Influence of cold walls on PET image quantification and volume segmentation: a phantom study.

Area of Science:

• Nuclear Medicine
• Medical Imaging Physics

Background:

• Commercially available positron emission tomography (PET) phantom inserts often feature thick plastic walls.
• These 'cold' walls can introduce partial volume effects, leading to inaccurate Standardized Uptake Value (SUV) estimations.
• This inaccuracy is a known issue in PET phantom studies for tumor simulation, quality control, and calibration.

Purpose of the Study:

• To investigate the impact of cold plastic wall thickness on 18F-fluorodeoxyglucose quantification in PET imaging.
• To assess the influence of wall thickness on image activity recovery and the performance of automatic segmentation algorithms.
• To quantify the effect of reduced wall thickness on accuracy in PET-based boundary detection.

Main Methods:

• Replicated commercial PET phantom inserts with up to 90% thinner plastic walls.
• Imaged thin- and thick-wall inserts simultaneously across various tumor-to-background ratios (TBRs).
• Compared SUV values (SUVmean, SUVmax, SUVpeak) and recovery coefficients (RC) against theoretical models; evaluated ten automatic segmentation methods against CT-derived ground truth.

Main Results:

• Thin-wall inserts yielded significantly higher SUVmean, SUVmax, and RC values (up to 25%) compared to thick-wall inserts.
• Observed substantial differences (>5%) for spheres up to 30 mm diameter and TBR up to 4.
• Thinner walls improved delineation accuracy for most segmentation methods, though quantification errors persisted.

Conclusions:

• A 90% reduction in insert wall thickness significantly impacts SUV quantification and PET boundary detection.
• Thick cold walls notably affect mean SUVs and recovery coefficients, aligning with theoretical predictions.
• Utilizing thin-wall inserts enhances segmentation algorithm accuracy, highlighting risks associated with thick-wall inserts in performance assessments.