Ultra fast symmetry and SIMD-based projection-backprojection (SSP) algorithm for 3-D PET image reconstruction

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Recent advancements in Positron Emission Tomography (PET) hardware, such as the High-Resolution Research Tomograph (HRRT), offer enhanced resolution and sensitivity.
• High-resolution PET scanners generate massive datasets (over 4.5 x 10^9 coincidence lines), posing significant challenges for image reconstruction.
• Current image reconstruction times for high-resolution PET data can take many hours, hindering applications like dynamic imaging.

Purpose of the Study:

• To develop an efficient algorithm for accelerating image reconstruction in high-resolution PET scanners.
• To overcome the data handling bottleneck in processing large datasets from advanced PET systems.
• To enable practical list-mode-based dynamic imaging by drastically reducing reconstruction times.

Main Methods:

• Developed a novel algorithm, the symmetry and SIMD-based projection-backprojection (SSP) technique.
• Leveraged symmetry properties of projection and backprojection processes within the 3-D OSEM algorithm.
• Integrated the single-instruction multiple-data (SIMD) technique to enhance computational efficiency.

Main Results:

• The SSP algorithm effectively reduces image reconstruction time for high-resolution PET data.
• Reconstruction times were decreased from many hours to just a few minutes.
• Demonstrated the application of the SSP technique to the HRRT's OSEM algorithm.

Conclusions:

• The SSP algorithm offers a viable solution to the computational challenges of high-resolution PET image reconstruction.
• This acceleration makes advanced PET applications, including dynamic imaging, more feasible.
• The developed technique represents a significant step forward in PET scanner performance and data analysis.