Automated abnormality classification of chest radiographs using deep convolutional neural networks
Automated abnormality classification of chest radiographs using deep convolutional neural networks
复制标题
DOI：
10.1038/s41746-020-0273-z
复制
发表时间：
2020-05-14
影响因子：
15.2
通讯作者：
Summers, Ronald M.
中科院分区：
文献类型：
--
作者：
Tang, Yu-Xing;Tang, You-Bao;Summers, Ronald M.
作为医疗实践中最普遍的诊断影像检查之一，胸部X光检查需要及时报告图像中的潜在发现和疾病诊断。基于胸部 X 线摄影的疾病自动、快速、可靠的检测是放射学工作流程中的关键步骤。在这项工作中，我们开发并评估了各种深度卷积神经网络 (CNN)，用于区分正常和异常的胸部 X 光片，以帮助提醒放射科医生和临床医生潜在的异常发现，作为工作列表分类和报告优先级的一种手段。基于 CNN 的模型对正常与异常胸片分类的 AUC 为 0.9824 +/- 0.0043（准确度为 94.64 +/- 0.45%，敏感性为 96.50 +/- 0.36%，特异性为 92.86 +/- 0.48%）。 CNN 模型获得的正常与肺不透明分类的 AUC 为 0.9804 +/- 0.0032（准确度为 94.71 +/- 0.32%，敏感性为 92.20 +/- 0.34%，特异性为 96.34 +/- 0.31%）。外部数据集的分类性能表明 CNN 模型可能具有高度泛化性，AUC 为 0.9444 +/- 0.0029。 CNN 模型在成人患者队列上进行预训练，并在儿童患者上进行微调，正常与肺炎分类的 AUC 为 0.9851 +/- 0.0046。使用自然图像进行预训练证明了对于包含约 8500 张图像的中等大小训练图像集的好处。本研究中观察到的诊断准确性的显着表现表明深度 CNN 可以准确有效地区分正常和异常的胸片，从而为放射学工作流程和患者护理提供潜在的好处。
As one of the most ubiquitous diagnostic imaging tests in medical practice, chest radiography requires timely reporting of potential findings and diagnosis of diseases in the images. Automated, fast, and reliable detection of diseases based on chest radiography is a critical step in radiology workflow. In this work, we developed and evaluated various deep convolutional neural networks (CNN) for differentiating between normal and abnormal frontal chest radiographs, in order to help alert radiologists and clinicians of potential abnormal findings as a means of work list triaging and reporting prioritization. A CNN-based model achieved an AUC of 0.9824 +/- 0.0043 (with an accuracy of 94.64 +/- 0.45%, a sensitivity of 96.50 +/- 0.36% and a specificity of 92.86 +/- 0.48%) for normal versus abnormal chest radiograph classification. The CNN model obtained an AUC of 0.9804 +/- 0.0032 (with an accuracy of 94.71 +/- 0.32%, a sensitivity of 92.20 +/- 0.34% and a specificity of 96.34 +/- 0.31%) for normal versus lung opacity classification. Classification performance on the external dataset showed that the CNN model is likely to be highly generalizable, with an AUC of 0.9444 +/- 0.0029. The CNN model pre-trained on cohorts of adult patients and fine-tuned on pediatric patients achieved an AUC of 0.9851 +/- 0.0046 for normal versus pneumonia classification. Pretraining with natural images demonstrates benefit for a moderate-sized training image set of about 8500 images. The remarkable performance in diagnostic accuracy observed in this study shows that deep CNNs can accurately and effectively differentiate normal and abnormal chest radiographs, thereby providing potential benefits to radiology workflow and patient care.