Normalization in PET group comparison studies--the importance of a valid reference region

Area of Science:

• Neuroimaging
• Cerebrovascular physiology
• Medical physics

Background:

• Positron emission tomography (PET) studies often normalize cerebral blood flow (CBF) to the global mean to reduce interindividual variation.
• This normalization method assumes no significant group differences in global CBF, which can lead to biased physiological interpretations if violated.

Purpose of the Study:

• To investigate the impact of global mean normalization on CBF measurements in aging and hepatic encephalopathy (HE).
• To evaluate alternative normalization strategies for more accurate CBF assessment.

Main Methods:

• Quantitative [15O]H2O PET scans were acquired in 45 healthy subjects and 14 HE patients.
• CBF was analyzed using volume-of-interest and voxel-based statistics.
• Comparisons included absolute CBF, and CBF normalized to gray matter (GM) and white matter (WM) means, alongside simulation experiments.

Main Results:

• In healthy aging, CBF remained unchanged in WM and central regions, but showed artifactual increases with GM mean normalization.
• Similar artifactual increases in CBF were observed in HE patients and simulation experiments when using global mean normalization.
• Normalization to the WM mean yielded less biased results.

Conclusions:

• Global mean normalization can introduce spurious findings in CBF studies, particularly when subtle group differences in global CBF exist.
• Previous studies using global mean normalization may have led to inaccurate physiological interpretations.
• Normalization to the central white matter (WM) is a more robust method for analyzing CBF in aging and HE, and potentially other conditions.