搜索鲁棒性增强方法以提高皮肤镜皮肤癌分类中的分布外泛化能力
Searching for Robust Augmentations to Improve Out-of-Domain Generalization in Dermoscopic Skin Cancer Classification
浏览论文内容
中文总结 AI 辅助
本研究针对皮肤癌分类的域偏移问题，在ISIC Archive等数据集上搜索增强策略，发现mix策略可显著提升分布外泛化性能，为皮肤癌分类器鲁棒性优化提供了新方向。
中文摘要 AI 辅助
背景/目标：皮肤镜皮肤病变分类器在跨成像设备、光照和采集伪影的域偏移下通常会损失准确率。我们研究数据增强如何提高二元良恶性分类器的鲁棒性，重点关注分布外（OOD）泛化能力。方法：在包含Derm7pt的多源ISIC Archive集合上搜索单一增强、光度组合和复合策略，使用ConvNeXt-Large作为骨干网络，以ROC-AUC为指标。按病变ID进行划分，将HAM10000和ISIC 2019-2020作为主要源不相交的OOD测试集保留。结果：最大的OOD增益来自mix策略，光度变换是最有用的OOD操作。在相同保留源的扩展集合上，增益为+0.053（95%置信区间+0.045至+0.061，p<0.001），在4个训练种子间一致（各种子ROC-AUC：基线0.761-0.775，mix策略0.806-0.829）。在小型独立临床集合上，单检查点灵敏度从0.591升至0.818，但该结果基于22例恶性病例且未在种子间持续。结论：对真实域偏移源建模的增强可能比最大化域内准确率更重要。由于策略是在用于评估的相同源上选择的，需采用源不相交选择协议才能将该效应量解读为无偏。
英文摘要
Background/Objectives: Dermoscopic skin-lesion classifiers lose accuracy when images arrive from a new clinic or a new device. We asked which data augmentations reduce that loss, and measured the effect under a protocol that keeps policy selection separate from policy evaluation. Methods: A ConvNeXt-Large binary malignant-versus-non-malignant classifier was trained on six dermoscopic sources (25,903 images); HAM10000 and ISIC 2016-2020 were held out of training entirely. Single augmentations, photometric combinations and eleven composite policies were ranked on a development split of 1511 held-out images. The winning policy was then evaluated on a confirmation set of 8073 held-out images that took no part in that ranking and from which we removed every image sharing a lesion identifier with the training data and every image contributed by an institution represented in training. Both policies were retrained with four random seeds each and compared with an exact permutation test. Results: The mix policy raised confirmation-set ROC-AUC from 0.787 to 0.826 (+0.039; per-seed ranges 0.772-0.797 and 0.815-0.840, non-overlapping; exact permutation p=0.029), with the same direction on each contributing source. At matched sensitivity the gain is larger in clinical terms: specificity rose from 0.612 to 0.713 at a sensitivity of 0.80, and from 0.284 to 0.397 at a sensitivity of 0.95. In-domain ROC-AUC was preserved (0.938 to 0.941). On an independent clinical cohort acquired with a different device at a different institution (472 images, 22 malignant), performance was maintained (0.934 versus 0.930). Conclusions: Augmentations that model the physical causes of domain shift improve cross-source transfer at no cost to in-domain accuracy, and the improvement survives a selection-disjoint, contamination-free evaluation.
