Section heading recognition in electronic health records using conditional random fields

Chih Wei Chen, Nai Wen Chang, Yung Chun Chang, Hong Jie Dai

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Electronic health records (EHRs) contain a wealth of information, such as discharge diagnoses, laboratory results, and pharmacy orders, which can be used to support clinical decision support systems and enable clinical and translational research. Unfortunately, the information is represented in a highly heterogeneous semi-structured or unstructured format with author- and domainspecific idiosyncrasies, acronyms and abbreviations. To take full advantage of health data, text-mining techniques have been applied by researchers to recognize named entities (NEs) mentioned in EHRs. However, the judgment of clinical data cannot be known solely from the NE level. For instance, a disease mention in the section of past medical history has different clinical significance when mentioned in the family medical history section. To obtain high-quality information and improve the understanding of clinical records, this work developed a machine learning-based section heading recognition system and evaluated its performance on a manually annotated corpus. The experiment results showed that the machine learning-based system achieved a satisfactory F-score of 0.939, which outperformed a dictionary-based system by 0.321.

Original languageEnglish
Pages (from-to)47-55
Number of pages9
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8916
Publication statusPublished - 2014

Keywords

  • Electronic health record
  • Information extraction
  • Natural language processing
  • Section recognition

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Fingerprint Dive into the research topics of 'Section heading recognition in electronic health records using conditional random fields'. Together they form a unique fingerprint.

  • Cite this