MARVEL: 面向长尾分布外检测的边际感知鲁棒 von Mises-Fischer 专家学习
MARVEL: Margin-Aware Robust von Mises-Fischer Expert Learning for Long-Tailed Out-of-Distribution Detection
AI总结：
提出 MARVEL 框架，通过非线性 von Mises-Fisher 分类器、边际感知多专家学习和离群专家，解决长尾医学图像分布外检测问题，在三个数据集上实现 FPR95 分别降低 8.45%、13.02% 和 36.90%。
摘要：
对于临床部署，自动诊断系统在面对前所未见的病例时保持可靠性至关重要，然而深度模型通常以高置信度错误分类分布外（OOD）输入，这凸显了更鲁棒的 OOD 检测方法的必要性。尽管已有大量工作致力于提高模型鲁棒性，但现有文献大多假设数据集平衡，在粗糙或非临床的 OOD 源上评估 OOD 检测，或缺乏对不同 OOD 场景的全面评估。为填补这些空白，我们提出了一种新颖的方法，该方法在多样且不平衡的医学数据集上训练，并在临床相关的 OOD 谱上进行评估。我们的框架包含三个关键组件：（1）非线性 von Mises-Fisher（NvMF）分类器，能够学习非线性决策边界，并理论证明了其与余弦分类器的渐近联系；（2）多专家框架，其中边际感知的 NvMF 分类器专门处理标签分布的不同区域，以更好地应对不平衡；（3）离群专家，专门训练以区分内点和离群点，从而增强 OOD 检测。在 RFMiD、ISIC2019 和 NCTCRC 数据集上的评估表明，与最先进方法相比，我们的方法持续改进，平均 FPR95 分别降低了 8.45%、13.02% 和 36.90%。这些增益通过全面的消融实验得到进一步支持，验证了每个组件的贡献。这使得能够可靠地识别不熟悉的病例，以便转诊给临床医生，从而在实际工作流程中支持更安全的 AI 辅助诊断。我们的代码可在以下网址获取：https://this URL。
英文摘要：
For clinical deployment, it is essential that automated diagnostic systems remain reliable when confronted with previously unseen cases, yet deep models routinely misclassify out-of-distribution (OOD) inputs with high confidence, underscoring the need for more robust OOD detection methods. Although substantial effort has been devoted to improving model robustness, most of the existing literature assumes balanced datasets, evaluates OOD detection on coarse or non-clinical OOD sources, or lacks comprehensive assessment across diverse OOD scenarios. To address the gaps, we propose a novel methodology trained on diverse and imbalanced medical datasets and evaluated across a clinically reflective OOD spectrum. Our framework comprises three key components: (1) a Nonlinear von Mises-Fisher (NvMF) classifier capable of learning non-linear decision boundaries, with theoretical proof of its asymptotic connection to cosine classifiers; (2) a multi-expert framework in which margin-aware NvMF classifiers specialise in different regions of label distribution to better handle imbalance; and (3) an outlier expert trained explicitly to distinguish inlier from outlier data, thereby strengthening OOD detection. Evaluation on RFMiD, ISIC2019, and NCTCRC datasets demonstrates consistent improvements over state-of-the-art methods, achieving mean FPR95 reductions of 8.45%, 13.02%, and 36.90% respectively. These gains are further supported by comprehensive ablations that validated the contributions of each component. This enables reliable identification of unfamiliar cases for deferral to clinicians, supporting safer AI-assisted diagnosis in real-world workflows. Our code is available at https://github.com/redboxup/MARVEL.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
