Harmonious parameters and performance: Lightweight convolutional stage and local feature weighted fusion MLP for medical image segmentation

Open Webpage

Yan-Xu Chen, Yu-Jie Xiong, Xihe Qiu, Chun-Ming Xia

Published: 2024, Last Modified: 08 Apr 2025Biomed. Signal Process. Control. 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A lightweight model (UConvNeXt) for medical image segmentation is proposed.•UConvNeXt surpasses SOTA methods with similar scale.•Compared to standard UNet, UConvNeXt is 17x smaller, and 3x faster.