Semantic-Enhanced Few-Shot Learning for Precise Dermatological Diagnosis

Accurate diagnosis of dermatological diseases remains challenging due to fine-grained inter-class similarities and high intra-class variability. Few-shot learning (FSL) offers a promising solution, yet existing methods often fail to capture subtle lesion characteristics. We propose the Semantic Feature Enhanced Prototypical Transformer (SFEPT), a unified framework for few-shot dermatological image classification. SFEPT employs a Swin Transformer backbone to extract hierarchical features and a semantic-aware enhancement module to enrich query representations. Additionally, a prototype rectification strategy refines class prototypes, enhancing robustness. Our evaluations on three dermatological benchmarks demonstrate superior performance, achieving 89.53% and 81.85% accuracy on SD-198 and Derm7pt, respectively, in 3-way 5-shot tasks. The code is available at https://github.com/AIYAU/FSL_dermatopathya.