Extraction of BI-RADS findings from breast ultrasound reports in Chinese using deep learning approaches

The wide adoption of electronic health record systems (EHRs) in hospitals in China has made large amounts of data available for clinical research including breast cancer. Unfortunately, much of detailed clinical information is embedded in clinical narratives e.g., breast radiology reports. The American College of Radiology (ACR) has developed a Breast Imaging Reporting and Data System (BI-RADS) to standardize the clinical findings from breast radiology reports.

This study aims to develop natural language processing (NLP) methods to extract BI-RADS findings from breast ultrasound reports in Chinese, thus to support clinical operation and breast cancer research in China.

We developed and compared three different types of NLP approaches, including a rule-based method, a traditional machine learning-based method using the Conditional Random Fields (CRF) algorithm, and deep learning-based approaches, to extract all BI-RADS finding categories from breast ultrasound reports in Chinese.

Using a manually annotated dataset containing 540 reports, our evaluation shows that the deep learning-based method achieved the best F1-score of 0.904, when compared with rule-based and CRF-based approaches (0.848 and 0.881 respectively).

This is the first study that applies deep learning technologies to BI-RADS findings extraction in Chinese breast ultrasound reports, demonstrating its potential on enabling international collaborations on breast cancer research.