Efficient Training Data Generation for Automatic Image Quality Assessment of FDG-PET Images by Machine Learning

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K08091
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 52040:Radiological sciences-related
• Research Institution: Kyoto College of Medical Science
• Principal Investigator: MATSUMOTO Keiichi 京都医療科学大学, 医療科学部, 准教授 (60393344)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: FDG-PET / 機械学習 / 画質自動判定

Outline of Final Research Achievements

The purpose of this study was to disseminate image quality criteria for [18F]-FDG whole-body PET imaging determined by the academic society in Japan. In this study, we developed a system to automatically determine the image quality of PET images and also investigated the characteristics of the automatic PET image evaluation system (Phantom Analysis Toolkit; PAT) using several numerical phantoms.
The results of the PAT analysis differed depending on the position of the numerical phantom in the field of view. We were able to collect a lot of training data for machine learning, which is necessary to automatically determine the image quality of PET images. It was thought that the image quality of PET images could be automatically determined using image quality criteria that are consistent with those of Western academic societies.

Free Research Field

放射線科学

Academic Significance and Societal Importance of the Research Achievements

本邦の関連学会が見直しを進めている[18F]-FDGを用いた全身PET撮像の画質基準は、欧米諸国の基準と比較して整合性が取れているだけでなく、より高い精度と再現性で定量的指標を取得することができるため、画質基準に適合したPET画像は定量的画像バイオマーカーとして活用できると考えられる。
画質基準を迅速かつ簡便に評価することができれば、質の高い診療を行うことができるだけでなく、定量的画像バイオマーカーを用いて短期間かつ低コストで新しい治療薬や治療法の有効性を実証できる可能性が示唆された。