Validation of a spatial normalization method using a principal component derived adaptive template for...

Area of Science:

• Nuclear Medicine
• Neuroimaging
• Biomedical Engineering

Background:

• Amyloid-β positron emission tomography (PET) quantification is crucial for diagnosing Alzheimer's disease.
• Spatial normalization of PET images is essential for accurate quantification but can be prone to bias.
• Existing methods often rely on magnetic resonance (MR) imaging, which may not always be available.

Purpose of the Study:

• To evaluate the efficacy of a principal component analysis (PCA) approach for spatial normalization of [18F]florbetaben PET data.
• To compare PCA-based normalization with conventional MR imaging-driven spatial normalization using SPM12.
• To determine if PCA can provide accurate and robust registration without requiring MR images.

Main Methods:

• PCA was applied to [18F]florbetaben PET data from 132 subjects (70 with Alzheimer dementia, 62 controls).
• An adaptive synthetic template was generated using PCA.
• Spatial normalization results were compared against SPM12's MR-driven algorithm, analyzing standardized uptake value ratios (SUVR).

Main Results:

• The PCA-based spatial normalization method demonstrated high agreement with SPM12's MR-driven algorithm.
• Minimal differences were observed in SUVR when using the cerebellum (R² = 0.997) or pons (R² = 0.996) as reference regions.
• The PCA approach proved robust and accurate for registering [18F]florbetaben PET images.

Conclusions:

• PCA provides a reliable method for spatial normalization of [18F]florbetaben PET images.
• This technique eliminates the need for MR imaging, offering a valuable alternative for quantification.
• The PCA-based approach holds significant potential for improving clinical diagnostics and research in Alzheimer's disease.