TinyU-Net: Lighter Yet Better U-Net with Cascaded Multi-receptive Fields

Junren Chen, Rui Chen, Wei Wang, Junlong Cheng, Lei Zhang, Liangyin Chen

Abstract: The lightweight models for automatic medical image segmentation have the potential to advance health equity, particularly in limited-resource settings. Nevertheless, their reduced parameters and computational complexity compared to state-of-the-art methods often result in inadequate feature representation, leading to suboptimal segmentation performance. To this end, we propose a Cascade Multi-Receptive Fields (CMRF) module and develop a lighter yet better U-Net based on CMRF, named TinyU-Net, comprising only 0.48M parameters. Specifically, the CMRF module leverages redundant information across multiple channels in the feature map to explore diverse receptive fields by a cost-friendly cascading strategy, improving feature representation while maintaining the lightweightness of the model, thus enhancing performance. Testing CMRF-based TinyU-Net on cost-effective medical image segmentation datasets demonstrates superior performance with significantly fewer parameters and computational complexity compared to state-of-the-art methods. For instance, in the lesion segmentation of the ISIC2018 dataset, TinyU-Net is \(52\times \), \(3\times \), and \(194\times \) fewer parameters, respectively, while being \(+3.90\%\), \(+3.65\%\), and \(+1.05\%\) higher IoU score than baseline U-Net, lightweight UNeXt, and high-performance TransUNet, respectively. Notably, the CMRF module exhibits adaptability, easily integrating into other networks. Experimental results suggest that TinyU-Net, with its outstanding performance, holds the potential to be implemented in limited-resource settings, thereby contributing to health equity. The code is available at https://github.com/ChenJunren-Lab/TinyU-Net.