Fast Count-Regulated OSEM Reconstruction With Adaptive Resolution Recovery

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Computational Science

Background:

• Ordered subsets expectation maximization (OSEM) accelerates tomographic reconstruction but faces trade-offs between speed and image quality.
• High subset numbers in OSEM can increase noise and quantitative inaccuracy in low-activity regions.
• Low subset numbers in OSEM lead to long reconstruction times and unrecovered details in high-activity regions.

Purpose of the Study:

• Introduce count-regulated OSEM (CROSEM) to locally adapt the effective number of subsets based on voxel photon counts.
• Evaluate CROSEM's performance in multi-pinhole SPECT simulations and in vivo imaging.
• Improve tomographic reconstruction by balancing noise and resolution across different activity levels.

Main Methods:

• Developed CROSEM to dynamically adjust the number of subsets per voxel.
• Tested CROSEM using multi-pinhole SPECT simulations.
• Validated CROSEM with in vivo imaging experiments.

Main Results:

• CROSEM achieved acceleration factors close to the maximum (128) in high-activity regions.
• Maintained quantitative accuracy in low-activity regions comparable to maximum likelihood expectation maximization (MLEM).
• Demonstrated lower noise in low-activity regions compared to MLEM at equal cold-lesion contrast in high-activity regions.

Conclusions:

• CROSEM offers a fast and stable alternative to OSEM for tomographic reconstruction.
• Effectively prevents excessive image noise and quantitative errors in low-activity regions.
• Achieves high-resolution recovery in structures with high activity uptake.