Development of a system that generates structured reports for chest x-ray radiography

Area of Science:

• Radiology
• Medical Informatics
• Natural Language Processing

Background:

• Traditional narrative radiology reports hinder advanced data analysis and decision support.
• Structured reporting is crucial for integrating AI and improving clinical workflows.

Purpose of the Study:

• To develop and evaluate a system for generating structured chest x-ray radiography reports.
• To overcome the limitations of narrative reporting in radiology.

Main Methods:

• Analyzed existing reports to define sentence structures (procedure, region, finding, diagnosis).
• Utilized Resource Description Framework (RDF) to model terms and relations.
• Developed a guided reporting system with term highlighting and templates.
• Compared system-generated reports with free-text reports from radiologists for 50 chest x-rays.

Main Results:

• System reduced report generation time by 12.5% compared to free-text.
• Generated sentences showed high concordance with free-text reports (F-measure = 90%).
• Radiologists found the system applicable to daily clinical practice.

Conclusions:

• Generating structured reports for chest x-rays is feasible and efficient.
• The system produces accurate reports in less time than traditional methods.
• Structured reporting facilitates advanced applications in radiology.