Catheterization difficulties and optimal catheter design for angiography using deep learning.

2021 Fiscal Year Final Research Report

• Project/Area Number: 20K22862
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: 0902:General internal medicine and related fields
• Research Institution: Hokkaido University
• Principal Investigator: Morita Ryo 北海道大学, 医学研究院, 助教 (30872626)
• Project Period (FY): 2020-09-11 – 2022-03-31
• Keywords: AI / Deep Learning / カテーテライゼーション / 難易度 / 血管 / IVR

Outline of Final Research Achievements

This study was conducted to develop artificial intelligence (AI) capable of determining the difficulty of catheter insertion into the target vessel for endovascular catheterization. AI analysis was performed based on a difficulty evaluation test for visibility using CT VR data from the celiac artery to the common hepatic artery performed by one specialist in 2020. The results showed that the overall accuracy was relatively good at 89.05% when cases were divided into difficult and non-difficult cases. In 2021, an AI analysis based on a visibility evaluation study by three IVR specialists was performed. The overall accuracy was more than 60% when discriminating between difficult and non-difficult catheter insertion cases and more than 80% when selecting non-difficult cases.

Free Research Field

Interventional Radiology

Academic Significance and Societal Importance of the Research Achievements

医療画像データを用いた画像診断の深層学習に関しては、肺結節の良悪性鑑別や脳動脈瘤の検出、肝腫瘍の鑑別など様々な報告がされ、深層学習が放射線診断専門医と同等の診断能や検出能があることも報告されている。一方、手術や血管内治療など医学的手技の難易度を、深層学習によって事前の画像データから解明するという報告はなく本研究は世界初の試みとなる。本研究で開発したカテーテライゼーションの難難易度及び適切なカテーテル判定システムは、頭頸部や脳、心臓、 骨盤、下肢など他の血管内治療領域への応用が可能である。これにより、従来経験に基づいて方針が決定されている医学的手技に科学的視点を導入する事が可能となる。