Iterative framework for image registration and partial volume correction in brain positron emission tomography.

Area of Science:

• Medical Imaging
• Neuroscience
• Radiochemistry

Background:

• Imprecise registration between PET and MR images is a major error source in MR imaging-guided partial volume correction (MR-PVC).
• This inaccuracy affects the reliability of quantitative measurements in PET studies, particularly for neurodegenerative diseases.

Purpose of the Study:

• To introduce a novel framework, PVC-optimized registration (PoR), designed to enhance the precision of image registration for MR-PVC.
• To evaluate the effectiveness of PoR in improving the accuracy of quantitative PET measurements compared to conventional registration methods.

Main Methods:

• The PoR framework iteratively refines registration by alternating between partial volume correction (PVC) and image registration using uncorrected PET and smoothed PV-corrected images.
• The method was applied to [11C]PiB PET data from 92 Alzheimer's Disease Neuroimaging Initiative participants.
• Registration accuracy, partial volume-corrected standardized uptake value ratio (SUVR), and intra-region coefficient of variation (CoV) were compared between PoR and conventional registration.

Main Results:

• Significant registration differences, up to 2.74 mm and 3.02°, were observed between PoR and conventional methods (effect size > 0.8).
• Substantial differences in SUVR were found across the brain, with a maximum of 62.3% in the sensory motor cortex.
• PoR significantly reduced intra-region CoV throughout the brain, indicating improved quantification consistency.

Conclusions:

• PoR effectively minimizes errors stemming from imprecise registration in partial volume correction.
• This novel method offers a valuable tool for the accurate quantification of amyloid burden in PET imaging, crucial for Alzheimer's disease research.