Open-Source GPU Toolkit for PET Image Reconstruction
Publication Title: YRT-PET: An Open-Source GPU-Accelerated Image Reconstruction Engine for Positron Emission Tomography
Summary
- Question
- This study introduced YRT-PET, an open-source image reconstruction toolkit for positron emission tomography (PET). The researchers aimed to create a flexible, high-speed software compatible with various PET scanners, offering features like motion correction, time-of-flight (TOF) integration, and GPU acceleration.
- Why it Matters
- PET imaging is a critical tool in medical fields such as neuroscience, cardiology, and oncology, enabling detailed visualization of biological processes. Current PET reconstruction engines are often proprietary and limited to specific scanner models, hindering innovation and comparative studies. YRT-PET addresses these challenges by providing an open-source solution that supports advanced imaging techniques, facilitates algorithm development, and enhances accessibility for researchers and developers. This could improve imaging accuracy and accelerate progress in PET scanner technology and diagnostic capabilities.
- Methods
- YRT-PET was developed using C++ and CUDA for GPU acceleration, with optional Python bindings for integration with external tools. It supports list-mode and histogram data formats, allowing flexibility in input data. The software includes corrections for motion, scatter, and randoms, and employs the ordered-subsets expectation-maximization (OS-EM) algorithm for image reconstruction. Evaluations were conducted using both commercial scanners and simulated data, comparing results to existing software.
- Key Findings
- YRT-PET demonstrated strong agreement with proprietary reconstruction software, producing comparable image quality and quantitative accuracy in dynamic and motion-corrected imaging. It showed high structural similarity (95.3%) between motion-corrected and static images and reduced noise in low-count data when paired with advanced algorithms like Deep Image Prior. Additionally, YRT-PET outperformed other open-source tools in reconstruction speed due to optimized GPU utilization.
- Implications
- The findings validate YRT-PET as a reliable and efficient alternative to proprietary software, enabling researchers to implement custom algorithms and explore novel PET imaging approaches. Its versatility could advance PET scanner development and improve diagnostic imaging in clinical and research settings. The toolkit’s open-source nature also encourages collaboration and innovation across institutions.
- Next Steps
- Future work will focus on improving scatter correction methods, optimizing GPU implementations, and adding support for sinogram-based reconstruction and non-rigid motion correction. The researchers also plan to expand compatibility with diverse scanner formats and enhance scalability for large datasets.
- Funding Information
- This research was supported by the National Institutes of Health (awards P41EB022544, R01EB035093, U01EB027003, and R01AG076153). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additional support was provided by the European Union’s Horizon Europe research and innovation program under grant agreement No 101099896 (PetVision project) and the Slovenian Research and Innovation Agency (Research core funding Nos. P1-0389 and P1-0135).
Full Citation
Najmaoui Y, Chemli Y, Toussaint M, Petibon Y, Marty B, Fontaine K, Gallezot J, Razdevsek G, Orehar M, Dhaynaut M, Guehl N, Dolenec R, Pestotnik R, Johnson K, Ouyang J, Normandin M, Tetrault M, Lecomte R, Fakhri G, Marin T. YRT-PET: An Open-Source GPU-Accelerated Image Reconstruction Engine for Positron Emission Tomography. IEEE Transactions On Radiation And Plasma Medical Sciences 2025, 10: 535-546. PMID: 41424471, PMCID: PMC12714321, DOI: 10.1109/trpms.2025.3619872.
This AI-assisted summary has been reviewed and approved by at least one of the study's authors to ensure it accurately reflects the research.
Authors
Yassir Najmaoui
First AuthorThibault Marin, PhD
Last AuthorAssistant Professor of Radiology and Biomedical Imaging
Additional Yale School of Medicine Authors
Other Authors
Research Themes
Concepts
- Open-source;
- Image reconstruction;
- Positron emission tomography image reconstruction;
- Reconstruction algorithm;
- Advanced image reconstruction algorithms;
- Open-source toolkit;
- Development of advanced algorithms;
- Time-of-flight;
- Scanner geometry;
- Novel reconstruction algorithm;
- Image reconstruction algorithm;
- Motion correction;
- Reusable code;
- GPU acceleration;
- Time-of-flight information;
- Python bindings;
- Plugin system;
- Software toolkit;
- Meaningful images;
- Data formats;
- Fast implementation;
- Rigid motion correction;
- Advanced algorithms;
- Point spread function model;
- Proprietary software