Technology Category
Computer Vision: SegmentationMachine Learning: Mixture of Experts (MoE)Search and Optimization: Learning to Search
ScaleFusionNet: Transformer-Guided Multi-Scale Feature Fusion for Skin Lesion Segmentation
🤖 AI Summary
Accurate skin lesion segmentation is critical for quantitative medical analysis, yet performance is limited by ambiguous lesion boundaries and large inter-lesion scale variations. To address these challenges, we propose a hybrid encoder architecture that jointly models local details and global semantics. Our key contributions are: (1) the Cross-Attention Transformer Module (CATM), a novel design that synergistically integrates Swin Transformer–guided self-attention with deformable convolution to bridge the semantic gap between encoder and decoder; and (2) an adaptive multi-scale feature fusion mechanism that explicitly enhances boundary delineation. Evaluated on the ISIC-2016 and ISIC-2018 benchmarks, our method achieves Dice scores of 92.94% and 91.65%, respectively—outperforming state-of-the-art approaches. The implementation is publicly available.
Technology Category
Application Category
📝 Abstract
Melanoma is a malignant tumor originating from skin cell lesions. Accurate and efficient segmentation of skin lesions is essential for quantitative medical analysis but remains challenging. To address this, we propose ScaleFusionNet, a segmentation model that integrates Cross-Attention Transformer Module (CATM) and AdaptiveFusionBlock to enhance feature extraction and fusion. The model employs a hybrid architecture encoder that effectively captures both local and global features. We introduce CATM, which utilizes Swin Transformer Blocks and Cross Attention Fusion (CAF) to adaptively refine encoder-decoder feature fusion, reducing semantic gaps and improving segmentation accuracy. Additionally, the AdaptiveFusionBlock is improved by integrating adaptive multi-scale fusion, where Swin Transformer-based attention complements deformable convolution-based multi-scale feature extraction. This enhancement refines lesion boundaries and preserves fine-grained details. ScaleFusionNet achieves Dice scores of 92.94% and 91.65% on ISIC-2016 and ISIC-2018 datasets, respectively, demonstrating its effectiveness in skin lesion analysis. Our code implementation is publicly available at GitHub.
Problem
Research questions and friction points this paper is trying to address.
Accurate segmentation of skin lesions for melanoma diagnosis.
Integration of Cross-Attention Transformer for feature fusion.
Enhancement of lesion boundary detection and detail preservation.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Integrates Cross-Attention Transformer Module
Uses AdaptiveFusionBlock for multi-scale fusion
Employs Swin Transformer for feature refinement
🔎 Similar Papers
No similar papers found.
