NiftyPET: a High-throughput Software Platform for High Quantitative Accuracy and Precision PET Imaging and Analysis.

Area of Science:

• Medical Imaging
• Computational Science
• Nuclear Medicine

Background:

• Accurate Positron Emission Tomography (PET) image reconstruction is crucial for quantitative analysis.
• Existing platforms may lack scalability, high-throughput capabilities, or precise modeling for advanced scanners.
• Need for integrated software for the entire PET image processing pipeline.

Purpose of the Study:

• To present a standalone, scalable, high-throughput software platform for PET image reconstruction and analysis.
• To achieve high accuracy and precision in quantitative imaging through high-fidelity modeling of acquisition processes.
• To facilitate the development of new reconstruction and analysis algorithms.

Main Methods:

• Developed a Python-based platform (NiftyPET) utilizing parallel computing for core routines.
• Implemented a comprehensive pipeline including list-mode processing, attenuation correction, normalization, exact forward/back projection, scatter and randoms estimation, and partial volume correction.
• Utilized span-1 ray tracing for accurate modeling of true, random, and scatter events.

Main Results:

• Demonstrated the platform's utility with an amyloid brain scan, executing all processing within a unified Python environment.
• Achieved high accuracy in quantitative imaging, particularly for large axial field-of-view scanners.
• Provided uncertainty estimation for image-derived statistics, aiding longitudinal study analysis.

Conclusions:

• The NiftyPET platform offers a scalable, high-throughput solution for accurate PET image reconstruction and analysis.
• Its high-fidelity modeling and integrated processing pipeline enhance quantitative imaging capabilities.
• The open-source availability and support for algorithm development promote further advancements in PET research.