We build a dataset of lung ultrasound images and videos. The dataset is assembled from publicly available resources in the web as well as from publications.
- You can donate your lung ultrasound recordings directly on our website:
- Please help us to find more data! Open an
issue if you
identified a promising data source. Please check here in our Google sheet whether the data is already included. Useful contributions are:
- Publications with ultrasound images/videos
- check sheet image_data_publications in the goole sheet linked above to see whether a publication was already added
- Images/Videos that are available via Creative Commens license (e.g. CC BY-NC-SA) in the web or on YouTube.
- Possible sources are:
- Publications with ultrasound images/videos
-
Convex:
- 134 videos (45x COVID, 29x bacterial pneumonia, 57x healthy, 3x viral pneumonia)
- 40 images (18x COVID, 7x bacterial pneumonia, 15x healthy)
- 21 videos from the Butterfly dataset (19 COVID, 2 healthy, see below how to use the provided scripts to process the data)
-
Linear:
- 21 videos (5x COVID, 2x bacterial pneumonia, 10x healthy, 4x viral pneumonia)
- 6 images (4x COVID, 2x bacterial pneumonia)
-
Data with artifacts (30 images, 2 videos): For the sake of completion, we provide a folder with images and videos that we found but did not include in our dataset due to artifacts (arrows, bars) in the image or because of unclear labels. For more information and comments by medical experts, see our metadata sheet or metadata csv.
-
We are constantly updating the dataset with new data - any contributions are appreciated!
- 25.8.2020: Update - added 10 new videos from the Nurthumbrian hospital
- 12.8.2020: Update - added 14 new COVID videos from publications
- 28.7.2020: Update - added 4 new videos and 7 new images from the Northumbria hospital
- 11.7.2020: Database update - We added an up to date csv file with all metadata, and new videos that were contributed from the Northumbria Specialist Emergency Care Hospital (17 images and 4 videos of healthy patients)
- 22.6.2020: Database update - We added 46 new videos (18x COVID, 1x bacterial pneumonia, 27x healthy).
- 16.5.2020: The ICLUS project released ~60 videos from COVID-patients (register here).
Unfortunately, not all data used to train/evaluate the model is in this repo as we do not have the right to host/distribute the data from Butterfly and ICLUS.
However, we provide a script that automatically processes the data from Butterfly. To reproduce the experiments from the paper, please first complete the following steps:
- Visit the COVID-19 ultrasound_gallery of Butterfly, scroll to the bottom and download the videos (we accessed this source on 17.04.2020 for training our models. Please note that it is not under control whether Butterfly will keep this data online. Feel free to notify us if you observe any changes).
- Place the
.zip
folder into the this folder (data
) cd
into thedata
folder.- Run:
NOTE: This step requires that you installed the
sh parse_butterfly.sh
pocovidnet
package before (see the pocovidnet README how to do that).
All current images should now be in data/image_dataset
.
In the current state, a user could input any image, e.g. of a house, and still receive a classification result as covid / pneumonia / healthy. In order to prevent this, we decided to include a fourth class called "uninformative", where we add Imagenet images and neck-ultrasound data from the Kaggle Nerve Segmentation Challenge.
Download the data here from google drive. It contains a folder uniform_class_nerves and one uniform_class_imagenet.
The data can be immediately used for training, simply combine it in a folder uninformative next to the covid, pneumonia and regular folders.
If you want to add them to an existing cross validation split (e.g after executing cross_val_splitter), we also provide a script:
Run:
python ../pocovidnet/scripts/add_uninformative_class.py -i uniform_class_imagenet -u uniform_class_nerves -o cross_validation -s 5
This script will split the data in the uniform_class_nerves and uniform_class_imagenet folders and add them in a folder uninformative to each fold.
Most data here is available under Creative Commons License. The following modifcations to videos/images were done:
- Cropped to the center to remove measuring bars, text etc.
- Removal of artifcats on the sample (few cases only)
We are deeply thankful to the authors and contributors to our datset, in particular
- Dr Avinash Aujayeb (MBBS MRCP (Edin 2008) PgCert ClinEd FHEA), Pleural Medicine Lead and Consultant in Respiratory and Acute Medicine for the Trustee for Mesothelioma UK and Northumbria Specialist Emergency Care Hospital, who contributes regularly to our database with clinical data. We greatly appreciate their efforts for open-access data.
- The maintainers of https://thepocusatlas.com, https://radiopaedia.org/, https://grepmed.com and https://litfl.com/ultrasound-library/ (thanks for Dr. Rippey in particular who gave interesting advice)
- Charlotte Buhre for recording data herself just for our dataset.
- The contributers to Stemlyn's blog.
