Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reliable COVID-19 Detection Using Chest X-ray Images

Aysen Degerli, Mete Ahishali, Serkan Kiranyaz, Muhammad E. H. Chowdhury, Moncef Gabbouj

TL;DR

This work tackles the challenge of reliable COVID-19 detection from chest X-rays under data scarcity by introducing ReCovNet, an end-to-end network that leverages a lung-segmentation encoder pretraining to ensure decisions focus on pulmonary regions. It also presents QaTa-COV19, the largest public CXR COVID-19 benchmark with 124{,}616 images (4{,}603 COVID-19 cases) spanning 14 thoracic diseases plus healthy controls, including a difficult early-COVID subset. The study demonstrates that ReCovNet-v2, built on a ResNet-50 encoder, achieves the highest sensitivity (98.57%) and specificity (99.77%) on unseen data, outperforming four state-of-the-art baselines, and provides Grad-CAM evidence that the model relies on lung areas rather than irrelevant regions. The dataset is publicly shared to foster robust evaluation, and the results highlight the potential of segmentation-informed transfer learning to enhance the reliability of AI-based COVID-19 detection in real-world settings.

Abstract

Coronavirus disease 2019 (COVID-19) has emerged the need for computer-aided diagnosis with automatic, accurate, and fast algorithms. Recent studies have applied Machine Learning algorithms for COVID-19 diagnosis over chest X-ray (CXR) images. However, the data scarcity in these studies prevents a reliable evaluation with the potential of overfitting and limits the performance of deep networks. Moreover, these networks can discriminate COVID-19 pneumonia usually from healthy subjects only or occasionally, from limited pneumonia types. Thus, there is a need for a robust and accurate COVID-19 detector evaluated over a large CXR dataset. To address this need, in this study, we propose a reliable COVID-19 detection network: ReCovNet, which can discriminate COVID-19 pneumonia from 14 different thoracic diseases and healthy subjects. To accomplish this, we have compiled the largest COVID-19 CXR dataset: QaTa-COV19 with 124,616 images including 4603 COVID-19 samples. The proposed ReCovNet achieved a detection performance with 98.57% sensitivity and 99.77% specificity.

Reliable COVID-19 Detection Using Chest X-ray Images

TL;DR

This work tackles the challenge of reliable COVID-19 detection from chest X-rays under data scarcity by introducing ReCovNet, an end-to-end network that leverages a lung-segmentation encoder pretraining to ensure decisions focus on pulmonary regions. It also presents QaTa-COV19, the largest public CXR COVID-19 benchmark with 124{,}616 images (4{,}603 COVID-19 cases) spanning 14 thoracic diseases plus healthy controls, including a difficult early-COVID subset. The study demonstrates that ReCovNet-v2, built on a ResNet-50 encoder, achieves the highest sensitivity (98.57%) and specificity (99.77%) on unseen data, outperforming four state-of-the-art baselines, and provides Grad-CAM evidence that the model relies on lung areas rather than irrelevant regions. The dataset is publicly shared to foster robust evaluation, and the results highlight the potential of segmentation-informed transfer learning to enhance the reliability of AI-based COVID-19 detection in real-world settings.

Abstract

Coronavirus disease 2019 (COVID-19) has emerged the need for computer-aided diagnosis with automatic, accurate, and fast algorithms. Recent studies have applied Machine Learning algorithms for COVID-19 diagnosis over chest X-ray (CXR) images. However, the data scarcity in these studies prevents a reliable evaluation with the potential of overfitting and limits the performance of deep networks. Moreover, these networks can discriminate COVID-19 pneumonia usually from healthy subjects only or occasionally, from limited pneumonia types. Thus, there is a need for a robust and accurate COVID-19 detector evaluated over a large CXR dataset. To address this need, in this study, we propose a reliable COVID-19 detection network: ReCovNet, which can discriminate COVID-19 pneumonia from 14 different thoracic diseases and healthy subjects. To accomplish this, we have compiled the largest COVID-19 CXR dataset: QaTa-COV19 with 124,616 images including 4603 COVID-19 samples. The proposed ReCovNet achieved a detection performance with 98.57% sensitivity and 99.77% specificity.

Paper Structure

This paper contains 9 sections, 1 equation, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Materials and Methodology
  3. The Benchmark QaTa-COV19 Dataset
  4. COVID-19 Detection with Deep Models
  5. ReCovNet: Reliable COVID-19 Detection Network
  6. Experimental Evaluation
  7. Experimental Setup
  8. Experimental Results
  9. Conclusions

Figures (2)

  • Figure 1: The proposed ReCovNet, where the transfer learning is performed from the segmentation network initially trained on CXRs for lung segmentation.
  • Figure 2: The activation maps are extracted by Grad-CAM selvaraju2017grad approach for the models. The top two rows are COVID-19 samples, whereas the bottom row is a CXR from the control group images.