Prediction of Intracardiac Pressure using Strain Echocardiography and Artifi

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K22504
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: 0403:Biomedical engineering and related fields
• Research Institution: Juntendo University
• Principal Investigator: Kagiyama Nobuyuki 順天堂大学, 保健医療学部, 准教授 (20722010)
• Project Period (FY): 2020-09-11 – 2024-03-31
• Keywords: 心エコー / 心臓超音波検査 / 人工知能 / 機械学習 / 説明可能な機械学習 / 血行動態 / 心腔内圧

Outline of Final Research Achievements

In this study, we conducted a multicenter collaborative research project in which approximately 1,000 cases of right heart catheterization and echocardiographic images were collected for strain analysis. Using machine learning, we developed a program to estimate intracardiac pressure. The traditional guideline-recommended algorithm was unable to estimate left ventricular pressure in 42.5% of patients. However, our algorithm, which employs explainable artificial intelligence (XAI) techniques, was able to estimate pressure in all cases. The area under the receiver operating characteristic curve (AUC) for external validation data was 0.844 (95% CI 0.793-0.894), demonstrating a significant improvement over the guideline-recommended algorithm (p=0.016). These findings were presented at a conference and have also been submitted for publication.

Free Research Field

循環器内科学

Academic Significance and Societal Importance of the Research Achievements

本研究によって、今までのガイドラインで推奨されていた心内圧推定プログラムよりも有意に精度がよく、また症例ごとの推定理由も説明可能なプログラムを作成することができた。さらにこれをWebページとして使用可能としたことで、実臨床でも使用可能となった。この意義は非常に大きく、今後臨床において広く使用される可能性がある。またこのような研究を通じて、説明可能な機械学習プログラムを臨床で使用する際のひとつの成功事例とすることができた。