AI Chest CT Analysis Distinguishes COVID-19 from Pneumonia

Researchers at Wuhan Huangpi People's Hospital (China), Shenzhen University Health Science Center (Shenzhen, China), and other institutions conducted a multicenter, retrospective study to develop the COVID-19 detection neural network (COVNet), a fully automatic AI framework that extracts visual features from volumetric chest CT exams for the detection of COVID-19, and differentiates them from CAP and other lung findings. Datasets were collected from six hospitals and diagnostic performance was assessed and evaluated.

In all, the collected dataset consisted of 4,356 volumetric chest CT exams from 3,322 patients (average age 49 years, 55% male). Overall per-exam sensitivity and specificity for detecting COVID-19 in the independent test set was 90% and 96%, respectively. The researchers also showed that the per-exam sensitivity and specificity for detecting CAP was 87% and 92%, respectively. The study was published on March 19, 2020, in Radiology.

“We were able to collect a large number of CT exams from multiple hospitals, which included 1,296 COVID-19 CT exams,” concluded lead author Lin Li, MD, of the Wuhan Huangpi People's Hospital department of radiology. “More importantly, 1,735 CAP and 1,325 non-pneumonia CT exams were also collected as the control groups in this study in order to ensure the detection robustness, considering that certain similar imaging features may be observed in COVID-19 and other types of lung diseases.”

The coronavirus Disease 2019 (COVID-19) outbreak has rapidly spread all over the world. On January 30, 2020, it was declared as a public health emergency of international concern by the World Health Organization (WHO). It is typically confirmed by reverse-transcription polymerase chain reaction (PCR), but CT can detect certain characteristic manifestations in the lung associated with COVID-19, such as bilateral ground-glass, consolidative pulmonary opacities on CT, and a crazy-paving pattern, among others.

Related Links:
Shenzhen University Health Science Center