Active Deep Learning-Based Annotation of Electroencephalography Reports for Cohort Identification
Ramon Maldonado1, Travis R Goodwin1, Sanda M Harabagiu1
1University of Texas at Dallas, Richardson, TX, U.S.A.
Summary
Annotating Electroencephalography (EEG) reports for medical concepts is challenging. A new framework combining active and deep learning shows promise for accurate, attribute-rich annotations.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Medicine
- Clinical Data Annotation
Background:
- Accurate annotation of Electroencephalography (EEG) reports is essential for developing patient cohort retrieval systems.
- Automated annotation of large-scale EEG data, including medical concepts, polarity, and modality, presents significant challenges.
Purpose of the Study:
- To develop and evaluate a novel framework for the automated annotation of EEG reports.
- To capture diverse attributes of medical concepts within EEG reports.
Main Methods:
- A hybrid framework integrating active learning and deep learning techniques was developed.
- The framework was designed to handle the complexities of Big Data in clinical settings.
Main Results:
- The proposed framework successfully produced annotations capturing various attributes of medical concepts.
- Preliminary results indicate significant promise for the framework's effectiveness in EEG report annotation.
Conclusions:
- The novel active and deep learning framework offers a promising solution for the automated annotation of EEG reports.
- This approach can enhance the development of specialized patient cohort retrieval systems by improving data quality and detail.
