SimPET

SimPET generates realistic brain positron emission tomography (PET) data via Monte Carlo simulation for validation and methodological studies.

Key Features:

• Monte Carlo simulation (SimSET): Performs Monte Carlo PET simulations using the SimSET package to produce synthetic PET images.
• Phantom and map-based generation: Uses phantom generation and activity and attenuation maps derived from PET/CT and MRI of 25 healthy subjects scanned with a GE Discovery ST.
• Tomographic reconstruction (STIR): Reconstructs simulated PET data using the STIR framework.
• Cross-scanner validation: Validates simulated images against acquired FDG-PET data from GE Advance NXi, GE Discovery ST, and Siemens Biograph mCT.
• Quantitative assessment: Reports voxel-wise agreement with errors below 10% for 98.09% (GE Discovery ST), 95.09% (GE Advance NXi), and 91.35% (Siemens Biograph mCT) of voxels.
• Statistical comparisons: Assesses Bland-Altman differences and performs voxel-wise comparisons using Statistical Parametric Mapping (SPM) to identify regional discrepancies.

Scientific Applications:

• Amyloid PET studies: Provides synthetic PET data to support development and testing of amyloid PET analysis methods.
• Ground-truth pathological models: Enables creation of ground-truth pathological patient models for validation of algorithms.
• Validation and benchmarking: Facilitates validation and benchmarking of PET reconstruction and analysis methods across GE Advance NXi, GE Discovery ST, and Siemens Biograph mCT scanners.
• Cross-scanner database conversion: Can be used to investigate conversion of control subject databases across scanners, with SPM revealing notable regional differences that require further investigation.

Methodology:

Activity and attenuation maps from PET/CT and MRI of 25 healthy subjects (GE Discovery ST) are used to generate synthetic PET data for GE Advance NXi, GE Discovery ST, and Siemens Biograph mCT via Monte Carlo simulations with SimSET, reconstructed with STIR, and validated using Bland-Altman and voxel-wise SPM comparisons.