[论文解读] Deep Convolutional Neural Networks to Diagnose COVID-19 and other Pneumonia Diseases from Posteroanterior Chest X-Rays

Pierre G. B. Moutounet-Cartan|arXiv (Cornell University)|May 2, 2020

COVID-19 diagnosis using AI参考文献 13被引用 31

一句话总结

本文评估多种CNN架构（VGG16/19、InceptionResNetV2、InceptionV3、Xception），在PA胸片上训练，用于将COVID-19、No Finding和Other Pneumonia进行分类，其中VGG16表现最好。

ABSTRACT

The article explores different deep convolutional neural network architectures trained and tested on posteroanterior chest X-rays of 327 patients who are healthy (152 patients), diagnosed with COVID-19 (125), and other types of pneumonia (48). In particular, this paper looks at the deep convolutional neural networks VGG16 and VGG19, InceptionResNetV2 and InceptionV3, as well as Xception, all followed by a flat multi-layer perceptron and a final 30% drop-out. The paper has found that the best performing network is VGG16 with a final $30$% drop-out trained over 3 classes (COVID-19, No Finding, Other Pneumonia). It has an internal cross-validated accuracy of $93.9(\pm3.4)$%, a COVID-19 sensitivity of $87.7(-1.9,+2)$%, and a No Finding sensitivity of $96.8(\pm0.8)$%. The respective external cross-validated values are $84.1(\pm13.5)$%, $87.7(-1.9,2)$%, and $96.8(\pm0.8)$%. The model optimizer was Adam with a 1e-4 learning rate, and categorical cross-entropy loss. It is hoped that, once this research will be put to practice in hospitals, healthcare professionals will be able in the medium to long-term to diagnosing through machine learning tools possible pneumonia, and if detected, whether it is linked to a COVID-19 infection, allowing the detection of new possible COVID-19 foyers after the end of possible "stop-and-go" lockdowns as expected by until a vaccine is found and widespread. Furthermore, in the short-term, it is hoped practitioners can compare the diagnosis from the deep convolutional neural networks with possible RT-PCR testing results, and if clashing, a Computed Tomography could be performed as they are more accurate in showing COVID-19 pneumonia.

研究动机与目标

• 评估深度CNN在PA胸片上诊断COVID-19及其他肺炎的可行性。
• 在三分类任务中比较多种架构（VGG16、VGG19、InceptionResNetV2、InceptionV3、Xception）的性能。
• 提供交叉验证性能指标并讨论在实际医院中的使用与局限性。

提出的方法

• 通过使用ImageNet预训练权重初始化网络来进行迁移学习。
• 使用Adam优化器、学习率0.0001，将所有模型训练200轮。
• 应用最终30% dropout和分类交叉熵损失。
• 使用分层的5折交叉验证来报告内部和外部测试性能。
• 通过将PA胸片调整为182x182并应用数据增强（旋转、平移、缩放、翻转、通道位移）来预处理。

实验结果

研究问题

• RQ1 基于PA胸片的CNN能否将COVID-19与No Finding及其他肺炎区分开来？
• RQ2 在这一三分类任务中，哪种CNN架构能取得最佳的交叉验证性能？
• RQ3 各架构在检测COVID-19和No Finding时的灵敏度是多少？
• RQ4 外部测试性能与内部交叉验证的对比如何？
• RQ5 在医院环境中使用PA X光和这些模型有哪些局限性？

主要发现

• VGG16在内部和外部数据集中均取得了被评估模型中的最佳整体性能。
• VGG16的内部5折交叉验证准确率：93.9%（±3.4）；COVID-19召回率：87.7%（±2.0）；No Finding召回率：96.8%（±0.8）。
• VGG16的外部测试准确率：84.1%（±13.5）；COVID-19召回率：87.7%（±2.0）；No Finding召回率：96.8%（±0.8）。
• InceptionResNetV2和InceptionV3显示中等性能，COVID-19灵敏度较低且方差较大；InceptionResNetV2的COVID-19召回率在约70.8–71.0%（内部/外部）。
• VGG19的性能与VGG16相似，但置信度提升略小且成本更高。
• 在该数据集中，Xception的表现不及基于VGG的模型。