从胸部X线片中萃取并学习细粒度标签
摘要
胸部X线片是当今急诊室与重症监护病房中最常见的诊断检查。近来，若干研究者着手构建大规模胸部X射线数据集，以开发用于识别少数粗粒度征象类别（如 opacity、mass 与 nodule）的深度学习模型。本文中，我们聚焦于从胸部X射线图像中提取并学习细粒度标签。具体而言，我们开发了一种新方法：将词汇驱动的概念抽取与依存句法分析树中的短语分组相结合，以实现修饰语与征象的关联，从而从放射学报告中提取细粒度标签。我们筛选出共计 457 个细粒度标签，描绘了迄今最大范围的征象谱，并获取了规模充足的数据集以训练一个专为细粒度分类设计的新深度学习模型。我们展示的结果表明，标签提取过程高度准确，且细粒度标签的学习可靠。据我们所知，所得网络是首个能够识别图像中征象细粒度描述（覆盖超过九类修饰语，包括侧别、部位、严重程度、大小与形态）的网络。
引用
@article{arxiv.2011.09517,
title = {Extracting and Learning Fine-Grained Labels from Chest Radiographs},
author = {Tanveer Syeda-Mahmood and Ph. D and K. C. L Wong and Ph. D and Joy T. Wu and M. D. and M. P. H and Ashutosh Jadhav and Ph. D and Orest Boyko and M. D. Ph. D},
journal= {arXiv preprint arXiv:2011.09517},
year = {2020}
}
备注
This paper won the Homer R. Warner Award at AMIA 2020 awarded to a paper that best describes approaches to improving computerized information acquisition, knowledge data acquisition and management, and experimental results documenting the value of these approaches. The paper shows a combination of textual and visual processing to automatically recognize complex findings in chest X-rays