The ongoing 2019 coronavirus disease (COVID-19) pandemic has seriously impacted global health and the economy. Computed tomography (CT) is the primary imaging technique for diagnosing lung infections in patients with COVID-19.
Data-driven and artificial intelligence solutions for automated processing of CT images rely primarily on large-scale, heterogeneous datasets.
Due to privacy and data availability concerns, it is difficult to obtain open and publicly available CT COVID-19 datasets, limiting the development of AI-enabled autodiagnostic solutions.
To address this problem, large computed tomography datasets covering various patterns of lung infections are in high demand.
Coronavirus Disease 2019 (COVID-19) is an infectious, highly contagious disease with serious global health implications. As of January 31, 2021, there were 103 million confirmed infections worldwide, claiming more than 2.2 million lives.
The main obstacle to managing and controlling COVID-19 is the availability of timely tests for screening and monitoring diseases.
Computed tomography (CT) is commonly used in clinical practice to diagnose, screen and treat COVID-19 around the world. The large number of scans required puts a heavy burden on radiographers and leaves them with limited time, which prevents timely delivery of CT scans.
In addition, in many underdeveloped rural areas, access to well-trained radiographers with sufficient experience in imaging COVID-19 is limited.
Collectively, they require data-driven artificial intelligence (AI) solutions to automatically detect and quantify COVID-19 infections.
To date, there have been many studies that have attempted to apply AI-based approaches, such as deep convolutional neural network (CNN) models, to automatically detect and quantify COVID-19 from CT images.
The key to the success of these models is the use of large datasets that encapsulate different patterns of lung infections. However, due to privacy and data collection concerns, the CT images used in these studies are of limited size and are not publicly available.
This will significantly impact the development of new AI-based solutions for more advanced diagnosis and quantification of COVID-19 infections.