Flexible Sampling for Long-tailed Skin Lesion Classification

TL;DR

This work tackles the challenge of long-tailed skin lesion classification where rare disease classes are underrepresented. It introduces Flexible Sampling, a curriculum-learning framework that first builds balanced representations via self-supervised learning, then constructs anchor-point prototypes and trains an inference model to estimate class learning difficulty. A learning-status-aware curriculum module dynamically queries additional samples, guided by per-class accuracy and instance uncertainty, to train the model from easy to hard. Experiments on ISIC-2019-LT and ISIC-Archive-LT show state-of-the-art performance, especially in tail classes, suggesting significant practical impact for improving rare-lesion recognition in clinical settings.

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.

Figures (3)

• Figure 1: An illustration of learning difficulty. RS denotes re-sampling. The sub-figure (a) shows two kinds of lesions melanoma and vascular lesion from the majority and minority classes respectively. The sub-figure (b) shows that naively over-sampling those minority classes can only achieve marginal performance gains on VASC but seriously hurts the recognition accuracy on MEL.
• Figure 2: The overview of our proposed framework.
• Figure 3: The performance on several classes from head/tail classes on three methods.