Prediction and stratification of acute kidney injury with a machine learning algorithm in intensive care unit

2023 Fiscal Year Final Research Report

• Project/Area Number: 19K18321
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 55060:Emergency medicine-related
• Research Institution: The University of Tokyo (2022-2023); Kyoto University (2019-2021)
• Principal Investigator: Sato Noriaki 東京大学, 医科学研究所, 助教 (90838997)
• Project Period (FY): 2019-04-01 – 2024-03-31
• Keywords: 機械学習 / 急性腎障害 / 集中治療部

Outline of Final Research Achievements

Acute kidney injury (AKI) occurs frequently in the intensive care unit due to a variety of conditions, including septic shock. It is clinically important to identify high-risk patients for AKI in advance and to intervene appropriately. In this study, we developed a model for real-time prediction of AKI onset and its rationale visualization using a one-dimensional convolutional neural network (CNN) and verified its accuracy. As a result, the model was able to predict the onset of AKI with high accuracy, and the basis for the prediction was clinically valid. Furthermore, we developed methods for evaluating pathological images in an unsupervised manner and quantifying uncertainty in the prediction basis in CNN.

Free Research Field

腎臓内科

Academic Significance and Societal Importance of the Research Achievements

集中治療部において高頻度に発症するAKIを高精度で予測し、その根拠をリアルタイムで予測する手法を開発した。このことから、例としてAKIアラートシステムへの応用といった有用性が示唆された。さらに、このようなモデルの不確実性を予測根拠に反映する手法を開発した。これは例として日常的に行われるモニタリングシステムへの導入など、医療現場への応用可能性が示唆される結果と考えられた。