Exploring Centiloid Robustness: Impact of Sample Size and Image Resolution on Centiloid Conversion Accuracy.

Area of Science:

• Nuclear Medicine
• Neuroimaging
• Alzheimer Disease Research

Background:

• Centiloids are widely used for quantifying amyloid-β (Aβ) positron emission tomography (PET) in clinical trials and settings.
• Accurate measurement error characterization is essential for reliable Aβ PET quantification.
• Potential sources of error include calibration dataset sampling and PET image resolution.

Purpose of the Study:

• To examine the impact of calibration dataset size and PET image resolution on Centiloid measurement accuracy.
• To evaluate the SUV ratio-to-Centiloid conversion equation's reliability under varying conditions.
• To identify optimal parameters for minimizing Centiloid quantification errors in Aβ PET analysis.

Main Methods:

• Analyzed [11C]PiB PET scans from 200 participants with clinically diagnosed Alzheimer disease (cAD) and 114 Aβ-negative individuals.
• Compared Centiloid conversion equations derived from subsamples (10-30 Aβ-negative, 15-50 cAD participants) and assessed PET image resolution effects (high, medium, low) on [11C]PiB, [18F]florbetaben, and [18F]florbetapir scans.
• Utilized standard Centiloid and MRI-based pipelines, alongside rPOP and CapAIBL PET-only pipelines.

Main Results:

• In the smallest calibration sample (15 cAD, 10 Aβ-negative), conversion errors were 1.7 and 3.4 Centiloids at 25 and 100 Centiloids, respectively.
• Increasing calibration participants to 50 cAD and 30 Aβ-negative reduced errors to 1.0 and 2.0 Centiloids.
• Lower PET image resolution systematically decreased Centiloid values, particularly in highly positive scans (e.g., 100 Centiloids at high resolution became 84.9 Centiloids at low resolution).

Conclusions:

• Accurate SUV ratio-to-Centiloid conversion for Level 2 analyses is achievable with minimally required datasets.
• Increasing cAD participants in calibration datasets effectively reduces Centiloid error at higher values.
• PET image resolution significantly impacts Centiloid values in highly positive scans and requires careful consideration during data interpretation.