ISRA May 2022

COVID-19 Classification of X-Ray Images Using Deep Neural Networks

Daphna Keidar 1 Daniel Yaron 2 Elisha Goldstein 3 Yair Shachar 4 Ayelet Blass 2 Leonid Charbinsky 5 Israel Aharony 5 Liza Lifshitz 5 Dimitri Lumelsky 5 Ziv Neeman 5 Matti Mizrachi 6,7 Majd Hajouj 6,7 Nethanel Eizenbach 6,7 Eyal Sela 6,7 Chedva S. Weiss 8 Philip Levin 8 Ofer Benjaminov 8 Gil N. Bachar 9,10 Shlomit Tamir 9,10 Yael Rapson 9,10 Dror Suhami 9,10 Eli Atar 9,10 Amiel A. Dror 6,7 Naama R. Bogot 8 Ahuva Grubstein 9,10 Nogah Shabshin 5 Yishai M. Elyada 11 Yonina C. Eldar 2
1Department of Computer Science, ETH Zrich, Switzerland
2Department of Math and Computer Science, Weizmann Institute of Science, Israel
3Bioinformatics Unit, Life Sciences Core Facilities, Weizmann Institute of Science, Israel
4Eyeway Vision, Ltd, Israel
5Department of Radiology, HaEmek Medical Center, Israel
6Department of Otolaryngology, Head and Neck Surgery, Galilee Medical Center, Israel
7The Azrieli Faculty of Medicine, Bar-Ilan University, Israel
8Cardiothoracic Imaging Unit, Shaare Zedek Medical Center, Israel
9Radiology Department, Rabin Medical Center, Israel
10Sakler School of Medicine, Tel-Aviv University, Israel
11Mobileye Vision Technologies, Ltd, Israel

Purpose: In the midst of the coronavirus disease 2019 (COVID-19) outbreak, chest X-ray (CXR) imaging is playing an important role in diagnosis and monitoring of patients with COVID-19. We propose a deep learning model for detection of COVID-19 from CXRs, as well as a tool for retrieving similar patients according to the model’s results on their CXRs. For training and evaluating our model, we collected CXRs from inpatients hospitalized in four different hospitals.

Methods: In this retrospective study, 1384 frontal CXRs, of COVID-19 confirmed patients imaged between March and August 2020, and 1024 matching CXRs of non-COVID patients imaged before the pandemic, were collected and used to build a deep learning classifier for detecting patients positive for COVID-19. The classifier consists of an ensemble of pre-trained deep neural networks (DNNS), specifically, ReNet34, ReNet50¸ ReNet152, and vgg16, and is enhanced by data augmentation and lung segmentation. We further implemented a nearest-neighbors algorithm that uses DNN-based image embeddings to retrieve the images most similar to a given image.

Results: Our model achieved accuracy of 90.3%, (95% CI: 86.3–93.7%) specificity of 90% (95% CI: 84.3–94%), and sensitivity of 90.5% (95% CI: 85–94%) on a test dataset comprising 15% (350/2326) of the original images. The AUC of the ROC curve is 0.96 (95% CI: 0.93–0.97).

Conclusions: We provide deep learning models, trained and evaluated on CXRs that can assist medical efforts and reduce medical staff workload in handling COVID-19.