Development of quantitative evaluation methods for emphysema which is not affected by CT scan condition using deep-learning based reconstruction

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K16701
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 52040:Radiological sciences-related
• Research Institution: Hiroshima University
• Principal Investigator: Fukumoto Wataru 広島大学, 医系科学研究科(医), 助教 (00726870)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 肺気腫 / CT / 人工知能 / Deep-learning

Outline of Final Research Achievements

The purpose of this study was to develop quantitative evaluation methods for emphysema score which was not affected by CT scan condition using deep-learning reconstruction.
In 2021-2022, we compared the conventional reconstruction and the deep-learning based reconstruction (DLR) for emphysema scores using clinical cases. It revealed that the the DLR reduced image noise and decreased scores, but since the true values were not determined, we could not prove which image reconstruction method was superior. Therefore, the phantom study was performed in 2023-2024. Our phantom study revealed that DLR may be the best reconstruction method for emphysema score because the image noise was reduced and the value was most accurate.

Free Research Field

CT

Academic Significance and Societal Importance of the Research Achievements

本研究は、AIを応用したCTの画像再構成法を用いることで、画像ノイズの低減が可能となり、CT装置や撮影条件等に影響されにくい肺気腫の定量評価が可能であることを明らかにした。
これは、2030年に世界の死因第3位になると予測されている慢性閉塞性肺疾患の早期発見、重症度分類、治療効果判定を行う際に有用となる可能性があり、社会的意義は非常に高い。また、近年は超高精細CTやphton-countiing CTといった新たなCTが登場しており、それらを活用するうえでも今回の我々の結果は非常に重要である。