To evaluate the impact of collaborative artificial intelligence assistance on chest x-ray report writing time and quality metrics.
Key Findings:
Mean writing time was 105 seconds with AI assistance vs. 114 seconds without.
Writing time decreased by 27% for one radiologist, 11% for another, and increased by 9% for the third, highlighting significant variability.
Suggestion acceptance rates ranged from 41% to 68%.
AI assistance led to an 18% reduction in writing time for longer reports but a 13% increase for shorter ones.
Automated report quality metrics were similar across conditions, with slight advantages for AI assistance in some scores.
Interpretation:
The study suggests potential efficiency gains from AI assistance in chest x-ray reporting, but results vary significantly by radiologist and case complexity.
Limitations:
Small sample size of 50 chest x-rays and only three radiologists.
No formal washout period and lack of structured templates in the unassisted condition.
Quality assessment based on automated metrics rather than independent radiologist review.
Potential for AI-generated text to be influenced by language patterns rather than image content.
Conclusion:
Further research is needed with larger samples and diverse radiologist experience to validate findings and assess AI adoption in clinical workflows.
