ELU-Net: An Efficient and Lightweight U-Net for Medical Image Segmentation

Recent years have witnessed a growing interest in the use of U-Net and its improvement. It is one classic semantic segmentation networks with an encoder-decoder architecture widely used medical image segmentation. In series versions U-Net, U-Net++ has been developed as improved by designing nested dense skip connections, 3+ taking advantage full-scale connections deep supervision on aggregated feature maps. Each network own advantages encoder decoder. this paper, we propose efficient lightweight (ELU-Net) connections. The include same- large-scale from to fully extract features encoder. addition, proposed ELU-Net different loss functions discussed improve effect brain tumor learning including WT (whole tumor), TC (tumor core) ET (enhance tumor) new function DFK designed. effectiveness method demonstrated for dataset BraTS 2018 Challenge liver ISBI LiTS 2017 Challenge.