Automatic ICD Coding by using Information Retrieval Method Okapi-25 and Exchangeable Word Pairs

DOI 23 References
  • Aramaki E
    The University of Tokyo Hospital
  • Imai T
    The University of Tokyo Hospital
  • Kajino M
    Japan Research Groupe for Medical Ontology, Japan
  • Miyo K
    The University of Tokyo Hospital
  • Ohe K
    The University of Tokyo Hospital

Bibliographic Information

Other Title
  • 情報検索尺度Okapi-BM25と交換可能語ペアを用いた自動ICDコーディングに関する研究

Search this article

Abstract

[OBJECTIVE] International Statistical Classification of Diseases and Related Health Problems (ICD10) is one of the most standard and important disease classifications. Because the ICD10 coding is the time/cost consuming task, computerized coding systems have much attention in the medical field. [METHOD] This paper proposes a new coding method based on document classification techniques. First, coding-examples that are similar to an input term are extracted. In calculating term similarity, we used a BM25-based similarity, dealing with exchangeable word pairs and their contexts. Then, by voting of the extracted coding-examples, the final ICD code is decided. [RESULT] The experimental results showed the 44.0% accuracy for 1,625 unseen disease names. Especially, long disease names, the accuracy is higher (50.8%), demonstrating the basic feasibility of the proposed approach.

Journal

References(23)*help

See more

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top