An automated COVID-19 triage pipeline using artificial intelligence based on chest radiographs and clinical data

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.

A real-world intelligent system for the diagnosis and triage of COVID-19 in the emergency department

“…One of the most outstanding fields in which AI has been applied in the last decade is computer vision, and the diagnosis of COVID-19 from CXR images seems a proper task to be tackled using this approach. In fact, many groups around the world have explored this idea, and several articles have been published on the subject to date [7][8][9][10]. Unfortunately, the quality of reported research is quite diverse and most proposals cannot be reproduced due to several issues, including lack of data availability, missing details in the description of processes or parameters, or incomplete descriptions of the methods followed.…”

Bending the patient safety curve: how much can AI help?

“…Another area of opportunity for AI in patient safety is automated interpretation of radiology imaging, which is one of the largest categories of healthcare AI publications over the last 5 years 7 . One example occurred at a large academic health system that had substantial AI resources and a widely used commercial EHR system.…”

“…The rapidly increasing use of artificial intelligence (AI) in operational clinical settings presents an opportunity for evaluation as there is limited research or funding for such about its efficacy or safety. There has been steady progress in methods and tools needed to manipulate and transform electronic clinical data, and increasingly mature data resources have supported the rapid development of sophisticated AI in some health care domains including patient safety [6][7][8] . The broad adoption of personal devices such as wrist watches that measure heart rhythms or portable glucose monitors or other patient self-monitoring devices offer far broader data types than traditional EHR data, and integration of this multi-modal data into EHRs seems likely to yield earlier and more actionable AI predictions 7 .…”

“…There has been steady progress in methods and tools needed to manipulate and transform electronic clinical data, and increasingly mature data resources have supported the rapid development of sophisticated AI in some health care domains including patient safety [6][7][8] . The broad adoption of personal devices such as wrist watches that measure heart rhythms or portable glucose monitors or other patient self-monitoring devices offer far broader data types than traditional EHR data, and integration of this multi-modal data into EHRs seems likely to yield earlier and more actionable AI predictions 7 . A recent study demonstrated the impact of clinical deterioration software on patient outcomes, this was homegrown in one large health system and has not been disseminated beyond this system.…”

“…There are few rigorous assessments of actual AI deployments in health care delivery systems, and while there is some limited evidence for improved safety processes or outcomes when these AI tools are deployed [4][5][6][7][8] , there is also evidence that these systems can increase risk if the algorithms are tuned to give overly confident results 9 . For example, within AI risk prediction models, the sizeable literature on model development and validation is in stark contrast to the scant data describing successful clinical deployment and impact of those models in health care settings.…”