An Accurate Deep Learning Model for Clinical Entity Recognition From Clinical Notes.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Biology

Background:

• Electronic health records (EHRs) generate vast amounts of clinical notes containing valuable medical information.
• Extracting clinical entities (diseases, treatments, drugs, etc.) from unstructured clinical notes is crucial for various applications.
• Existing machine learning and deep learning models face challenges in accurately extracting these entities.

Purpose of the Study:

• To develop a novel deep learning-based technique for accurate clinical entity extraction from clinical notes.
• To improve upon existing models by incorporating both local and global contextual information.
• To enhance the utility of clinical notes for applications like data analysis and privacy preservation.

Main Methods:

• A novel deep learning model combining Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (Bi-LSTM), and Conditional Random Fields (CRF).
• The model utilizes a combination of local and global context for entity extraction, unlike previous methods relying solely on global context.
• Employs non-complex embeddings for efficient processing.

Main Results:

• The proposed model achieved superior performance compared to existing methods.
• Outperformed existing models by 4-10% (i2b2-2010) and 5-12% (i2b2-2012) in F1-score.
• Demonstrated the effectiveness of combining CNN, Bi-LSTM, and CRF with local and global context.

Conclusions:

• The novel deep learning model offers a more accurate approach to clinical entity extraction.
• Accurate entity detection from clinical notes can significantly aid in medical data privacy preservation.
• Enhanced trust in sharing medical data between users and organizations can be fostered through improved data privacy.