Extracting Possibly Representative COVID-19 Biomarkers from X-ray Images with Deep Learning Approach and Image Data Related to Pulmonary Diseases
Extracting Possibly Representative COVID-19 Biomarkers from X-ray Images with Deep Learning Approach and Image Data Related to Pulmonary Diseases
复制标题
DOI：
10.1007/s40846-020-00529-4
复制
发表时间：
2020-05-14
影响因子：
2
通讯作者：
Tzani, Mpesiana A.
中科院分区：
文献类型：
--
作者：
Apostolopoulos, Ioannis D.;Aznaouridis, Sokratis I.;Tzani, Mpesiana A.
目的随着COVID-19的传播增加，新的、自动的和可靠的准确检测方法对于减少医学专家对疫情的暴露至关重要。X线成像，虽然限于特定的可视化，可能有助于诊断。在这项研究中，考虑了从X射线图像中自动分类肺部疾病（包括最近出现的COVID-19）的问题。方法深度学习已被证明是一种从医学图像中提取大量高维特征的显着方法。具体来说，在本文中，国家的最先进的卷积神经网络称为移动的网，从头开始训练，调查的重要性提取的特征的分类任务。利用3905张X射线图像的大规模数据集（对应于6种疾病）来训练MobileNet v2，该方法已被证明在相关任务中取得了优异的结果。结果从头开始训练CNN的性能优于其他迁移学习技术，无论是在区分七个类别之间的X射线还是在区分Covid-19和非Covid-19之间。七类之间的分类准确率达到87.66%。此外，该方法在COVID-19检测中的准确度为99.18%，灵敏度为97.36%，特异性为99.42%。结论结果表明，从头开始训练CNN可以揭示与COVID-19疾病相关但不限于COVID-19疾病的重要生物标志物，而最高的分类准确率表明进一步检查X射线成像潜力。
Purpose While the spread of COVID-19 is increased, new, automatic, and reliable methods for accurate detection are essential to reduce the exposure of the medical experts to the outbreak. X-ray imaging, although limited to specific visualizations, may be helpful for the diagnosis. In this study, the problem of automatic classification of pulmonary diseases, including the recently emerged COVID-19, from X-ray images, is considered. Methods Deep Learning has proven to be a remarkable method to extract massive high-dimensional features from medical images. Specifically, in this paper, the state-of-the-art Convolutional Neural Network called Mobile Net is employed and trained from scratch to investigate the importance of the extracted features for the classification task. A large-scale dataset of 3905 X-ray images, corresponding to 6 diseases, is utilized for training MobileNet v2, which has been proven to achieve excellent results in related tasks. Results Training the CNNs from scratch outperforms the other transfer learning techniques, both in distinguishing the X-rays between the seven classes and between Covid-19 and non-Covid-19. A classification accuracy between the seven classes of 87.66% is achieved. Besides, this method achieves 99.18% accuracy, 97.36% Sensitivity, and 99.42% Specificity in the detection of COVID-19. Conclusion The results suggest that training CNNs from scratch may reveal vital biomarkers related but not limited to the COVID-19 disease, while the top classification accuracy suggests further examination of the X-ray imaging potential.