Transition to GPU-based reconstruction for clinical organ-targeted PET scanner.

Area of Science:

• Medical Imaging
• Computational Imaging
• Nuclear Medicine

Background:

• Traditional whole-body PET/CT scans face challenges with high radiation exposure and limited spatial resolution.
• Organ-targeted PET scanners, like the Radialis technology, aim to improve specificity and reduce dose using planar detectors.
• Existing reconstruction methods, often CPU-based, can be time-consuming.

Purpose of the Study:

• To explore and implement graphics processing unit (GPU)-based techniques for efficient image reconstruction in organ-targeted PET scanners.
• To evaluate the performance and clinical utility of GPU-accelerated reconstruction for the Radialis low-dose PET system.
• To demonstrate the advantages of GPU computing in accelerating PET image reconstruction.

Main Methods:

• Applied GPU-based reconstruction algorithms to the Radialis organ-targeted PET technology.
• Transitioned from Central Processing Unit (CPU)-based Maximum Likelihood Expectation Maximization (MLEM) to GPU-based MLEM.
• Utilized standardized PET performance tests and clinical image analysis for evaluation.

Main Results:

• Achieved a tenfold overall speedup and a hundredfold improvement in iteration speed for image reconstruction using GPUs.
• Demonstrated that GPU-based reconstruction maintains diagnostic image quality comparable to CPU-based methods.
• Significantly reduced examination times in clinical applications, such as positron emission mammography, enhancing patient comfort and throughput.

Conclusions:

• GPU-based image reconstruction is a viable and efficient method for organ-targeted PET imaging.
• This advancement significantly optimizes the clinical workflow by reducing scan and reconstruction times.
• Leveraging GPU parallel processing capabilities offers substantial benefits for the future of PET imaging.