NiftyPET

NiftyPET enables high-throughput reconstruction and quantitative analysis of Positron Emission Tomography (PET) images, providing high-fidelity modeling of acquisition processes to achieve accurate and precise quantitative imaging for scanners including large axial field-of-view systems.

Key Features:

• High Fidelity Modeling: Employs span-1 ray tracing to model true, random, and scatter events without axial compression, improving reconstruction accuracy.
• Parallel Computing Integration: Implements core routines using parallel computing within Python for efficient data access, manipulation, and visualization at any processing stage.
• Comprehensive Processing Pipeline: Performs list-mode data processing, accurate attenuation coefficient map generation, detector normalisation, exact forward and back projection, reduced-variance random events estimation, high-accuracy 3D scatter event estimation, and voxel-based partial volume correction.
• Advanced Image Analysis: Supports region- and voxel-level analysis and provides uncertainty estimation for image-derived statistics to aid tracking of physiological changes in longitudinal studies.
• Flexible Reconstruction Environment: Allows restriction of the axial field of view to specific rings to enable faster, fully 3D reconstruction and corrections using real data.

Scientific Applications:

• Amyloid brain PET imaging: Executes end-to-end processing from MR and raw PET input to quantitative analysis, supporting neurological research and longitudinal studies.

Methodology:

Computational methods include span-1 ray tracing for true, random, and scatter modeling, list-mode data processing, attenuation coefficient map generation, detector normalisation, exact forward and back projection, reduced-variance random events estimation, high-accuracy 3D scatter estimation, voxel-based partial volume correction, region- and voxel-level analysis with uncertainty estimation, and optional restriction of axial field-of-view to rings for fully 3D reconstruction and corrections; core routines use parallel computing in Python.