Related Experiment Video
Updated: May 27, 2025
07:50
A Metadata Extraction Approach for Clinical Case Reports to Enable Advanced Understanding of Biomedical Concepts
Published on: September 20, 2018
15.7K
Leveraging natural language processing for efficient information extraction from breast cancer pathology reports:
Phillip Park1,2, Yeonho Choi2, Nayoung Han3
1Department of Digital Health, Samsung Advanced Institute for Health Sciences and Technology, Sung Kyun Kwan University, Seoul, Korea.
Plos One
|February 18, 2025
Summary
Natural language processing (NLP) models, specifically BioBERT, demonstrate superior accuracy in extracting data from breast cancer pathology reports compared to traditional regular expressions. This advancement enhances the analysis of critical diagnostic information.
Area of Science:
- Oncology
- Medical Informatics
- Computational Linguistics
Background:
- Accurate pathology reports are crucial for breast cancer diagnosis and treatment.
- Breast cancer is a leading global cancer affecting women.
Purpose of the Study:
- To compare the accuracy of regular expressions and natural language processing (NLP) for data extraction from breast cancer pathology reports.
- To evaluate different NLP models, including BERT variants, for this task.
Main Methods:
- Utilized 1,215 breast cancer pathology reports for NLP model development.
- Trained and evaluated three BERT models (BERT-basic, BioBERT, ClinicalBERT) using k-fold cross-validation.
- Compared NLP model performance against regular expressions using named entity recognition (NER).
Main Results:
- BioBERT achieved the highest accuracy (average performance = 0.99901 with k=5) among the evaluated NLP models.
- BioBERT demonstrated a lower error rate across all report items (accuracy ≥ 0.9) compared to other BERT models.
- BioBERT significantly outperformed regular expressions in extracting key items like intraductal component, lymph node status, and lymphovascular invasion.
Conclusions:
- The NLP model BioBERT offers higher accuracy than regular expressions for processing breast cancer pathology reports.
- BioBERT's superior performance highlights its importance in automated analysis of oncological pathology data.
