Deep-COVID: Predicting COVID-19 From Chest X-Ray Images Using Deep\n Transfer Learning

2020/04/20 by Shervin Minaee, Minaee, Shervin, Rahele Kafieh +7 · 3 citations
Medicine · #COVID-19 diagnosis using AI #Radiomics and Machine Learning in Medical Imaging #Lung Cancer Diagnosis and Treatment

paper · pdf · doi:10.48550/arxiv.2004.09363

Abstract

The COVID-19 pandemic is causing a major outbreak in more than 150 countries\naround the world, having a severe impact on the health and life of many people\nglobally. One of the crucial step in fighting COVID-19 is the ability to detect\nthe infected patients early enough, and put them under special care. Detecting\nthis disease from radiography and radiology images is perhaps one of the\nfastest ways to diagnose the patients. Some of the early studies showed\nspecific abnormalities in the chest radiograms of patients infected with\nCOVID-19. Inspired by earlier works, we study the application of deep learning\nmodels to detect COVID-19 patients from their chest radiography images. We\nfirst prepare a dataset of 5,000 Chest X-rays from the publicly available\ndatasets. Images exhibiting COVID-19 disease presence were identified by\nboard-certified radiologist. Transfer learning on a subset of 2,000 radiograms\nwas used to train four popular convolutional neural networks, including\nResNet18, ResNet50, SqueezeNet, and DenseNet-121, to identify COVID-19 disease\nin the analyzed chest X-ray images. We evaluated these models on the remaining\n3,000 images, and most of these networks achieved a sensitivity rate of 98%\n(\± 3%), while having a specificity rate of around 90%. Besides sensitivity\nand specificity rates, we also present the receiver operating characteristic\n(ROC) curve, precision-recall curve, average prediction, and confusion matrix\nof each model. We also used a technique to generate heatmaps of lung regions\npotentially infected by COVID-19 and show that the generated heatmaps contain\nmost of the infected areas annotated by our board certified radiologist. While\nthe achieved performance is very encouraging, further analysis is required on a\nlarger set of COVID-19 images, to have a more reliable estimation of accuracy\nrates. The dataset, model implementations (in PyTorch), and evaluations, are\nall made publicly available for research community at\nhttps://github.com/shervinmin/DeepCovid.git\n

Cited by

• MAPI-GNN: Multi-Activation Plane Interaction Graph Neural Network for Multimodal Medical Diagnosis
• Intuitions of Machine Learning Researchers about Transfer Learning for Medical Image Classification
• Application of discrete Ricci curvature in pruning randomly wired neural networks: A case study with chest x-ray classification of COVID-19

Related

• Empowering COVID-19 Detection: Optimizing Performance Through Fine-Tuned EfficientNet Deep Learning Architecture
• Evaluation of Contemporary Convolutional Neural Network Architectures for Detecting COVID-19 from Chest Radiographs
• A Novel Implementation of Machine Learning for the Efficient, Explainable Diagnosis of COVID-19 from Chest CT
• DeepCOVIDExplainer: Explainable COVID-19 Diagnosis Based on Chest X-ray Images
• COVID-Net: a tailored deep convolutional neural network design for detection of COVID-19 cases from chest X-ray images