Detection of Interstitial Lung Abnormalities on Chest Radiographs: Diagnostic Performance of Radiologists and...

Area of Science:

• Pulmonary Medicine
• Radiology
• Artificial Intelligence in Medicine

Background:

• Interstitial lung abnormalities (ILA) detected via CT scans are increasingly recognized for their link to interstitial lung disease progression.
• The diagnostic utility of standard chest radiography for identifying ILA remains underexplored.

Purpose of the Study:

• To assess the diagnostic performance of both human radiologists and an AI model in detecting ILA on chest radiographs.
• To compare their performance against CT scans, which serve as the gold standard.

Main Methods:

• A retrospective analysis of 1168 adult patients who underwent both chest CT and radiography.
• Radiologists used a 5-point Likert scale, while an AI model generated probability scores for ILA detection.
• CT scans were used as the reference standard for classifying fibrotic and nonfibrotic ILA.

Main Results:

• Both radiologists and AI demonstrated high specificity but suboptimal sensitivity in detecting ILA on radiographs.
• AI performance showed a trend towards higher accuracy and AUC compared to radiologists for fibrotic ILA.
• Radiographic ILA identified by AI was independently associated with increased respiratory disease-related mortality.

Conclusions:

• Chest radiography, interpreted by either radiologists or AI, has limitations in sensitivity for ILA detection.
• Despite a significant association with respiratory mortality, current findings do not support using chest radiography for ILA screening.