Development of an AI System That Uses the SUVmax of FDG-PET/CT as a Lesion Identifier
FDG-PET/CT is an imaging method useful for cancer diagnosis. After the examination, specialists create a report, which the primary care physician reads to develop a treatment plan. This report includes the intensity of signals from areas such as cancer (SUVmax), recorded as numerical values. By utilizing these values, it is possible to automatically identify cancer locations from a large database of past images and reports (1). AI development requires large datasets for training, and this method is considered useful for efficiently creating such datasets. This study aims to establish and validate an analytical method for this purpose.
Development Section in Charge
AI Development for Diagnostic Imaging Assistant Section
Principal Investigator
HIRATA Kenji, Associate Professor, Department of Diagnostic Imaging, Hokkaido University Graduate School of Medicine
University-Affiliated Researchers
OGAWA Takahiro, Professor, Laboratory of Media Dynamics, Graduate School of Information Science and Technology
ENDO Hiroki, PhD Student, Department of Diagnostic Imaging, Hokkaido University Graduate School of Medicine
TAKENAKA Junki, Clinical Fellow (PhD Student), Department of Diagnostic Imaging, Hokkaido University Graduate School of Medicine
Non-University-Affiliated Researcher
MIYAKE Mototaka, National Cancer Center Hospital, Department of Diagnostic Radiology
External Funding
JSPS KAKENHI Grant-in-Aid for Scientific Research (C) (2023–2025)