Deep Learning-Based Prediction of nCOVID-19 Disease Using Chest X-ray Images (CXRIs)
Deep Learning-Based Prediction of nCOVID-19 Disease Using Chest X-ray Images (CXRIs)
复制标题
使用胸部 X 射线图像 (CXRI) 基于深度学习的 nCOVID-19 疾病预测
DOI：
10.1007/978-981-16-5411-4_3
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
P. Krishna
中科院分区：
文献类型：
--
作者：
M. Muqeet;Mohammed Umair Quadri;K. Sasidhar;P. Krishna
新冠肺炎已演变成一场大流行，影响了日常生活、健康和全球经济。尽早发现感染患者，避免新冠病毒进一步传播，并迅速对患者进行治疗至关重要。最近的研究表明，这些CXRIs包含有关新冠病毒的重要细节。将深度学习应用于这些CXRIs，可以支持与常规的新冠病毒RT-PCR检测一起精确检测该疾病。在本章中，我们研究了深度学习（DL）模型的应用，通过考虑CXRIs，从正常患者中检测nCOVID-19患者。我们首先从现有的公开数据库中准备了1800个cxri的数据集。在80%的数据集上进行迁移学习，用于训练三种流行的卷积神经网络（cnn），包括VGG16、VGG19和ResNet50，以对CXRIs中的nCOVID-19感染患者进行分类和预测。这些模型在CXRIs上进行了评估，大多数cnn都获得了良好的敏感性和特异性值。我们还展示了所选历元数的精度和损失值曲线。所提出的工作在谷歌实验室进行了实验。
The nCOVID-19 turned into a pandemic and has affected routine lives, health, and the global economy. It is crucial to identify the infectious patients as early as possible to avoid further spread of the nCOVID-19 and to rapidly treat the affected patients. Recent studies have suggested that such CXRIs contain salient details about the nCOVID-19. Application of deep learning to such CXRIs can be supportive for the precise detection of this disease along with the regular RT-PCR test for nCOVID-19. In this chapter, we examined the application of deep learning (DL) models to detect nCOVID-19 patients from normal patients via considering the CXRIs. We first prepared a dataset of 1800 CXRIs from the publicly existing database. Transfer learning on 80% of the dataset was applied to train three popular convolutional neural networks (CNNs), including VGG16, VGG19, and ResNet50, to classify and predict nCOVID-19 infected patients from the CXRIs. These models are evaluated on the CXRIs, and most of these CNNs achieved good sensitivity and specificity values. We also exhibited the accuracy and loss value curves for the selected number of epochs. The experimentations of the proposed work are carried out in Google Colab.