Development of Radiation Dose Reduction Stragegy Using Deep-learning Reconstruction for Pediatric CT

2022 Fiscal Year Final Research Report

• Project/Area Number: 19K17173
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 52040:Radiological sciences-related
• Research Institution: Kumamoto University
• Principal Investigator: Nagayama Yasunori 熊本大学, 病院, 助教 (60791762)
• Project Period (FY): 2019-04-01 – 2023-03-31
• Keywords: 小児CT被ばく / 深層学習 / CT画像再構成法

Outline of Final Research Achievements

Children are known to be more radiosensitive than adults, highlighting the importance of optimizing radiation dose in CT scans. A trade-off exists between radiation dose and image quality, necessitating the incorporation of image reconstruction algorithms capable of reducing image noise and enhancing spatial/contrast resolution. In recent years, deep learning reconstruction (DLR) has emerged as an artificial intelligence-based technology for improving image quality. This research project investigated the image quality characteristics of DLR and evaluated its clinical application in reducing radiation exposure in pediatric CT scans.

Free Research Field

放射線診断学

Academic Significance and Societal Importance of the Research Achievements

本研究では、近年のAI技術の発達に伴い開発されたdeep-learning reconstruction (DLR) の小児CTにおける被ばく低減効果を示した。CT被ばくに起因する潜在的な発がんリスクの低下に寄与する成果であり、社会的意義は大きい。また、AIを活用した画像生成技術の有益な臨床応用例として、将来の研究の基礎となるという観点からも学術的意義の高い研究成果と考える。