Also, we obtained videos and images from publications on ultrasound, and we appreciate very much that we could include data from the following publications in our database:
- 3 COVID-19 images from [@inchingolo2020diagnosis] - The Diagnosis of Pneumonia in a Pregnant Woman with COVID-19 Using Maternal Lung Ultrasound
- 8 COVID-19 images from [@huang2020preliminary] - A Preliminary Study on the Ultrasonic Manifestations of Peripulmonary Lesions of Non-Critical Novel Coronavirus Pneumonia (COVID-19)
- 4 videos (healthy and pneumonia) from [@irwin2016advances] - Advances in Point-of-Care Thoracic Ultrasound
- 2 videos (healthy and pneumonia) from [@bouhemad2007clinical] - Clinical review: Bedside lung ultrasound in critical care practice
- 3 COVID-19 images from [@lomoro2020covid] - COVID-19 pneumonia manifestations at the admission on chest ultrasound, radiographs, and CT: single-center study and comprehensive radiologic literature review
- 2 viral pneumonia videos (H1N1) from [@testa2012early] - Early recognition of the 2009 pandemic influenza A (H1N1) pneumonia by chest ultrasound
- 1 healthy and 7 COVID-19 videos from [@yassa2020lung] - Lung Ultrasound Can Influence the Clinical Treatment of Pregnant Women With COVID ‐19
- 2 healthy and 1 viral image from [@stadler2017lung] - Lung ultrasound for the diagnosis of community-acquired pneumonia in children
- 1 healthy and 2 regular images from [@reissig2014lung] - Lung Ultrasound in Community-Acquired Pneumonia and in Interstitial Lung Diseases
- 4 pneumonia images from [@claes2017performance] - Performance of chest ultrasound in pediatric pneumonia
- 4 videos (viral and bacterial pneumonia) from [@tsung2012prospective] - Prospective application of clinician-performed lung ultrasonography during the 2009 H1N1 influenza A pandemic: distinguishing viral from bacterial pneumonia
- 6 videos (COVID-19, healthy, bacterial pneumonia) from [@vieira2020role] - Role of point-of-care ultrasound during the COVID-19 pandemic: our recommendations in the management of dialytic patients
- 9 COVID-19 images from [@sofia2020thoracic] - Thoracic ultrasound and SARS-COVID-19: a pictorial essay
- 8 COVID-19 videos from [@denault2020proposed] - A proposed lung ultrasound and phenotypic algorithm for the care of COVID-19 patients with acute respiratory failure
- 5 COVID-19 videos from [@volpicelli2020sonographic] - Sonographic signs and patterns of COVID-19 pneumonia
- 1 COVID-19 videos from [@abramspoint] - Point-of-Care Ultrasound in the Evaluation of COVID-19
@article{abramspoint,
title={Point-of-Care Ultrasound in the Evaluation of COVID-19},
author={Abrams, Eric R and Rose, Gabriel and Fields, J Matthew and Esener, Dasia},
journal={The Journal of emergency medicine},
pages={S0736--4679}
}
@article{volpicelli2020sonographic,
title={Sonographic signs and patterns of COVID-19 pneumonia},
author={Volpicelli, Giovanni and Gargani, Luna},
journal={The Ultrasound Journal},
volume={12},
number={1},
pages={1--3},
year={2020},
publisher={SpringerOpen}
}
@article{denault2020proposed,
title={A proposed lung ultrasound and phenotypic algorithm for the care of COVID-19 patients with acute respiratory failure},
author={Denault, Andr{\'e} Y and Delisle, St{\'e}phane and Canty, David and Royse, Alistair and Royse, Colin and Serra, Ximena Cid and Gebhard, Caroline E and Couture, {\'E}tienne J and Girard, Martin and Cavayas, Yiorgos Alexandros and others},
journal={Canadian Journal of Anaesthesia},
pages={1},
year={2020},
publisher={Nature Publishing Group}
}
@article{inchingolo2020diagnosis,
title={The Diagnosis of Pneumonia in a Pregnant Woman with COVID-19 Using Maternal Lung Ultrasound},
author={Inchingolo, Riccardo and Smargiassi, Andrea and Moro, Francesca and Buonsenso, Danilo and Salvi, Silvia and Del Giacomo, Paola and Scoppettuolo, Giancarlo and Demi, Libertario and Soldati, Gino and Testa, Antonia Carla},
journal={American Journal of Obstetrics and Gynecology},
year={2020},
publisher={Elsevier}
}
@article{huang2020preliminary,
title={A preliminary study on the ultrasonic manifestations of peripulmonary lesions of non-critical novel coronavirus pneumonia (COVID-19)},
author={Huang, Yi and Wang, Sihan and Liu, Yue and Zhang, Yaohui and Zheng, Chuyun and Zheng, Yu and Zhang, Chaoyang and Min, Weili and Zhou, Huihui and Yu, Ming and others},
