Proposing New RadLex Terms by Analyzing Free-Text Mammography Reports

Area of Science:

• Radiology
• Medical Informatics
• Natural Language Processing

Background:

• The RadLex terminology, essential for radiology, has existing gaps.
• Developing comprehensive terminologies requires continuous updates and expansion.

Purpose of the Study:

• To develop a data-driven method for discovering new terms to enhance the RadLex terminology, specifically for mammography.
• To leverage natural language processing (NLP) to analyze free-text radiology reports for candidate RadLex terms.

Main Methods:

• A NLP system was developed to extract and classify noun phrases from free-text mammography reports.
• The system was evaluated using expert radiologist annotations on a test set of 100 reports.
• Performance metrics including precision and recall were calculated.

Main Results:

• The NLP system achieved a precision of 0.77 and a recall of 0.94 in identifying candidate RadLex terms.
• Overall system accuracy was 0.80.
• Analysis of 270,540 reports identified 31,800 unique noun phrases as potential RadLex candidates.

Conclusions:

• A data-driven NLP approach effectively identifies new candidate terms for expanding the RadLex terminology in mammography.
• This method shows promise for discovering new terms across various radiology domains.
• The findings contribute to a more comprehensive breast imaging lexicon within RadLex.