This research introduces a semantic consistency-based uncertainty quantification framework to identify and reduce hallucinations in AI-generated radiology reports.

• Addresses critical factuality challenges in automated medical report generation
• Proposes a new uncertainty measurement based on semantic consistency
• Demonstrates improved accuracy by identifying areas where the AI model might be uncertain
• Enhances patient safety by flagging potentially inaccurate diagnostic information

For healthcare organizations, this advancement represents a significant step toward safer deployment of AI-assisted radiology reporting systems that radiologists can trust, reducing risks and improving clinical workflow efficiency.

Semantic Consistency-Based Uncertainty Quantification for Factuality in Radiology Report Generation