Skin lesion classification with multi-level fusion of Swin-T and ConvNeXt

• Published in: Sheng wu yi xue gong cheng xue za zhi Vol. 41; no. 3; p. 544
• Main Authors: Wang, Zetong, Zhang, Junhua, Wang, Xiao
• Format: Journal Article
• Language: Chinese
• Published: China Sichuan Society for Biomedical Engineering 25.06.2024
• Subjects: Accuracy; Algorithms; Cancer; Diagnosis; Diagnosis, Computer-Assisted - methods; Feature extraction; Humans; Image Interpretation, Computer-Assisted - methods; Information processing; Lesions; Modules; Public health; Skin - pathology; Skin cancer; Skin diseases; Skin lesions; Skin Neoplasms - classification; Skin Neoplasms - diagnostic imaging; Skin Neoplasms - pathology; Skin tests; Swin-T; Hierarchical inverted residual fusion module; Multi-level attention mechanisms; ConvNeXt; Skin lesion images
• ISSN: 1001-5515
• DOI: 10.7507/1001-5515.202305025

Skin cancer is a significant public health issue, and computer-aided diagnosis technology can effectively alleviate this burden. Accurate identification of skin lesion types is crucial when employing computer-aided diagnosis. This study proposes a multi-level attention cascaded fusion model based on Swin-T and ConvNeXt. It employed hierarchical Swin-T and ConvNeXt to extract global and local features, respectively, and introduced residual channel attention and spatial attention modules for further feature extraction. Multi-level attention mechanisms were utilized to process multi-scale global and local features. To address the problem of shallow features being lost due to their distance from the classifier, a hierarchical inverted residual fusion module was proposed to dynamically adjust the extracted feature information. Balanced sampling strategies and focal loss were employed to tackle the issue of imbalanced categories of skin lesions. Experimental testing on the ISIC2018 and ISIC2019 datasets yielded acc