🤖 AI Summary
This work addresses the challenge of skin lesion classification, which often relies on segmentation masks or auxiliary models that hinder clinical deployment and increase computational overhead. The authors propose the PLCRD framework, which leverages lesion masks during training to construct a teacher model and transfers structured knowledge about lesion–context relationships to a student model that requires only raw images for inference, thereby enabling mask-free prediction. Innovatively, privileged mask information is transformed into transferable relational knowledge, circumventing direct feature alignment between heterogeneous architectures. The approach integrates multiple mechanisms—including diagnostic distribution transfer, attention propagation, lesion similarity alignment, and lesion–context affinity matching. Evaluated on HAM10000 and ISIC 2018, the method achieves macro F1 scores of 0.773 ± 0.018 and 0.732 ± 0.008, respectively, significantly advancing classification performance under mask-free conditions.
📝 Abstract
Accurate skin lesion classification can benefit from lesion segmentation masks, but requiring masks or an auxiliary segmentation model during inference reduces clinical practicality and increases computational complexity. This work introduces Privileged Lesion-Context Relational Distillation (PLCRD), a teacher-student framework that exploits lesion masks exclusively during training while preserving image-only inference. The privileged teacher jointly analyzes the original dermoscopic image and its mask-guided lesion region to learn lesion-specific and contextual diagnostic representations. An image-only student is then trained through complementary knowledge-transfer mechanisms that convey the teacher's diagnostic distribution, lesion-focused attention, inter-lesion relational geometry, and lesion-context structure. PLCRD decomposes deep representations into lesion and contextual embeddings and transfers their relational organization through inter-lesion similarity alignment, lesion-context affinity matching, separation regularization, and class-aware relational learning. This formulation avoids direct feature matching between heterogeneous teacher and student architectures and enables the student to internalize mask-informed diagnostic structure without accessing masks at deployment. The framework was evaluated on HAM10000 using lesion-disjoint data partitioning and externally validated on ISIC 2018 without retraining. PLCRD achieved a lesion-level macro-F1 of 0.773 +/- 0.018, balanced accuracy of 0.764 +/- 0.023, and macro-AUROC of 0.976 +/- 0.002 on HAM10000, together with a macro-F1 of 0.732 +/- 0.008 on ISIC 2018. The results indicate that privileged lesion annotations can be transformed into transferable relational knowledge, yielding a practical and interpretable approach to mask-free skin lesion classification.