The Development of an Automatic Amyloid PET Quantification (AmPQ) Software with MR-based and MR-free Spatial Normalization

Purpose: This study aimed to develop and evaluate an in-house software tool, Amyloid PET Quantification (AmPQ), for automatic Centiloid calculation. AmPQ supports both MR-based and MR-free adaptive PET template-based spatial normalization. Methods: The Centiloid quantification procedure was developed according to the standard GAAIN pipeline, with two-tier validation involving PiB and three [¹⁸F]-labeled amyloid PET tracers: Florbetapir (FBP), Flutemetamol (FMM), and Florbetaben (FBB). Additional independent validation was performed using three external cohorts: T-ADNI (n = 92), ADNI (n = 50), and AIBL (n = 50). Simple linear regression was used for quality control (QC) and to assess the performance of the MR-free spatial normalization. Results: For level-1 replication, both MR-based (slope = 1.001; intercept = 0.0477; R² = 0.9994) and MR-free (slope = 0.9890; intercept = 0.6073; R² = 0.9890) approaches satisfied the QC criteria. In level-2 calibration, all tracers exceeded the R² threshold of 0.70: FBP (MR-based = 0.895; MR-free = 0.868), FMM (MR-based = 0.960; MR-free = 0.947), and FBB (MR-based = 0.953; MR-free = 0.941). Centiloid values derived from MR-free and MR-based methods were highly correlated (R² >0.98 in GAAIN datasets and R² >0.96 in all external cohorts). Conclusion: AmPQ demonstrates robust replicability and high concordance with the standard Centiloid pipeline. The high correlation between MR-free and MR-based spatial normalization supports the feasibility of MR-free quantification in large-scale or resource-limited settings.