Utilizing AI CAD for early pandemic screening in chest radiographs

Area of Science:

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

Background:

• The COVID-19 pandemic created an urgent need for rapid diagnostic tools.
• Existing artificial intelligence (AI) software, initially for pulmonary nodule detection, was repurposed.
• Chest radiographs are crucial for diagnosing pneumonia and COVID-19.

Purpose of the Study:

• To evaluate the performance of pre-trained commercial AI software in diagnosing COVID-19 and other pneumonia on chest radiographs.
• To assess the AI's utility as a screening tool during the pandemic.

Main Methods:

• Utilized a public dataset of chest radiographs from the early COVID-19 pandemic.
• Assessed the performance of AI-assisted computer-aided detection (CAD) software.
• Analyzed sensitivity and specificity for pneumonia detection, including COVID-19.

Main Results:

• The AI software achieved 86.83% sensitivity and 59.59% specificity for detecting pneumonia.
• Performance improved with posterior-anterior (PA) radiographs, reaching 89.16% sensitivity and 67.41% specificity.
• The AI CAD demonstrated potential for initial screening of infectious pulmonary diseases.

Conclusions:

• Existing AI CAD software can be adapted for diagnosing infectious pulmonary diseases like COVID-19.
• AI offers a valuable triage tool, particularly in areas with limited medical resources.
• Rapid adaptation of AI is crucial for addressing healthcare challenges during outbreaks.