Mammography reporting dataset with BI-RADS system for natural language processing applications: Addressing public...

Area of Science:

• Medical Informatics
• Natural Language Processing
• Radiology

Background:

• Automating clinical data analysis using Natural Language Processing (NLP) is crucial for enhancing diagnostic accuracy and healthcare efficiency.
• Mammography reports contain vital information for breast cancer screening and diagnosis.

Purpose of the Study:

• To present a novel, anonymized dataset of Spanish mammography reports for NLP and machine learning research.
• To facilitate the development of automated systems for classifying BI-RADS (Breast Imaging Reporting and Data System) categories and analyzing mammography findings.

Main Methods:

• A dataset of 4357 Spanish mammography reports was compiled from multiple medical units.
• Reports were segmented into sections (observations, conclusions, recommendations) and included BI-RADS classifications.
• Anonymization and manual verification ensured data quality and patient confidentiality.

Main Results:

• The dataset comprises 4357 records with 15 variables, including full report text, section-specific text, BI-RADS classification, and relevant metadata.
• The data is structured for direct application in developing and validating NLP and machine learning models.

Conclusions:

• This dataset represents a valuable resource for advancing AI in medical imaging analysis, particularly for mammography reporting.
• It supports research in automated BI-RADS classification and the extraction of clinical insights from unstructured text data.