研究人员开发了一种名为正常引导（Normal Guidance）的新型正则化技术，用于3D医学图像分类中基于注意力机制的多实例学习（MIL）。该方法鼓励学习到的注意力分布遵循钟形曲线，旨在提高弱监督设置下的切片级预测准确性。在三个数据集上对超过400万个2D切片进行的实验表明，正常引导显著增强了基于注意力机制和基于Transformer的MIL方法的切片级定位能力，同时在整个扫描分类中保持了有竞争力的性能。 AI
影响 通过改进定位能力，这项技术有望实现更准确的AI驱动的3D医学扫描诊断。
排序理由 该集群包含一篇详细介绍医学影像AI新技术的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.27306v1 Announce Type: new Abstract: We consider training classifiers for 3D medical images using only one binary label for the entire volume rather than a label for each 2D slice. In such weakly supervised settings, can we learn accurate classifiers for slice-level pr…
We consider training classifiers for 3D medical images using only one binary label for the entire volume rather than a label for each 2D slice. In such weakly supervised settings, can we learn accurate classifiers for slice-level predictions? Attention-based multiple instance lea…