Robust breast cancer detection in mammography and digital breast tomosynthesis using an annotation-efficient deep learning approach.
Robust breast cancer detection in mammography and digital breast tomosynthesis using an annotation-efficient deep learning approach.
复制标题
DOI：
10.1038/s41591-020-01174-9
复制
发表时间：
2021-03
期刊：
影响因子：
82.9
通讯作者：
Gregory Sorensen A
中科院分区：
文献类型：
--
作者：
Lotter W;Diab AR;Haslam B;Kim JG;Grisot G;Wu E;Wu K;Onieva JO;Boyer Y;Boxerman JL;Wang M;Bandler M;Vijayaraghavan GR;Gregory Sorensen A
乳腺癌仍然是一个全球性的挑战，2018年造成60多万人死亡(参考文献)。)。为了更早地发现癌症，世界各地的卫生组织建议筛查乳房X光检查，据估计，它可以将乳腺癌死亡率降低20%-40%。尽管筛查乳房X光检查有明显的价值，但显著的假阳性和假阴性率以及专家阅读器可用性的不一致为提高质量和获取机会留下了机会。为了解决这些限制，最近人们对将深度学习应用于乳房X光检查很感兴趣，这些努力突出了两个关键困难：获得大量带注释的训练数据和确保在人群、获取设备和模式之间推广。在这里，我们提出了一种注释高效的深度学习方法，它(1)在乳房X光照片分类中实现了最先进的性能，(2)成功地扩展到数字乳房断层合成(DBT；‘3D乳房X光摄影’)，(3)在癌症患者的临床阴性乳房X光照片中检测癌症，(4)很好地适用于筛查率较低的人群，(5)比五名全职乳房成像专家中的五名表现更好，敏感度平均提高了14%。通过从DBT数据创建新的“最大可疑投影”(MSP)图像，我们逐步训练的多实例学习方法有效地训练了仅使用乳房水平标签的DBT检查，同时保持了基于本地化的可解释性。总而言之，我们的结果展示了软件的前景，该软件可以提高全球乳房X光检查的准确性和获得机会。
Breast cancer remains a global challenge, causing over 600,000 deaths in 2018 (ref. ). To achieve earlier cancer detection, health organizations worldwide recommend screening mammography, which is estimated to decrease breast cancer mortality by 20–40% (refs.,). Despite the clear value of screening mammography, significant false positive and false negative rates along with non-uniformities in expert reader availability leave opportunities for improving quality and access,. To address these limitations, there has been much recent interest in applying deep learning to mammography, , , , , , , , , , , –, and these efforts have highlighted two key difficulties: obtaining large amounts of annotated training data and ensuring generalization across populations, acquisition equipment and modalities. Here we present an annotation-efficient deep learning approach that (1) achieves state-of-the-art performance in mammogram classification, (2) successfully extends to digital breast tomosynthesis (DBT; ‘3D mammography’), (3) detects cancers in clinically negative prior mammograms of patients with cancer, (4) generalizes well to a population with low screening rates and (5) outperforms five out of five full-time breast-imaging specialists with an average increase in sensitivity of 14%. By creating new ‘maximum suspicion projection’ (MSP) images from DBT data, our progressively trained, multiple-instance learning approach effectively trains on DBT exams using only breast-level labels while maintaining localization-based interpretability. Altogether, our results demonstrate promise towards software that can improve the accuracy of and access to screening mammography worldwide.
登录
查看更多内容
影响因子：
158.5
作者：
Fenton, Joshua J.;Taplin, Stephen H.;Elmore, Joann G.
通讯作者：
Elmore, Joann G.
影响因子：
19.7
作者：
Ikeda, DM;Birdwell, RL;Sickles, EA
通讯作者：
Sickles, EA
影响因子：
19.7
作者：
Lehman, Constance D.;Arao, Robert F.;Miglioretti, Diana L.
通讯作者：
Miglioretti, Diana L.
影响因子：
3.7
作者：
Evans KK;Birdwell RL;Wolfe JM
通讯作者：
Wolfe JM
影响因子：
8.8
作者：
Hart, D;Shochat, E;Agur, Z
通讯作者：
Agur, Z