Swin-SaSNet: Swin Transformer with Symmetry-aware and Soft-attention Network for Skin Lesion Classification

Read full text Summarise with VIRAT
Authors: Gururaj Nase, Arun A
Cited by: 0
References: 26

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Skin cancer is one of the most frequently diagnosed malignancies, and early detection is vital for improving patient survival.Automated skin lesion classification remains challenging due to visual similarity between lesion types, class imbalance, and background artifacts.This paper proposes Swin-SaSNet, a hybrid attention-based framework for accurate and interpretable dermoscopic skin lesion classification.The model employs a Swin Transformer backbone to capture hierarchical image representations and integrates two complementary attention modules.A Symmetry-Aware Cross-Attention mechanism models structural asymmetry by comparing features from original and mirrored images, while a Soft-Attention Boundary Enhancement module emphasizes lesion margins and suppresses irrelevant background regions.This dual-attention strategy improves the discrimination of color, texture, and shape patterns.Experiments conducted on the HAM10000 dataset demonstrate that Swin-SaSNet outperforms conventional convolutional and transformer-based approaches while providing interpretable attention maps that support transparent clinical decision-making.

Fields of Study & Subjects

Cutaneous Melanoma Detection and Management AI in cancer detection Advanced Neural Network Applications

Publication details

Journal / source
International journal of intelligent engineering and systems
Volume
19
Issue
4
Pages
607–619
Publication date
23 March 2026
Language
EN