Open source release of COVID-Net CXR-3, our third-generation deep neural network tailored for COVID-19 screening using chest x-ray (CXR) imaging, as part of the COVID-Net initiative!
Trained using over 19,000 CXR images from a multinational cohort of over 16,656 patient cases in 51 countries (largest, most diverse of its kind), COVID-Net CXR-3 achieves state-of-the-art accuracy, sensitivity, and PPV of 98.3%, 97.5%, and 99%, respectively. Furthermore, we demonstrated state-of-art accuracy on predicting the severity of a COVID-positive patient's condition to illustrate its multi-faceted utility across different clinical decision support tasks.
#computervision @kdnuggets @datasciencechatsPaper: https://arxiv.org/abs/2110.06063
Repo: https://github.com/lindawangg/COVID-Net/blob/master/docs/models.md
COVID-Net portal: http://www.covid-net.ml