Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Segment convolutional neural networks (Seg-CNNs) for classifying relations in clinical notes.

Yuan Luo1, Yu Cheng2, Özlem Uzuner3

  • 1Department of Preventive Medicine, Northwestern University, Chicago, IL, USA.

Journal of the American Medical Informatics Association : JAMIA
|October 13, 2017
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Solutions for Unexpected Challenges Encountered when Integrating Research Genomics Results into the EHR.

ACI open·2026
Same author

The Maturity of Maturity Models in Health Care and Informatics.

AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science·2026
Same author

A Privacy Nihilist Perspective on Clinical Data Sharing: Open Clinical Data Sharing is Dead, Long Live the Walled Garden.

Journal of the Society for Clinical Data Management·2026
Same author

Automated identification of incidentalomas requiring follow-up: A multi-anatomy evaluation of LLM-based and supervised approaches.

Journal of biomedical informatics·2026
Same author

A scoping review of natural language processing in addressing medically inaccurate information: Errors, misinformation, and hallucination.

Journal of biomedical informatics·2025
Same author

Toward responsible AI governance: Balancing multi-stakeholder perspectives on AI in healthcare.

International journal of medical informatics·2025

Segment Convolutional Neural Networks (Seg-CNNs) classify relations in clinical notes using word embeddings. This approach achieves state-of-the-art results without manual feature engineering.

Area of Science:

  • Natural Language Processing
  • Machine Learning
  • Biomedical Informatics

Background:

  • Clinical notes contain complex relationships between medical concepts.
  • Accurate relation classification is crucial for clinical decision support and research.
  • Existing methods often require extensive manual feature engineering.

Purpose of the Study:

  • To introduce Segment Convolutional Neural Networks (Seg-CNNs) for automated relation classification in clinical text.
  • To evaluate the performance of Seg-CNNs on a benchmark dataset without manual feature engineering.
  • To demonstrate the effectiveness of learning segment-level representations for relation extraction.

Main Methods:

  • Seg-CNNs utilize word-embedding features, eliminating the need for manual feature engineering.
Keywords:
convolutional neural networkmachine learningmedical relation classificationnatural language processing

Related Experiment Videos

  • The model learns separate representations for distinct text segments (preceding, concept1, middle, concept2, succeeding) within a sentence.
  • Evaluation was performed on the i2b2/VA relation classification challenge dataset.

    • Main Results:

    • Seg-CNN achieved a state-of-the-art micro-average F-measure of 0.742 for overall relation classification.
    • Specific F-measures include 0.686 for problem-treatment, 0.820 for problem-test, and 0.702 for problem-problem relations.
    • The study demonstrated the benefit of segment-level representations and the positive impact of medical domain word embeddings.

    Conclusions:

    • Seg-CNNs offer a powerful, automated approach for classifying relations in clinical notes.
    • The model achieves state-of-the-art performance, highlighting the efficacy of segment-level representation learning.
    • Seg-CNNs are computationally efficient, trainable on GPUs, and eliminate the need for manual feature engineering.