Generative Artificial Intelligence for Chest Radiograph Interpretation in the Emergency Department

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Multimodal generative artificial intelligence (AI) offers potential for optimizing emergency department (ED) care through automated radiology report generation.
• Evaluating AI's accuracy and quality in interpreting ED chest radiographs is crucial for clinical integration.

Purpose of the Study:

• To assess the accuracy and quality of AI-generated chest radiograph interpretations in an emergency department setting.
• To compare AI interpretations with those of on-site radiologists and teleradiology services.

Main Methods:

• A retrospective diagnostic study analyzed 500 ED chest radiographs.
• Interpretations from AI, on-site radiologists, and teleradiology were rated by emergency physicians on a 5-point Likert scale.
• Statistical models were used to compare Likert scores and the probability of clinically significant discrepancies.

Main Results:

• AI and radiologist reports received significantly higher ratings than teleradiology reports (P < .001).
• AI and radiologist reports showed no significant difference in quality ratings.
• No significant differences in the probability of clinically significant discrepancies were found between AI, radiologist, and teleradiology reports across various findings.

Conclusions:

• Generative AI models produce chest radiograph reports with clinical accuracy and textual quality comparable to human radiologists.
• AI reports demonstrated superior textual quality compared to teleradiology reports.
• Clinical implementation of AI could enhance timely detection of critical findings and aid imaging interpretation in the ED.