COVID-Net: an Open Source Initiative for COVID-19 Detection and Risk Stratification using Chest X-rays
With the unprecedented number of COVID-19 infections on a global scale, chest X-rays has now become an important screening tool alongside viral testing in countries that are most affected by COVID-19, with key advantages being: i) speed, ii) widespread availability (including portable X-ray systems), ii) provides useful information for risk assessment, which one cannot obtain from viral testing. The goal of the COVID-Net initiative is to accelerate the open collaborative development of deep learning AI solutions for COVID-19 infection detection and risk stratification, with AI explanations to gain transparency into visual indicators of COVID-19. We have provided not only COVIDx, a large dataset of over 16,000 images across over 13,000 patient cases, but also open source reference models trained on this dataset so that the global community can built upon and improve. By leveraging deep learning AI with COVID-Net, we aim to help clinicians improve both sensitivity and specificity by better differentiate COVID-19 infections from other forms of viral infections, which is a current challenge faced given their similarities, as well as assist clinicians with additional knowledge about how COVID-Net detects COVID-19 infections through AI explanations. Furthermore, we aim to build deep learning AI for risk stratification to aid hospitals and clinical sites to help improve patient population management and individualized care based on risk level.