k-Dense-UNet for Biomedical Image Segmentation

Abstract

Medical image segmentation is the premise of many medical image applications including disease diagnosis, anatomy, and radiation therapy. This paper presents a k-Dense-UNet for segmentation of Electron Microscopy (EM) images. Firstly, based on the characteristics of the long skip connection of U-Net and the mechanism of short skip connection of DenseNet, we propose a Dense-UNet by embedding the dense blocks into U-Net, leading to deeper layers for better feature extraction. We experimentally show that Dense-UNet outperforms the popular U-Net. Secondly, we combine Dense-UNet with one of the newest U-Net variants called kU-Net into a network called k-Dense-UNet, which consists of multiple Dense-UNet submodules. Skip connections are added between the adjacent submodules, to pass information efficiently, helping the model to identify fine features. Experimental results on the ISBI 2012 EM dataset show that k-Dense-UNet achieves better performance than U-Net and some of its variants.