tetano
Editor, Senior Moderator
Radiology
. 2020 Sep 24;202944.
doi: 10.1148/radiol.2020202944. Online ahead of print.
Diagnosis of COVID-19 Pneumonia Using Chest Radiography: Value of Artificial Intelligence
Ran Zhang[SUP] 1 [/SUP], Xin Tie[SUP] 1 [/SUP], Zhihua Qi[SUP] 1 [/SUP], Nicholas B Bevins[SUP] 1 [/SUP], Chengzhu Zhang[SUP] 1 [/SUP], Dalton Griner[SUP] 1 [/SUP], Thomas K Song[SUP] 1 [/SUP], Jeffrey D Nadig[SUP] 1 [/SUP], Mark L Schiebler[SUP] 1 [/SUP], John W Garrett[SUP] 1 [/SUP], Ke Li[SUP] 1 [/SUP], Scott B Reeder[SUP] 1 [/SUP], Guang-Hong Chen[SUP] 1 [/SUP]
Affiliations
Abstract
Background Radiologists are proficient in differentiating between chest x-ray radiographs (CXRs) with and without symptoms of pneumonia, but have found it more challenging to differentiate CXRs with COVID-19 pneumonia symptoms from those without. Purpose To develop an artificial intelligence algorithm to differentiate COVID-19 pneumonia from other causes of CXR abnormalities. Materials and Methods In this retrospective study, a deep neural network, CV19-Net, was trained, validated, and tested on CXRs from patients with and without COVID-19 pneumonia. For the COVID-19 positive CXRs, patients with reverse transcriptase polymerase chain reaction positive results for severe acute respiratory syndrome coronavirus 2 with positive pneumonia findings between February 1, 2020 and May 30, 2020 were included. For the non-COVID-19 CXRs, patients with pneumonia who underwent CXR between October 1, 2019 and December 31, 2019 were included. Area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were calculated to characterize diagnostic performance. To benchmark the performance of CV19-Net, a randomly sampled test dataset containing 500 CXRs from 500 patients was evaluated by both the CV19-Net and three experienced thoracic radiologists. Results A total of 2060 patients (5806 CXRs; mean age 62 ? 16, 1059 men) with COVID-19 pneumonia and 3148 patients (5300 CXRs; mean age 64 ? 18, 1578 men) with non-COVID-19 pneumonia were included and split into training + validation and test datasets. For the test set, CV19-Net achieved an AUC of 0.92 (95% confidence interval [CI]: 0.91, 0.93) corresponding to a sensitivity of 88% (95% CI: 87%, 89%) and a specificity of 79% (95% CI: 77%, 80%) using a high sensitivity operating threshold, or a sensitivity of 78% (95% CI: 77%, 79%) and a specificity of 89% (95% CI: 88%, 90%) using a high specificity operating threshold. For the 500 sampled CXRs, CV19-Net achieved an AUC of 0.94 (95% CI: 0.93, 0.96) compared to a 0.85 AUC (95% CI: 0.81, 0.88) of radiologists. Conclusion CV19-Net was able to differentiate COVID-19 related pneumonia from other types of pneumonia with performance exceeding that of experienced thoracic radiologists.
. 2020 Sep 24;202944.
doi: 10.1148/radiol.2020202944. Online ahead of print.
Diagnosis of COVID-19 Pneumonia Using Chest Radiography: Value of Artificial Intelligence
Ran Zhang[SUP] 1 [/SUP], Xin Tie[SUP] 1 [/SUP], Zhihua Qi[SUP] 1 [/SUP], Nicholas B Bevins[SUP] 1 [/SUP], Chengzhu Zhang[SUP] 1 [/SUP], Dalton Griner[SUP] 1 [/SUP], Thomas K Song[SUP] 1 [/SUP], Jeffrey D Nadig[SUP] 1 [/SUP], Mark L Schiebler[SUP] 1 [/SUP], John W Garrett[SUP] 1 [/SUP], Ke Li[SUP] 1 [/SUP], Scott B Reeder[SUP] 1 [/SUP], Guang-Hong Chen[SUP] 1 [/SUP]
Affiliations
- PMID: 32969761
- DOI: 10.1148/radiol.2020202944
Abstract
Background Radiologists are proficient in differentiating between chest x-ray radiographs (CXRs) with and without symptoms of pneumonia, but have found it more challenging to differentiate CXRs with COVID-19 pneumonia symptoms from those without. Purpose To develop an artificial intelligence algorithm to differentiate COVID-19 pneumonia from other causes of CXR abnormalities. Materials and Methods In this retrospective study, a deep neural network, CV19-Net, was trained, validated, and tested on CXRs from patients with and without COVID-19 pneumonia. For the COVID-19 positive CXRs, patients with reverse transcriptase polymerase chain reaction positive results for severe acute respiratory syndrome coronavirus 2 with positive pneumonia findings between February 1, 2020 and May 30, 2020 were included. For the non-COVID-19 CXRs, patients with pneumonia who underwent CXR between October 1, 2019 and December 31, 2019 were included. Area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were calculated to characterize diagnostic performance. To benchmark the performance of CV19-Net, a randomly sampled test dataset containing 500 CXRs from 500 patients was evaluated by both the CV19-Net and three experienced thoracic radiologists. Results A total of 2060 patients (5806 CXRs; mean age 62 ? 16, 1059 men) with COVID-19 pneumonia and 3148 patients (5300 CXRs; mean age 64 ? 18, 1578 men) with non-COVID-19 pneumonia were included and split into training + validation and test datasets. For the test set, CV19-Net achieved an AUC of 0.92 (95% confidence interval [CI]: 0.91, 0.93) corresponding to a sensitivity of 88% (95% CI: 87%, 89%) and a specificity of 79% (95% CI: 77%, 80%) using a high sensitivity operating threshold, or a sensitivity of 78% (95% CI: 77%, 79%) and a specificity of 89% (95% CI: 88%, 90%) using a high specificity operating threshold. For the 500 sampled CXRs, CV19-Net achieved an AUC of 0.94 (95% CI: 0.93, 0.96) compared to a 0.85 AUC (95% CI: 0.81, 0.88) of radiologists. Conclusion CV19-Net was able to differentiate COVID-19 related pneumonia from other types of pneumonia with performance exceeding that of experienced thoracic radiologists.