npj Digital Medicine (Jan 2022)
An automated COVID-19 triage pipeline using artificial intelligence based on chest radiographs and clinical data
- Chris K. Kim,
- Ji Whae Choi,
- Zhicheng Jiao,
- Dongcui Wang,
- Jing Wu,
- Thomas Y. Yi,
- Kasey C. Halsey,
- Feyisope Eweje,
- Thi My Linh Tran,
- Chang Liu,
- Robin Wang,
- John Sollee,
- Celina Hsieh,
- Ken Chang,
- Fang-Xue Yang,
- Ritambhara Singh,
- Jie-Lin Ou,
- Raymond Y. Huang,
- Cai Feng,
- Michael D. Feldman,
- Tao Liu,
- Ji Sheng Gong,
- Shaolei Lu,
- Carsten Eickhoff,
- Xue Feng,
- Ihab Kamel,
- Ronnie Sebro,
- Michael K. Atalay,
- Terrance Healey,
- Yong Fan,
- Wei-Hua Liao,
- Jianxin Wang,
- Harrison X. Bai
Affiliations
- Chris K. Kim
- Department of Diagnostic Imaging, Rhode Island Hospital
- Ji Whae Choi
- Department of Diagnostic Imaging, Rhode Island Hospital
- Zhicheng Jiao
- Department of Diagnostic Imaging, Rhode Island Hospital
- Dongcui Wang
- Department of Radiology, Xiangya Hospital, Central South University
- Jing Wu
- Department of Radiology, Xiangya Hospital, Central South University
- Thomas Y. Yi
- Department of Diagnostic Imaging, Rhode Island Hospital
- Kasey C. Halsey
- Department of Diagnostic Imaging, Rhode Island Hospital
- Feyisope Eweje
- Perelman School of Medicine at University of Pennsylvania
- Thi My Linh Tran
- Department of Diagnostic Imaging, Rhode Island Hospital
- Chang Liu
- Department of Radiology, Xiangya Hospital, Central South University
- Robin Wang
- Perelman School of Medicine at University of Pennsylvania
- John Sollee
- Department of Diagnostic Imaging, Rhode Island Hospital
- Celina Hsieh
- Department of Diagnostic Imaging, Rhode Island Hospital
- Ken Chang
- Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital
- Fang-Xue Yang
- Department of Radiology, Xiangya Hospital, Central South University
- Ritambhara Singh
- Department of Computer Science, Brown University
- Jie-Lin Ou
- Department of Radiology, Xiangya Hospital, Central South University
- Raymond Y. Huang
- Department of Radiology, Brigham and Women’s Hospital
- Cai Feng
- Department of Radiology, Xiangya Hospital, Central South University
- Michael D. Feldman
- Perelman School of Medicine at University of Pennsylvania
- Tao Liu
- Department of Biostatistics, Brown University
- Ji Sheng Gong
- Department of Radiology, Xiangya Hospital, Central South University
- Shaolei Lu
- Department of Radiology, Xiangya Hospital, Central South University
- Carsten Eickhoff
- Center for Biomedical Informatics, Brown University
- Xue Feng
- Carina Medical
- Ihab Kamel
- Department of Radiology and Radiological Sciences, Johns Hopkins University
- Ronnie Sebro
- Perelman School of Medicine at University of Pennsylvania
- Michael K. Atalay
- Department of Diagnostic Imaging, Rhode Island Hospital
- Terrance Healey
- Department of Diagnostic Imaging, Rhode Island Hospital
- Yong Fan
- Perelman School of Medicine at University of Pennsylvania
- Wei-Hua Liao
- Department of Radiology, Xiangya Hospital, Central South University
- Jianxin Wang
- School of Computer Science and Engineering, Central South University
- Harrison X. Bai
- Department of Diagnostic Imaging, Rhode Island Hospital
- DOI
- https://doi.org/10.1038/s41746-021-00546-w
- Journal volume & issue
-
Vol. 5,
no. 1
pp. 1 – 9
Abstract
Abstract While COVID-19 diagnosis and prognosis artificial intelligence models exist, very few can be implemented for practical use given their high risk of bias. We aimed to develop a diagnosis model that addresses notable shortcomings of prior studies, integrating it into a fully automated triage pipeline that examines chest radiographs for the presence, severity, and progression of COVID-19 pneumonia. Scans were collected using the DICOM Image Analysis and Archive, a system that communicates with a hospital’s image repository. The authors collected over 6,500 non-public chest X-rays comprising diverse COVID-19 severities, along with radiology reports and RT-PCR data. The authors provisioned one internally held-out and two external test sets to assess model generalizability and compare performance to traditional radiologist interpretation. The pipeline was evaluated on a prospective cohort of 80 radiographs, reporting a 95% diagnostic accuracy. The study mitigates bias in AI model development and demonstrates the value of an end-to-end COVID-19 triage platform.
