Data structuring of electronic medical records and development of artificial intelligence-based model for disease diagnostic support using a novel natural language processing technology

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K18874
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 58010:Medical management and medical sociology-related
• Research Institution: Gunma University
• Principal Investigator: Noguchi Rei 群馬大学, 医学部附属病院, 助教 (50828861)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 電子カルテデータ / 自然言語処理 / 非構造化データ / 診断支援AI / 疾患判別モデル / 症例マトリクス / テキストデータ構造化 / 類似症例予測AI

Outline of Final Research Achievements

This study aims to build an AI that utilises text data from electronic medical records to provide diagnostic support for disease names based on descriptions in medical records.
Through this research, a methodology was established to extract disease and symptom names from text data in electronic medical records using natural language processing technology, and to automatically generate structured data (case matrix) of diseases and symptoms for each case. In addition, using the case matrix as training data, a machine learning model for disease classification (disease classification AI model) that can detect specific cardiovascular diseases with a maximum reproduction rate of 87% could be constructed.

Free Research Field

医療情報学

Academic Significance and Societal Importance of the Research Achievements

真の診断支援AI の構築には、電子カルテのテキストデータの活用が不可欠であるが、非構造化データのために扱いが難しくまだ十分に活用されていない。本研究は既存の電子カルテデータの活用可能性を広げるとともに、将来的な診断支援AI の実現に向けたコア技術となるものであり、医療の質向上や均てん化、医師の負担軽減に大きく貢献できると考えられる。