Construction of an infectious disease epidemic prediction model by deep learning

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K10614
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 58030:Hygiene and public health-related: excluding laboratory approach
• Research Institution: Gunma University
• Principal Investigator: Uchida Mitsuo 群馬大学, 大学院医学系研究科, 准教授 (00377251)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 感染症 / AI / RNN / LSTM / 数理モデル / 流行予測

Outline of Final Research Achievements

The purpose of this study was to create a model to predict future epidemics of infectious diseases using AI technology with past epidemiological data. Data were collected at the Gunma Prefectural Institute of Public Health and Environment for the past 11 years from 2009 to 2019, including influenza, RS virus, pharyngoconjunctival fever, group A streptococcus, infectious gastroenteritis, varicella, hand-foot-and-mouth disease, erythema infectiosum, sudden rash, herpangina, epidemic mumps, epidemic keratitis conjunctivitis, and Mycoplasma. As a result of constructing a LSTM model, we were able to create a highly accurate prediction model for influenza and RS virus which were with enough cases.

Free Research Field

感染症の疫学

Academic Significance and Societal Importance of the Research Achievements

本研究は，毎年周期的に流行を引き起こす感染症に対し，“その年の流行を予測できれば医療資源の準備や病床の確保を行うための参考情報にできるのではないか？”という発想の下で行われた。本研究の成果より，毎年多数報告されるインフルエンザやRSウイルスは予測精度の高いモデルを構築することができたが，他方，報告数の多くない感染症の予測精度は高くなかった。現在の学習型のAIは，学習のために多数のサンプル数を必要とするため，報告数の少ない感染症への対応に課題が残された。この課題は，今後の研究により解決することが望まれる。