EU-Net: Efficient Dense Skip-Connected Autoencoder for Medical Image Segmentation
Lizhuang Liu, Jiacun Qiu, Ke Wang, Qiao Zhan, Jiaxi Jiang, Zhenqi Han, Jianxin Qiu, Tian Wu, Jinghang Xu, Zheng Zeng · IEEE Access · 2023
Recently, deep learning has made an important contribution to the development of medical image segmentation, such as U-Net and its variants. However, the existing methods based deep learning perform poorly in the small structures, structure continuity and efficiency. In this paper, we present EU-Net, an efficient and more powerful U-Net-like architecture for medical image segmentation, which consists of a lightweight encoder and decoder that are connected through dense skip-connection. To further improve the robustness of the EU-Net, chain EU-Net is proposed. Chain EU-Net is based on a streamlined architecture that uses multiple EU-Net to build light weight deep neural networks by dense skip-connection. We evaluate the EU-Net and chain EU-Net on three typical medical image segmentation tasks: GLAS (Gland segmentation) dataset, RITE (Retinal Images Vessel Tree Extraction) dataset and LiTS (Liver Tumor Segmentation Challenge) dataset. In addition, we used PUFH (Peking University First Hospital) dataset. Experimental results show that the proposed methods achieve state-of-the-art performance with very few parameters.