Evaluation of a COVID-19 Pneumonia CXR AI Detection Algorithm

• Conditions: Covid19
• Interventions: Diagnostic Test: AI model

• Registration Number: NCT04561024
• Lead Sponsor: Ensemble Group Holdings, LLC

Brief Summary

This study investigates the diagnostic performance of an AI algorithm in the detection of COVID-19 pneumonia on chest radiographs.

Detailed Description

This is an international multi-center study. Chest radiographs (CXR) from different participating centers will be collected to develop an AI algorithm to detect COVID-19 pneumonia. This will be tested on external hold out datasets from different centers using SARS-CoV-2 by Real-Time Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) Assay as ground truth.

Study Timeline

• Study Start: 2020-03-01
• Status Verified: 2020-09
• Study First Submitted: 2020-09-22
• Study First Submitted QC: 2020-09-22
• Last Update Submitted: 2020-09-22
• Study First Posted: 2020-09-23
• Last Update Posted: 2020-09-24
• Primary Completion: 2020-10-30
• Study Completion: 2020-12-31

Recruitment & Eligibility

• Status: UNKNOWN
• Sex: All
• Target Recruitment: 4000

Inclusion Criteria

• All adult patients >18 years of age
• Attended any of the participating institutes between February 1, 2020 until September, 2020
• Underwent both RT-PCR testing and frontal CXR (within 48 hours of PCR testing) for COVID-19 infection
• frontal CXR of patients pre-covid pandemic

Exclusion Criteria

• Unavailability of patient demographics and clinical data
• Inconclusive RT-PCR results
• CXR considered to be of non-diagnostic quality by the clinical radiology research team at each site
• CXR not in a retrievable or processable format for AI inference

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Arm && Interventions

Group	Intervention	Description
RT-PCR Positive Patients	AI model	RT-PCR confirmed patients positive for SARS-CoV-2
Negative patients	AI model	RT-PCR confirmed patients negative for SARS-CoV-2 or patients with CXR performed before the emergence of COVID-19 pandemic

Primary Outcome Measures

Name	Time	Method
Diagnostic Performance of AI model	9 months	Performance (accuracy, sensitivity, specificity, false-positive rate (FPR), false-negative rate (FNR), and Area Under the Curve (AUC)) of the AI model in detection of COVID-19 pneumonia on their baseline CXR using RT-PCR and historical controls as gold standard in a multi-center / multi-national cohort.

Trial Locations

Locations (1)

University of Hong Kong; 🇭🇰 Hong Kong, Hong Kong