• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2023 Fiscal Year Final Research Report

The Development of an Automatic Diagnosis System for Coronary Artery Atherosclerosis Using Deep Learning and Its Clinical Application

Research Project

  • PDF
Project/Area Number 21K08065
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Review Section Basic Section 53020:Cardiology-related
Research InstitutionKansai Medical University

Principal Investigator

Fujii Kenichi  関西医科大学, 医学部, 講師 (90434943)

Co-Investigator(Kenkyū-buntansha) 廣田 誠一  兵庫医科大学, 医学部, 教授 (50218856)
塩島 一朗  関西医科大学, 医学部, 教授 (90376377)
植田 大樹  大阪公立大学, 大学院医学研究科, 前期臨床研究医 (90779480)
Project Period (FY) 2021-04-01 – 2024-03-31
Keywords人工知能 / 血管内イメージング / 冠動脈疾患 / 動脈硬化
Outline of Final Research Achievements

This study aimed to automate the diagnosis of coronary artery tissue characteristics from OFDI images using deep learning. The subjects comprised 109 coronary arteries obtained from 45 autopsied hearts. After conducting ex vivo OFDI examinations, a total of 1,103 matched sets of pathological section images and OFDI images with atherosclerotic changes were obtained. These images were divided into three datasets: training, validation, and testing. Subsequently, a deep learning model based on PSPNet was constructed using the training set. Upon evaluating the model, an average F-score of 0.6255 and an IoU of 0.488 were achieved in the validation dataset, while in the test dataset, an average F-score of 0.6577 and an IoU of 0.5166 were attained.

Free Research Field

冠動脈粥状硬化

Academic Significance and Societal Importance of the Research Achievements

本研究の結果は、目的であるOFDI画像からの冠動脈組織性状の自動診断が一定の精度で可能であることを示しています。冠動脈カテーテルインターベンションの際に医師が行う冠動脈硬化の組織性状診断は定性評価であるため、読影者間でのバラツキが大きく、精度は経験値に依存します。今回我々が開発した人工知能モデルを用いることで、経験の浅い医師や非専門医がOFDI画像を読影しても経験豊富な専門医と同等もしくはそれ以上の精度で診断することが可能になると考えます。また、本モデルを用いることで、長年解明されてこなかった冠動脈粥状硬化形成のメカニズム解明につながると考えます。

URL: 

Published: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi