RGSS-ID: an approach to new radiologic reporting system

Area of Science:

• Computer Science
• Medical Informatics
• Artificial Intelligence

Background:

• Radiology reporting often uses unstructured natural language, hindering automated analysis.
• Integrating artificial intelligence (AI) with existing reporting systems presents challenges in data representation.
• Bridging the gap between human language and computational methods is crucial for advancing medical informatics.

Purpose of the Study:

• To develop a developmental computer system, RGSS-ID, applying AI to radiology reporting.
• To propose the Generalized Finding Representation (GFR) scheme for seamless integration of natural language and AI.
• To enable computer parsing of radiologist-composed sentences for enhanced data encoding.

Main Methods:

• RGSS-ID utilizes AI methods within a reporting system framework.
• The Generalized Finding Representation (GFR) scheme is central to encoding findings.
• Data entry involves item selection, allowing radiologists to compose computer-parsable sentences.
• Encoded findings are stored in a knowledge database using GFR expressions.

Main Results:

• RGSS-ID successfully encodes natural language radiology findings into a structured format (GFR).
• The system facilitates the creation of computer-parsable sentences from radiologist input.
• A knowledge database is populated with structured findings for potential AI application.
• The final report is generated in natural language, maintaining clinical readability.

Conclusions:

• RGSS-ID demonstrates a viable approach to integrating AI with radiology reporting systems.
• The Generalized Finding Representation (GFR) effectively bridges natural language and AI methods.
• This system enhances the potential for automated analysis and knowledge discovery in radiology data.