DEU-Net: Dual-Encoder U-Net for Automated Skin Lesion Segmentation
Open Access
- 28 November 2023
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 134804-134821
Abstract
The computer-aided diagnosis (CAD) of skin diseases relies heavily on automated skin lesion segmentation, albeit presenting considerable challenges due to lesion diversity in shape, size, color, and texture, as well as potential blurry boundaries with surrounding tissues. Traditional Convolutional Neural Networks (CNN) typically underperform in this domain, given their inherent constraints in global context information capture. In the present study, we present a new U-shaped network, Dual-Encoder U-Net (DEU-Net), which is based on an encoder-decoder architecture. DEU-Net integrates a dual-encoder branch comprising a convolutional encoder and a transformer encoder, thereby facilitating the concurrent extraction of local features and global contextual information. Additionally, in order to enhance the performance of DEU-Net, we employ an integrated test-time augmentation technique. To ascertain the efficiency and superiority of our proposed methodology, we performed comprehensive experiments across four widely accessible skin lesion datasets, namely ISIC 2016, ISIC 2017, ISIC 2018, and PH2. The Dice coefficients achieved on these datasets were 92.90%, 87.16%, 90.81%, and 95.65%, respectively. These results demonstrate superior performance compared to most current state-of-the-art methods. The source code is released at https://github.com/alikm6/DEU-Net.Keywords
This publication has 60 references indexed in Scilit:
- RMMLP:Rolling MLP and matrix decomposition for skin lesion segmentationBiomedical Signal Processing and Control, 2023
- MaxViT: Multi-axis Vision Transformer2022
- UNeXt: MLP-Based Rapid Medical Image Segmentation Network2022
- NCRNet: Neighborhood Context Refinement Network for skin lesion segmentationComputers in Biology and Medicine, 2022
- Fully convolutional neural network with attention gate and fuzzy active contour model for skin lesion segmentationMultimedia Tools and Applications, 2022
- Boundary-Aware Transformers for Skin Lesion Segmentation2021
- DSNet: Automatic dermoscopic skin lesion segmentationComputers in Biology and Medicine, 2020
- UNet++: A Nested U-Net Architecture for Medical Image Segmentation2018
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015
- ImageNet Large Scale Visual Recognition ChallengeInternational Journal of Computer Vision, 2015