Related Experiment Video
Updated: Jul 4, 2025
07:50
A Metadata Extraction Approach for Clinical Case Reports to Enable Advanced Understanding of Biomedical Concepts
Published on: September 20, 2018
15.9K
Entity relationship extraction from Chinese electronic medical records based on feature augmentation and cascade
Xiaoqing Lu1, Jijun Tong2, Shudong Xia3
1School of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou, China.
Mathematical Biosciences and Engineering : MBE
|February 2, 2024
Summary
This study introduces a new neural network for extracting relations from long Chinese electronic medical records. The model achieves high precision and recall, improving medical information extraction from complex documents.
Area of Science:
- Natural Language Processing
- Medical Informatics
- Machine Learning
Background:
- Extracting entity relations from unstructured Chinese electronic medical records is crucial for medical information extraction.
- Existing models struggle with the document-level length of Chinese electronic medical records, leading to performance issues.
- The need for advanced models capable of handling long text sequences in medical contexts is evident.
Purpose of the Study:
- To propose a novel neural network framework for entity relation extraction from document-level Chinese electronic medical records.
- To address the limitations of existing models in processing long medical text sequences.
- To develop a model that improves the accuracy and efficiency of medical information extraction.
Main Methods:
- A neural network model employing a feature augmentation and cascade binary tagging framework.
- Utilizing a pre-trained model for text tokenization and obtaining word embedding vectors.
- Feature augmentation network to fuse word vectors with original and positional features, followed by a cascade binary tagging decoder.
Main Results:
- The proposed model achieved a precision of 87.82% and a recall of 88.47% on the newly built VSCMeD dataset.
- Validation on the CMeIE-V2 dataset demonstrated the model's effectiveness with a precision of 54.51% and recall of 48.63%.
- The model successfully handles document-level Chinese electronic medical records, outperforming existing approaches.
Conclusions:
- The feature augmentation and cascade binary tagging neural network framework is effective for entity relation extraction in Chinese electronic medical records.
- The model demonstrates significant improvements in handling long text sequences characteristic of medical documents.
- The developed VSCMeD dataset provides a valuable resource for advancing research in Chinese medical information extraction.
