Comparison between MRI-based attenuation correction methods for brain PET in dementia patients.

Area of Science:

• Neuroimaging
• Medical Physics
• Radiochemistry

Background:

• Hybrid PET/MRI offers advanced brain structure and function analysis.
• Accurate attenuation correction (AC) is crucial for PET data reconstruction, especially considering bone density.
• Existing MRI-based AC methods lack systematic comparison for brain PET studies.

Purpose of the Study:

• To systematically compare the performance of novel MRI-based attenuation correction (AC) methods.
• To evaluate AC methods in the context of [18F]-Fluorodeoxyglucose-PET/MRI for neurodegenerative dementias like Alzheimer's disease (AD).
• To assess accuracy in quantifying glucose metabolism in specific brain regions affected by AD.

Main Methods:

• Four novel methods (atlas-based, segmentation, hybrid) were developed for μ-map generation from MRI data.
• Dixon-based and UTE-based vendor methods were included for comparison.
• Performance evaluated by tissue identification accuracy, quantitative PET accuracy, and diagnostic precision for hypometabolic areas in 15 AD patients.

Main Results:

• Novel methods achieved quantitative regional errors within ±5%, outperforming vendor methods (-20% to -10%).
• Precision in identifying hypometabolic areas improved from 62.9-79.5% (vendor methods) to 87.5-93.3% (novel methods).
• The four novel methods demonstrated similar, high performance levels.

Conclusions:

• Vendor-provided AC methods (Dixon, UTE) are inferior for brain PET/MRI attenuation correction.
• Recently proposed atlas-based, template-based, and segmentation-based AC methods show comparable and superior performance.
• Accurate AC is vital for reliable quantitative analysis in neurodegenerative disease PET/MRI studies.