journal={Available at SSRN 3544750},
year={2020}
}
@article{irwin2016advances,
title={Advances in point-of-care thoracic ultrasound},
author={Irwin, Zareth and Cook, Justin O},
journal={Emerg Med Clin North Am},
volume={34},
number={1},
pages={151--7},
year={2016}
}
@article{bouhemad2007clinical,
title={Clinical review: bedside lung ultrasound in critical care practice},
author={Bouhemad, B{\'e}la{\"\i}d and Zhang, Mao and Lu, Qin and Rouby, Jean-Jacques},
journal={Critical care},
volume={11},
number={1},
pages={205},
year={2007},
publisher={Springer}
}
@article{lomoro2020covid,
title={COVID-19 pneumonia manifestations at the admission on chest ultrasound, radiographs, and CT: single-center study and comprehensive radiologic literature review},
author={Lomoro, Pascal and Verde, Francesco and Zerboni, Filippo and Simonetti, Igino and Borghi, Claudia and Fachinetti, Camilla and Natalizi, Anna and Martegani, Alberto},
journal={European journal of radiology open},
pages={100231},
year={2020},
publisher={Elsevier}
}
@article{testa2012early,
title={Early recognition of the 2009 pandemic influenza A (H1N1) pneumonia by chest ultrasound},
author={Testa, Americo and Soldati, Gino and Copetti, Roberto and Giannuzzi, Rosangela and Portale, Grazia and Gentiloni-Silveri, Nicol{\`o}},
journal={Critical Care},
volume={16},
number={1},
pages={R30},
year={2012},
publisher={Springer}
}
@article{yassa2020lung,
title={Lung Ultrasound Can Influence the Clinical Treatment of Pregnant Women With COVID-19},
author={Yassa, Murat and Birol, Pinar and Mutlu, Ali Memis and Tekin, Arzu Bilge and Sandal, Kemal and Tug, Niyazi},
journal={Journal of Ultrasound in Medicine},
year={2020},
publisher={Wiley Online Library}
}
@article{stadler2017lung,
title={Lung ultrasound for the diagnosis of community-acquired pneumonia in children},
author={Stadler, Jacob AM and Andronikou, Savvas and Zar, Heather J},
journal={Pediatric radiology},
volume={47},
number={11},
pages={1412--1419},
year={2017},
publisher={Springer}
}
@article{reissig2014lung,
title={Lung ultrasound in community-acquired pneumonia and in interstitial lung diseases},
author={Reissig, Angelika and Copetti, Roberto},
journal={Respiration},
volume={87},
number={3},
pages={179--189},
year={2014},
publisher={Karger Publishers}
}
@article{claes2017performance,
title={Performance of chest ultrasound in pediatric pneumonia},
author={Claes, Anne-Sophie and Clapuyt, Philippe and Menten, Renaud and Michoux, Nicolas and Dumitriu, Dana},
journal={European journal of radiology},
volume={88},
pages={82--87},
year={2017},
publisher={Elsevier}
}
@article{tsung2012prospective,
title={Prospective application of clinician-performed lung ultrasonography during the 2009 H1N1 influenza A pandemic: distinguishing viral from bacterial pneumonia},
author={Tsung, James W and Kessler, David O and Shah, Vaishali P},
journal={Critical ultrasound journal},
volume={4},
number={1},
pages={1--10},
year={2012},
publisher={SpringerOpen}
}
@article{vieira2020role,
title={Role of point-of-care ultrasound during the COVID-19 pandemic: our recommendations in the management of dialytic patients},
author={Vieira, Ana Luisa Silveira and J{\'u}nior, Jos{\'e} Muniz Pazeli and Bastos, Marcus Gomes},
journal={The Ultrasound Journal},
volume={12},
number={1},
pages={1--9},
year={2020},
publisher={SpringerOpen}
}
@article{sofia2020thoracic,
title={Thoracic ultrasound and SARS-COVID-19: a pictorial essay},
author={Sofia, Soccorsa and Boccatonda, Andrea and Montanari, Marco and Spampinato, Michele and D’ardes, Damiano and Cocco, Giulio and Accogli, Esterita and Cipollone, Francesco and Schiavone, Cosima},
journal={Journal of ultrasound},
pages={1--5},
year={2020},
publisher={Springer}
}
The paper is available here.
If you build upon our work or find it useful, please cite our paper:
@article{born2020pocovid,
title={POCOVID-Net: Automatic Detection of COVID-19 From a New Lung Ultrasound Imaging Dataset (POCUS)},
author={Born, Jannis and Br{\"a}ndle, Gabriel and Cossio, Manuel and Disdier, Marion and Goulet, Julie and Roulin, J{\'e}r{\'e}mie and Wiedemann, Nina},
journal={arXiv preprint arXiv:2004.12084},
year={2020}
}