Artificial Intelligence for understanding kidney disease

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K08725
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 53040:Nephrology-related
• Research Institution: Osaka University
• Principal Investigator: Matsui Isao 大阪大学, 医学系研究科, 講師 (60456986)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 腎生検画像 / 人工知能

Outline of Final Research Achievements

A database of renal biopsy images of 5002 patients who underwent renal biopsy between 2014 and 2018 at 24 institutions in Japan was created. We developed several artificial intelligence (AI) models for renal biopsy image diagnosis. Although the accuracy of detection of crescents was not sufficient, we constructed a model that generally reflected the histological findings made at each facility. In addition, it was possible to detect characteristics of diabetic nephropathy in patients with a history of diabetes but not diagnosed as diabetic nephropathy. Unsupervised learning identified slide format and the facility as the major factors in image discrimination.

Free Research Field

腎臓内科学

Academic Significance and Societal Importance of the Research Achievements

本研究では、人工知能を用いて腎生検画像診断が可能である事を示した。また、糖尿病歴を有するが糖尿病性腎症と診断されていない症例に糖尿病性腎症の特徴を見出すことなどが可能になった。腎病理診断は腎病変を詳しく評価するために必須であるが、その診断一致率については改善余地があるとされている。AIを用いて診断の均てん化を図ることにより、よりよい腎疾患治療構築が可能になると考えられる。