Model-based clinical note entity recognition for rheumatoid arthritis using bidirectional encoder representation from...

Area of Science:

• Natural Language Processing
• Artificial Intelligence in Medicine
• Biomedical Informatics

Background:

• Rheumatoid arthritis (RA) is a disabling immune system disease.
• Clinical notes contain valuable diagnostic and treatment information.
• Artificial intelligence (AI) can effectively mine this clinical data.

Purpose of the Study:

• To develop an effective AI method for identifying and classifying RA-related medical entities in clinical notes.
• To utilize these entity identification results for subsequent research.

Main Methods:

• Utilized the Bidirectional Encoder Representation from Transformers (BERT) pre-training model to enhance word vector semantic representation.
• Employed a model combining BERT with bidirectional long short-term memory (BiLSTM) and conditional random field (CRF) for named entity recognition (NER).
• Input combined token, segment, and position embeddings into the BERT model, fine-tuning during training.

Main Results:

• The BERT model significantly outperformed the traditional Word2vec model for word vector generation.
• Achieved a best F1-score of 0.936 for the named entity recognition task on RA clinical notes.

Conclusions:

• Confirms the effectiveness of advanced AI for NER on large clinical note corpora.
• This application shows promise for medical settings.
• Provides foundational data for relation extraction, knowledge graph construction, and disease reasoning.