COVIDGR is a set of X-ray images to assist the diagnosis of COVID-19 built in close collaboration with radiology expert teams in Spain.

Thanks to the collaboration with radiology expert teams, it has been possible to build the COVIDGR database. The data set are composed of anonymized radiographs of patients from different hospitals, following a strict labelling protocol. This allows the construction of diagnostic models of COVID-19 previously hindered by the lack of public data sets carefully inspected by experts. X-rays are faster and more widespread tests than others such as CT scanners or RT-PCR test, so they can facilitate the diagnosis of COVID-19 to many people in a short time.


All images have been obtained considering only the front/rear view. In addition, they have been trimmed by centering them on the lungs using the U-Net segmentation model, applying a 2.5% margin around the lung ends. This minimizes the potential noise introduced by radiographic notations and the inclusion of other parts of the body.


The Positive images correspond to patients who have tested positive for COVID-19 according to the RT-PCR test with less than 24 hours difference between taking the X-ray and performing the test. In addition, positive class X-rays have been scored according to the severity of the case using the RALE scale. Those that do not present symptoms but are positive according to the RT-PCR test are scored as Normal. The rest are labeled Mild, Moderate or Severe.


The first version of the COVIDGR set contains 426 positive and 426 negative cases (852 total), collected in hospitals in Granada (Spain). The positive cases cover the entire spectrum of RALE severity: 76 images are normal positive PCR, 100 are mildly severe, 171 are moderate and 79 are severe cases. Available versions of COVIDGR are downloadable at the following repository.

Reference paper:

S. Tabik, A. Gómez-Ríos, J. L. Martín-Rodríguez, I. Sevillano-García, M. Rey-Arena, D. Charte, E. Guirado, J. L. Suárez, J. Luengo, M. A. Valero-González, P. García-Villanova, E. Olmedo-Sánchez and F. Herrera, “COVIDGR Dataset and COVID-SDNet Methodology for Predicting COVID-19 Based on Chest X-Ray Images,” in IEEE Journal of Biomedical and Health Informatics, vol. 24, no. 12, pp. 3595-3605, Dec. 2020, doi: 10.1109/JBHI.2020.3037127.

Publication date:

June 2020


Siham Tabik

Scroll Up