[RGSS-IDJ and its application to cranial computed tomography]

Area of Science:

• Medical Informatics
• Artificial Intelligence in Radiology
• Natural Language Processing

Background:

• Radiographic reporting is crucial for diagnosis.
• Integrating artificial intelligence (AI) into reporting systems presents challenges.
• Bridging natural language and AI requires effective data representation.

Purpose of the Study:

• To develop RGSS-IDJ, the Japanese version of the Report Generation Support System for Imaging Diagnosis (RGSS-ID).
• To support AI-driven report generation for cranial computed tomography (CT).
• To propose and implement the Generalized Finding Representation (GFR) scheme for natural language-AI integration.

Main Methods:

• Developed RGSS-IDJ as a Japanese AI-based reporting system for cranial CT.
• Employed the Generalized Finding Representation (GFR) scheme, consistent with RGSS-ID.
• Utilized a dialogue system for inputting radiographic findings via query-response and mouse selection.
• Encoded findings into network expressions (LISP list form) and stored them in a knowledge database.

Main Results:

• RGSS-IDJ successfully supports report generation for cranial CT in Japanese.
• The GFR scheme effectively bridges natural language findings and AI processing.
• Encoded findings are stored in a knowledge base for AI analysis.
• The system generates final radiographic reports in natural Japanese language.

Conclusions:

• RGSS-IDJ is a functional AI system for Japanese cranial CT reporting.
• GFR facilitates the use of radiographic report content for AI-driven analyses.
• The system enables efficient and structured generation of radiological reports.