Artificial Intelligence-Based Automated Detection of Chest X-ray Abnormalities as Support for Young Radiologists

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Chest X-ray (CXR) interpretation varies among radiologists.
• Automatic detection tools are emerging to aid CXR interpretation.
• Evaluating AI's impact on radiologist performance is crucial.

Purpose of the Study:

• To assess if AI software enhances young radiologists' performance in CXR interpretation.
• To compare AI-assisted radiologist performance against an experienced radiologist.

Main Methods:

• 500 CXRs analyzed by AI software and junior radiologists.
• Junior radiologists re-evaluated CXRs with AI assistance after three months.
• Key findings: atelectasis, calcification, cardiomegaly, consolidation, fibrosis, nodules, mediastinal widening, pleural effusion, pneumoperitoneum, pneumothorax.
• Sensitivity and specificity calculated against an expert reader.

Main Results:

• AI demonstrated significantly higher sensitivity than junior radiologists alone.
• Junior radiologists showed significantly higher specificity without AI.
• AI assistance significantly increased sensitivity for both radiologists.
• AI assistance led to a significant decrease in specificity.

Conclusions:

• AI software excels in high sensitivity, minimizing false negatives.
• AI assistance can enhance junior radiologists' sensitivity in CXR interpretation.
• A potential risk of overdiagnosis exists with AI assistance due to decreased specificity.