Dataset for COVID-19 segmentation and severity scoring

In this dataset, we present a set of chest X-ray images of COVID-19 positive and negative patients, with the aim of performing segmentation and scoring of disease-affected areas in the lungs. The total number of images is 1,364. Of them, 580 are COVID-19 positive images (43%) and 784 images show no findings (57%). Each COVID-19 image contains the segmentation mask of the COVID-19-affected area and the associated severity score.

We collected and pre-processed four publicly available datasets used for the classification of COVID-19 and pneumonia, namely:
• Actualmed COVID-19 Chest X-ray Dataset containing 49 images of COVID-19 patients [1];
• COVID-19 Radiography Database containing 104 images of COVID-19 patients [2];
• COVID Chest X-Ray Dataset containing 399 images of COVID-19 patients [3];
• Figure1 COVID Chest X-ray Dataset containing 28 images of COVID-19 patients [4].

Then, two senior radiologists in our team, from the United States and Russia, annotated anteroposterior and posteroanterior radiographs. Each annotator has more than 10 years of experience and has been working extensively during the COVID-19 pandemic. The annotators labeled the datasets independently. Despite being time-consuming, this allowed us to get a consensus severity score of COVID-19-positive patients. Each patient has been labeled on a scale from 0 to 6, where 0 means no findings, while 6 is a severe form of COVID-19 which affects more than 85% of the lungs.

In addition to COVID-19 data, we also used two datasets with normal lungs without any pneumonia or any other abnormalities. Those healthy lung images have also been validated by the same two radiologists. Those images are represented by two datasets, namely:
• Chest X-ray Normal Dataset containing 431 images of patients with no findings [5];
• RSNA Normal Dataset containing 353 images of patients with no findings [6].

References:
1. Wang L, Wong A, Qiu ZL, McInnis P, Chung A, Gunraj H, et al. Actualmed COVID-19 Chest X-ray Dataset Initiative. 2020. Available: https://github.com/agchung/Actualmed-COVID-chestxray-dataset
2. COVID-19 Radiography Database. Available: https://www.kaggle.com/tawsifurrahman/covid19-radiography-database
3. COVID-19 Image Data Collection. 2020. Available: https://github.com/ieee8023/covid-chestxray-dataset
4. Wang L, Wong A, Qiu ZL, McInnis P, Chung A, Gunraj H, et al. Figure 1 COVID-19 Chest X-ray Dataset Initiative. 2020. Available: https://github.com/agchung/Figure1-COVID-chestxray-dataset
5. Chest X-Ray Images (Pneumonia). Available: https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia
6. RSNA Pneumonia Detection Challenge. Available: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge