Interactive software for generation and visualization of structured findings in radiology reports

Area of Science:

• Radiology Informatics
• Medical Natural Language Processing
• Clinical Decision Support

Background:

• Traditional radiology reporting lacks standardization, hindering efficient data retrieval and analysis.
• Integrating natural language processing (NLP) and speech recognition can streamline the reporting workflow.

Purpose of the Study:

• To develop a user-friendly graphic interface for a module integrating radiology reporting, NLP, and editing.
• To facilitate structured radiology reports in routine clinical practice.
• To implement a hardware-independent module using commercial speech recognition for online transcription.

Main Methods:

• Development of a graphic interface module combining traditional reporting, NLP, and editing.
• Integration of a commercial speech recognition module for real-time transcription.
• Implementation in a hardware-independent environment.

Main Results:

• The module was tested with 150 chest radiology reports by two radiologists.
• Achieved an accuracy close to 90% with high user satisfaction.
• Demonstrated the feasibility of structuring reports for intelligent indexing and retrieval.

Conclusions:

• Structured radiology reports, facilitated by this module, enable intelligent indexing and retrieval.
• The system supports teaching and research by making data more accessible.
• High accuracy and user satisfaction indicate the module's clinical utility.