Variance reduction with synaptic density imaging in Parkinson's disease using direct-4D PET image reconstruction.

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Neuroscience

Background:

• Dynamic Positron Emission Tomography (PET) data analysis traditionally uses indirect reconstruction (IR), involving frame-by-frame reconstruction followed by kinetic modeling.
• Direct reconstruction (DR) of parametric images from dynamic PET data offers potential for noise reduction compared to IR.
• Synaptic vesicle glycoprotein 2A (SV2A) PET imaging with [11C]UCB-J is crucial for assessing synaptic density in neurodegenerative diseases like Parkinson's disease (PD).

Purpose of the Study:

• To compare the performance of direct reconstruction (DR) versus indirect reconstruction (IR) in parametric imaging.
• To evaluate the impact of DR and IR on within-subject and between-subject variability.
• To assess noise-induced bias in parametric images generated by DR and IR methods.

Main Methods:

• The PMOLAR-1T DR method was evaluated using multiple [11C]UCB-J PET scans from Parkinson's disease patients.
• Parametric images (K1, k2, VT) were estimated using both DR and IR at full and reduced radioligand count levels (20%, 10%, 5%).
• Within-subject and between-subject variability, as well as noise-induced bias, were quantitatively compared between the two reconstruction methods.

Main Results:

• DR significantly reduced within-subject variability for K1 (79-353%) and VT (62-79%) compared to IR at lower count levels.
• DR substantially decreased between-subject variability for K1 (20-221%) and VT (45-48%) compared to IR.
• DR demonstrated minimal noise-induced bias (1% for K1, 7-18% for VT) compared to IR (3-28% for K1, 12-31% for VT).

Conclusions:

• Direct reconstruction (DR) offers superior performance over indirect reconstruction (IR) for dynamic PET parametric imaging.
• DR significantly lowers both within-subject and between-subject variability, enhancing the reliability of quantitative PET analyses.
• This study provides the first evidence that 4D direct reconstruction improves variability and bias in both within- and between-subject analyses.