UCM-NetV2: An efficient and accurate deep learning model for skin lesion segmentation

Abstract

Accurate segmentation of skin lesions from dermoscopic images is crucial for early skin cancer detection, yet variations in lesion appearance and image artifacts present challenges. This study proposes an efficient deep learning model, UCM-NetV2, to improve accuracy and computational efficiency. UCM-NetV2 enhances the UCM-Net architecture with a novel "cyber-structure" com- bining Multilayer Perceptron and CNN layers, improving prediction accuracy while maintaining an ultra-lightweight design with only 0.046 million parameters. Evaluations on the ISIC2017 and ISIC2018 datasets demonstrate that UCM-NetV2 outperforms existing methods in accuracy and com- putational efficiency, achieving up to 67 times faster inference speeds than U-Net and requiring less than 0.04 GFLOPs. These advancements make skin lesion analysis more accessible, particularly in resource-limited settings, enabling proactive skin health monitoring and facilitating teledermatology. To foster further innovation in mobile health diagnostics, the source code for UCM-NetV2 is on https://github.com/chunyuyuan/UCMV2-Net.