COVID-19 Diagnosis on Chest Radiograph Using Artificial Intelligence

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Infectious Disease Diagnostics

Background:

• The COVID-19 pandemic has caused global morbidity and mortality.
• Disparities exist in the availability and efficacy of diagnostic tools worldwide.
• Chest radiography analysis is explored for COVID-19 diagnosis.

Purpose of the Study:

• To assess the diagnostic capability of commercial artificial intelligence (AI) software for COVID-19 using chest radiographs.
• To compare the performance of AI software against board-certified radiologists in identifying COVID-19.

Main Methods:

• Symptomatic patients' chest radiographs were analyzed by AI software and radiologists.
• A total of 120 radiographs from patients with confirmed positive or negative RT-PCR tests were included.
• Performance was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC) values.

Main Results:

• The AI software demonstrated performance significantly better than chance (p = 0.001).
• AI software's diagnostic performance was not statistically different from that of radiologists (p = 0.78).

Conclusions:

• Commercial AI software is non-inferior to radiologists in diagnosing COVID-19 via chest radiograph analysis.
• AI advances offer a valuable tool for analyzing chest radiographs in COVID-19 diagnosis.
• RT-PCR testing remains the gold standard, but AI complements diagnostic capabilities.