Cross-platform Comparison of Quantitative Florbetapir Amyloid PET Using MIMneuro and Neurophet SCALE PET

Background: Quantitative amyloid PET is increasingly used in clinical practice, but the interchangeability of quantification software remains uncertain. We compared quantitative amyloid PET measurements, diagnostic performance, and platform-specific thresholds between two FDA-cleared software packages, MIMneuro and Neurophet SCALE PET, using clinical Florbetapir 18 F-PET/CT. Patients and Methods: In this retrospective single-center study, 95 consecutive patients who underwent Florbetapir 18 F-PET/CT were independently analyzed using MIMneuro and Neurophet SCALE PET. Quantitative measures included standardized uptake value ratios (SUVRs) and global Centiloid values. Cross-platform agreement was evaluated using Pearson correlation, Bland–Altman analysis, and Cohen κ statistics. Receiver operating characteristic analysis was used to determine platform-specific thresholds for predicting visual amyloid positivity. Results: Regional SUVRs showed moderate correlations ( r = 0.414 to 0.674; all P < 0.001), with the strongest correlation observed for global Centiloid values ( r = 0.684). Neurophet yielded significantly higher global SUVR values than MIMneuro (mean bias, −0.20; P < 0.001), whereas the difference in Centiloid values was not significant. The optimal thresholds for predicting visual amyloid positivity were 1.10 versus 1.40 for global SUVR and 20.7 versus 34.4 for global Centiloids for MIMneuro and Neurophet, respectively. Cross-platform agreement was moderate using SUVR (κ = 0.42) and improved with Centiloids (κ = 0.60). Conclusions: Although MIMneuro and Neurophet SCALE PET showed moderate agreement, their quantitative measurements were not directly interchangeable at clinically relevant thresholds. These findings suggest that platform-specific quantitative thresholds may be needed when amyloid PET is used for clinical interpretation, longitudinal follow-up, or selection of patients for antiamyloid therapy.