Bourgeat Pierrick, Fripp Jurgen, Lebrat Leo, Xia Ying, Feizpour Azadeh, Cox Timothy, Zisis Georgios, Gillman Ashley, Goyal Manu, Tosun Duygu, Benzinger Tammie, LaMontagne Pamela, Breakspear Michael, Lupton Michelle, Short Cathy, Adam Robert, Robertson Joanne, Sperling Reisa, O'Bryant Sid, Johnson Sterling, Jack Clifford, Schwarz Christopher, Park Denise C, Barkhof Frederik, Farrar Gill, Bollack Ariane, Collij Lyduine, Landau Susan, Koeppe Robert, Morris John, Weiner Michael, Villemagne Victor, Masters Colin, Rowe Christopher, Doré Vincent
CSIRO.
Queensland University of Technology.
Res Sq. 2025 Jul 8:rs.3.rs-7015694. doi: 10.21203/rs.3.rs-7015694/v1.
The Centiloid scale is the standard for Amyloid PET quantification, widely used in research, clinical settings, and trial stratification. However, variability between tracers and scanners remains a challenge. This study introduces DeepSUVR, a deep learning method to correct Centiloid quantification, by penalising implausible longitudinal trajectories during training. The model was trained using data from 2,098 participants (6,762 PET scans) in AIBL/ADNI and validated using 15,806 PET scans from 10,543 participants across 10 external datasets. DeepSUVR increased correlation between tracers, and reduced variability in the -negatives. It showed the strongest association with cognition, highest AUC against visual reads and best longitudinal consistency between studies. DeepSUVR also increased the effect size for detecting lower Centiloid increase per year in the A4 study. DeepSUVR advances PET quantification, outperforming standard approaches, which is particularly important for consistent decision making and to detect subtle and early changes in clinical interventions.
百分制量表是淀粉样蛋白PET定量的标准，广泛应用于研究、临床环境和试验分层。然而，示踪剂和扫描仪之间的变异性仍然是一个挑战。本研究引入了DeepSUVR，这是一种深度学习方法，通过在训练期间惩罚不合理的纵向轨迹来校正百分制定量。该模型使用来自AIBL/ADNI的2098名参与者（6762次PET扫描）的数据进行训练，并使用来自10个外部数据集的10543名参与者的15806次PET扫描进行验证。DeepSUVR增加了示踪剂之间的相关性，并降低了阴性结果的变异性。它显示出与认知的最强关联、针对视觉读数的最高AUC以及研究之间最佳的纵向一致性。DeepSUVR还增加了A4研究中每年检测较低百分制增加的效应量。DeepSUVR推动了PET定量，优于标准方法，这对于一致的决策以及检测临床干预中的细微和早期变化尤为重要。
