YRT-PET: An Open-Source GPU-accelerated Image Reconstruction Engine for Positron Emission Tomography.

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Positron emission tomography (PET) image reconstruction relies on accurate scanner models, often proprietary, limiting research.
• Existing open-source tools lack a unified engine for reusable code, speed, and extensibility.

Purpose of the Study:

• Introduce YRT-PET, an open-source toolkit for flexible, reproducible, and fast PET image reconstruction.
• Enable interoperability and extensibility for advanced algorithm development and scanner system research.

Main Methods:

• Developed YRT-PET in C++ with CUDA-enabled GPU acceleration and Python bindings.
• Implemented a plugin system for multi-scanner support and incorporated features like time-of-flight and motion correction.

Main Results:

• Evaluated YRT-PET on two scanners across dynamic imaging, motion correction, deep image prior, and limited-angle geometry.
• Demonstrated good agreement in image quality and correction effectiveness compared to existing tools.

Conclusions:

• YRT-PET provides a versatile, high-performance open-source solution for PET image reconstruction research.
• The toolkit facilitates the development of novel reconstruction algorithms and new PET scanner systems.