Accuracy of 3-dimensional reconstruction algorithms for the high-resolution research tomograph

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Neuroscience

Background:

• Iterative reconstruction algorithms are standard for High-Resolution Research Tomograph (HRRT) brain PET scans.
• Existing methods like ordinary Poisson OSEM (OP-OSEM) exhibit bias in short-duration frames.
• Novel algorithms aim to improve quantitative accuracy and reduce bias.

Purpose of the Study:

• To evaluate the quantitative accuracy of novel iterative reconstruction algorithms for HRRT brain PET.
• Compare shifted Poisson OSEM (SP-OSEM) and ordered-subsets weighted least squares (OSWLS) against 3D filtered backprojection (3D-FBP).

Main Methods:

• Implemented 3D iterative reconstruction methods: OP-OSEM, SP-OSEM, and OSWLS on the HRRT.
• Conducted quantitative evaluations using phantom studies and a human brain study (n=5).

Main Results:

• OSWLS demonstrated low coefficient of variation and good agreement with 3D-FBP for gray matter-to-white matter contrast ratios.
• SP-OSEM and OP-OSEM showed fewer errors in contrast ratios across varying frame durations compared to other iterative methods.
• OSWLS exhibited minimal artifacts and strong correlation with 3D-FBP for parametric imaging.

Conclusions:

• OP-OSEM, SP-OSEM, and OSWLS performed well in phantom studies.
• OSWLS showed superior results for parametric analysis in clinical studies.
• OSWLS is recommended for quantitative HRRT brain PET imaging.