Objectives: The primary endpoint of this study is to establish a reliable SUVR cutoff threshold to distinguish patients with Alzheimer's disease (AD), excluding those with mild cognitive impairment (MCI), from normal control (NC) individuals using [18F]florapronol PET imaging and deep learning-based automated quantification software. The secondary endpoint is to evaluate whether combining partial volume correction (PVC) with SUVR analysis improves diagnostic accuracy in detecting AD.
Methods: A total of 141 participants, including 55 AD patients (excluding MCI) and 86 NC controls, were enrolled. Each participant underwent [18F]florapronol PET imaging, and SUVR values were calculated for six amyloid-prone brain regions using deep learning-based software. SUVRs were computed with and without PVC, using the cerebellar cortex as the reference region. Receiver operating characteristic (ROC) analysis identified optimal SUVR thresholds for distinguishing AD (excluding MCI) from NC and for determining visual positivity. Age-matched subgroup analyses ensured consistent diagnostic performance across different age groups.
Results: In the full cohort (n = 141), visual analysis achieved a sensitivity of 90.9% and specificity of 94.1% for distinguishing AD from NC. SUVR without PVC reached a similar sensitivity of 90.9% and specificity of 86.0% (optimal threshold > 1.26), while PVC-adjusted SUVR further improved accuracy with a sensitivity of 90.9% and specificity of 94.2% at a threshold of > 1.31. For visual positivity, SUVR without PVC achieved 92.7% sensitivity and 89.5% specificity, while PVC-adjusted SUVR improved these metrics to 96.4% sensitivity and 94.2% specificity. Age-matched analyses confirmed diagnostic consistency across different age groups. The visual analysis and the quantitative analysis using SUVR with PVC as the threshold were consistent in 134 out of 141 subjects (95.0%).
Conclusions: Automated SUVR quantification with PVC adjustment provides a reliable and objective method for distinguishing AD from NC, aligning closely with visual assessment accuracy and supporting clinical use of [18F]florapronol PET imaging for AD diagnosis. This standardized approach enhances diagnostic consistency, particularly in settings with limited access to PET specialists, and establishes robust SUVR thresholds for broader clinical application in amyloid PET imaging.
Keywords: Alzheimer’s disease; Deep learning; Partial volume correction; SUVR; [18F]florapronol.
© The Author(s), under exclusive licence to Korean Society of Nuclear Medicine 2025. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.