Artificial intelligence (AI) algorithms applied to chest computed tomography (CT) images and clinical history can quickly and accurately identify patients with COVID-19, according to a study at the Icahn School of Medicine at Mount Sinai. This system achieved an AUC (a metric of machine-learning accuracy) of 0.92 and demonstrated sensitivity equal to that of a senior thoracic radiologist.
Rapid and accurate testing for COVID-19 is urgently needed. The current method used – a SARS-CoV-2 virus–specific reverse-transcriptase polymerase chain reaction (RT-PCR) test – can take up to two days to complete, and repeat testing may be required to rule out the possibility of false-negative results. There is also a shortage of available RT-PCR test kits.
Chest CT is a valuable tool used in the evaluation of patients with suspected SARS-CoV-2 infection. However, CT imaging alone cannot rule out COVID-19 in certain cases of patients with other types of lung disease. In addition, some patients in the early stages of the disease may have apparently normal CT results.
Yang Yang and colleagues used AI algorithms to integrate chest CT results with clinical symptoms, exposure history and laboratory testing to rapidly diagnose COVID-19-positive patients. The researchers trained and tested the model on a dataset of CT scans and clinical information collected between 17 January 2020 and 3 March 2020 from 905 patients in 18 medical centres in 13 provinces of China. There were 488 male and 417 female patients, who were from 1-91 years of age. A total of 419 patients tested positive for SARS-CoV-2 by RT-PCR assay.
The authors evaluated their AI model on a test set of 279 cases, of the 905 samples, and compared its performance to that of two thoracic radiologists, a senior radiologist and a fellow. Of 145 COVID-19-negative cases in the test set, 113 were correctly classified by both the model and the senior radiologist. The authors also found that the AI system improved the detection of RT-PCR-positive patients with COVID-19 who presented with apparently normal CT scans, correctly identifying 17 of 25 patients as COVID-19 positive, whereas the radiologists classified all of these patients as COVID-19 negative.
The authors conclude that when CT scans and associated clinical history are available, the proposed AI system can help to rapidly diagnose patients with COVID-19.
Abstract
For diagnosis of coronavirus disease 2019 (COVID-19), a SARS-CoV-2 virus-specific reverse transcriptase polymerase chain reaction (RT–PCR) test is routinely used. However, this test can take up to 2 d to complete, serial testing may be required to rule out the possibility of false negative results and there is currently a shortage of RT–PCR test kits, underscoring the urgent need for alternative methods for rapid and accurate diagnosis of patients with COVID-19. Chest computed tomography (CT) is a valuable component in the evaluation of patients with suspected SARS-CoV-2 infection. Nevertheless, CT alone may have limited negative predictive value for ruling out SARS-CoV-2 infection, as some patients may have normal radiological findings at early stages of the disease. In this study, we used artificial intelligence (AI) algorithms to integrate chest CT findings with clinical symptoms, exposure history and laboratory testing to rapidly diagnose patients who are positive for COVID-19. Among a total of 905 patients tested by real-time RT–PCR assay and next-generation sequencing RT–PCR, 419 (46.3%) tested positive for SARS-CoV-2. In a test set of 279 patients, the AI system achieved an area under the curve of 0.92 and had equal sensitivity as compared to a senior thoracic radiologist. The AI system also improved the detection of patients who were positive for COVID-19 via RT–PCR who presented with normal CT scans, correctly identifying 17 of 25 (68%) patients, whereas radiologists classified all of these patients as COVID-19 negative. When CT scans and associated clinical history are available, the proposed AI system can help to rapidly diagnose COVID-19 patients.
Authors
Xueyan Mei, Hao-Chih Lee, Kai-yue Diao, Mingqian Huang, Bin Lin, Chenyu Liu, Zongyu Xie, Yixuan Ma, Philip M Robson, Michael Chung, Adam Bernheim, Venkatesh Mani, Claudia Calcagno, Kunwei Li, Shaolin Li, Hong Shan, Jian Lv, Tongtong Zhao, Junli Xia, Qihua Long, Sharon Steinberger, Adam Jacobi, Timothy Deyer, Marta Luksza, Fang Liu, Brent P Little, Zahi A Fayad, Yang Yang
[link url="https://www.nature.com/articles/s41591-020-0931-3"]Nature abstract[/link]