Using image references in radiology reports to support enhanced report-to-image navigation

Area of Science:

• Medical Informatics
• Radiology
• Natural Language Processing

Background:

• Radiology reports often cite specific image slices for findings.
• Current methods require manual navigation through multiple steps to access referenced images.

Purpose of the Study:

• To develop and evaluate an automated system for creating hyperlinks from radiology report image references to the actual image slices.
• To enhance the efficiency of accessing visual data within radiology reports.

Main Methods:

• A regular expressions-based algorithm was developed to identify image references within radiology report sentences.
• The algorithm was validated on 314 image references from general radiology reports.

Main Results:

• The algorithm achieved high performance metrics: 99.35% precision, 98.08% recall, and 98.71% F-measure.
• Demonstrated the feasibility of integrating these hyperlinks into Picture Archiving and Communication Systems (PACS) report viewers.

Conclusions:

• Automated image reference extraction using regular expressions is a viable approach for improving report-to-image navigation.
• Hyperlinking enables one-click access to referenced images, streamlining clinical workflows.