🤖 AI Summary
Skin lesion classification inherently exhibits a long-tailed distribution due to inter-subject variability and the rarity of certain diseases. Conventional long-tailed learning methods apply uniform class-level resampling, ignoring intra-class difficulty heterogeneity—thereby compromising discriminative capability. To address this, we propose a curriculum learning-driven dynamic sampling framework. It introduces a learnability-aware soft sampling mechanism anchored on class prototypes for difficulty estimation, coupled with a dynamic curriculum scheduling strategy that enables differential modeling and progressive training of easy versus hard samples. Evaluated under two long-tailed settings on the ISIC dataset, our method significantly outperforms existing state-of-the-art approaches, establishing new performance benchmarks for long-tailed skin lesion classification. This work is the first to integrate difficulty-aware sampling with prototype-guided curriculum learning, markedly enhancing discrimination for rare classes and overall generalization.
📝 Abstract
Most of the medical tasks naturally exhibit a long-tailed distribution due to the complex patient-level conditions and the existence of rare diseases. Existing long-tailed learning methods usually treat each class equally to re-balance the long-tailed distribution. However, considering that some challenging classes may present diverse intra-class distributions, re-balancing all classes equally may lead to a significant performance drop. To address this, in this paper, we propose a curriculum learning-based framework called Flexible Sampling for the long-tailed skin lesion classification task. Specifically, we initially sample a subset of training data as anchor points based on the individual class prototypes. Then, these anchor points are used to pre-train an inference model to evaluate the per-class learning difficulty. Finally, we use a curriculum sampling module to dynamically query new samples from the rest training samples with the learning difficulty-aware sampling probability. We evaluated our model against several state-of-the-art methods on the ISIC dataset. The results with two long-tailed settings have demonstrated the superiority of our proposed training strategy, which achieves a new benchmark for long-tailed skin lesion classification.
