rPOP

rPOP performs MRI-free preprocessing and quantification of amyloid-PET scans to enable standardized measurement of amyloid burden (Centiloid) without requiring structural MRI.

Key Features:

• MRI-Free Processing: rPOP replaces reliance on structural MRI (sMRI) by operating on PET-only data for amyloid quantification.
• Compatibility with FDA-Approved Radiotracers: Supports amyloid-PET acquired with ^18F-florbetapir/FBP, ^18F-florbetaben/FBB, and ^18F-flutemetamol/FLUTE.
• Spatial Normalization with Non-linear Warping: Performs spatial normalization using non-linear warping to weighted PET templates.
• Data-Driven Differential Smoothing: Applies differential smoothing derived from the PET data to standardize image resolution across scans.
• Flexible Image Origin Handling: Provides an option to automatically reset image origin to the center of mass or proceed without modification.
• Validation and Performance: Validated on 2,258 amyloid-PET scans from IDEAS-BHR and ADNI with accurate warping on first attempt in 98.9% of cases and manual intervention rescuing most failures.
• High Concordance with Established Methods: rPOP-based amyloid status shows concordance with local visual reads and MRI-pipeline status (Cohen's k = 0.72 for IDEAS-BHR; k = 0.88 for ADNI) and a strong linear relationship with MRI-based Centiloid values (R² = 0.95).
• Implementation: Implemented in MATLAB.

Scientific Applications:

• Alzheimer's disease and dementia research: Preprocessing and quantification of amyloid-PET for studies where amyloid deposition is a key biomarker.
• Multi-site dataset harmonization and analysis: Comparison and merging of heterogeneous amyloid-PET datasets and clinical-trial data when structural MRI is unavailable.

Methodology:

Implemented in MATLAB, the pipeline performs non-linear warping-based spatial normalization to weighted PET templates followed by data-driven differential smoothing, with an optional automatic origin reset to the image center of mass.