Simulation-based partial volume correction for dopaminergic PET imaging: Impact of segmentation accuracy

Area of Science:

• Neuroimaging
• Nuclear Medicine
• Medical Physics

Background:

• Quantitative analysis in Positron Emission Tomography (PET) relies on accurate partial volume correction (PVC).
• Brain PET imaging, particularly with (18)F-FDOPA, necessitates robust PVC methods to overcome partial volume effects.
• The accuracy of PVC is significantly influenced by the quality of MR-based image segmentation.

Purpose of the Study:

• To evaluate PVELab, a free software tool, for partial volume correction (PVC) in (18)F-FDOPA brain-PET.
• To assess the impact of different MR-based segmentation approaches on PVC accuracy.
• To investigate the influence of PVC on quantitative metrics like the influx constant (Ki).

Main Methods:

• Four PVC algorithms (M-PVC, MG-PVC, mMG-PVC, R-PVC) were tested on simulated (18)F-FDOPA brain-PET data.
• MR image segmentation was performed using FSL, SPM, and an adapted SPML method.
• Comparisons focused on regional activity deviations, time-activity curves (TACs), and influx constant (Ki) accuracy in the occipital cortex, caudate nucleus, and putamen.

Main Results:

• MR segmentation significantly impacted PVC, with SPML showing the lowest misclassification errors in the putamen (<30%) compared to FSL (>70%).
• M-PVC accurately recovered occipital cortex activity (recovery coefficient 0.99-1.10).
• MG-PVC and mMG-PVC were more suitable for subcortical regions, and PVC generally improved Ki quantification accuracy for the caudate nucleus and putamen, except for M-PVC.

Conclusions:

• Accurate MR image segmentation is crucial for reliable PVC in (18)F-FDOPA brain-PET using the PVELab framework.
• The choice of PVC algorithm and segmentation method should be tailored to the specific anatomical region of interest.
• While PVC enhances accuracy, it may reduce precision, potentially biasing influx constant (Ki) values and impacting group comparison studies.