Development of a system for predicting the onset of hypertensive disorder of pregnancy using an AI system

2021 Fiscal Year Final Research Report

• Project/Area Number: 20K18168
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 56040:Obstetrics and gynecology-related
• Research Institution: Yamaguchi University
• Principal Investigator: Shinagawa Masahiro 山口大学, 医学部, 特別医学研究員 (50814472)
• Project Period (FY): 2020-04-01 – 2022-03-31
• Keywords: 妊娠高血圧症候群 / AI / 妊婦健診データ

Outline of Final Research Achievements

Blood pressure, urinary protein, and body mass index (BMI) data were obtained from the pregnant women's health examination data. A hidden Markov model (HMM) was used for statistical machine learning.　We also attempted to extend the conventional HMM to allow time (weeks of pregnancy) dependent covariates to be modeled as parameters of the state transition matrix. Using this Markov-dependent mixture model, we calculated the number of states that can efficiently classify the state at the time of the antenatal checkup using systolic blood pressure, diastolic blood pressure, and proteinuria as response variables, and the transition probability to estimate the state at the next checkup at each checkup session. In the validation data, the accuracy of predicting the onset of HDP during pregnancy and the accuracy of predicting the onset of HDP thereafter from data before 31 weeks of pregnancy were calculated.

Free Research Field

産婦人科

Academic Significance and Societal Importance of the Research Achievements

本予測システムは我が国で統一された妊婦定期健診データに基づいているため、新たなデータ取得を必要とすることなく、そのまま他施設での運用が可能である。本研究において、AIを用いたHDP発症予測システムが確立されれば、妊娠管理中に数週間先のHDP発症を予測することが可能となり、早期の治療介入が可能となる。HDPは突然に発症して急激に増悪する疾患のため、発症前に高精度に予測できることは極めて重要な意義を持つ。