Informatics in radiology: RADTF: a semantic search-enabled, natural language processor-generated radiology teaching...

Area of Science:

• Radiology Informatics
• Medical Education Technology
• Natural Language Processing in Healthcare

Background:

• Storing and retrieving radiology cases for education and research is often time-consuming.
• Current methods may omit incidental findings or lack structure and de-identification.
• Searching unstructured radiology reports requires advanced methods.

Purpose of the Study:

• To develop an open-source, RadLex-compatible solution for creating searchable radiology teaching files.
• To leverage natural language processing (NLP) for extracting and de-identifying teaching-relevant information.
• To convert existing radiology information system (RIS) and picture archiving and communication system (PACS) archives into on-demand educational resources.

Main Methods:

• Developed RADTF, an open-source teaching file solution compatible with RadLex.
• Utilized NLP to process radiology reports, extracting and de-identifying teaching-relevant statements.
• Created a stand-alone, searchable database from RIS and PACS data.

Main Results:

• Generated a semantic search-enabled, web-based radiology archive.
• The archive contains over 700,000 cases with millions of images.
• RADTF effectively combines teaching content representation with a versatile search engine.

Conclusions:

• RADTF provides an efficient method for creating a comprehensive, searchable radiology teaching resource.
• The system enhances the utility of existing RIS-PACS archives for educational and research purposes.
• NLP-driven extraction and de-identification improve the accessibility and usability of radiology case data.