COVID-19 Research Results Published in Radiology

March 26, 2020

Image Source: NIAID-RML.

On March 20, the research results of Keya Medical’s COVID-19 AI Assisted Diagnosis algorithm were published in the paper, Artificial Intelligence Distinguishes COVID-19 from Community Acquired Pneumonia on Chest CT, in Radiological Society of North America’s peer-reviewed journal, Radiology*. Keya Medical used deep learning to develop a three-dimensional (3D) detection neural network for coronavirus disease 2019 (COVID-19) in pulmonary CT images with an area under the receiver operating characteristic curve (AUC) of 0.96.

Early diagnosis of COVID-19 is important for the treatment of patients and their subsequent isolation to prevent the spread of the disease. To improve the speed and efficiency of detection and diagnosis, we developed a deep learning model that can detect COVID-19 on chest CT exams, referred to as COVNet.

This retrospective and multi-center study included 4,536 three-dimensional (3D) volumetric chest CT exams from 3,506 patients acquired in six medical centers between Aug. 16, 2016 and Feb. 17, 2020. The dataset included 1,296 COVID-19 exams, 1,735 community-acquired pneumonia, and 1,325 non-pneumonia cases. All COVID-19 cases were confirmed as positive via nucleic acid testing with reverse transcription polymerase chain reaction (RT-PCR) testing and were acquired from Dec. 31, 2019 to Feb. 17, 2020.

After training the deep learning model, the average processing time for a new CT exam was 4.51 seconds on a workstation. On an independent testing dataset, COVNet achieved high sensitivity and high specificity in detecting COVID-19 in chest CT images. The AUC values for COVID-19 and community acquired pneumonia were 0.96 and 0.95 respectively.

There is overlap in the chest CT imaging findings of all viral pneumonias with other chest diseases that encourages a multidisciplinary approach to the final diagnosis used for patient treatment. As a next step, it would be important to predict the severity degree to further help monitor and treat patients.

At Keya Medical, we are committed to applying our AI expertise to help respond to this global health crisis. We are actively recruiting public health agencies, non-profit organizations, and other AI labs around the world to accelerate progress.

*Radiology is a monthly, peer-reviewed medical journal owned and published by the Radiological Society of North America. Radiology features the most current, high-quality and clinically relevant research in the field of radiology every month.