Comparison of MRI based and PET template based approaches in the quantitative analysis of amyloid imaging with PIB-PET

Rationale: [(11)C]Pittsburgh compound-B (PIB) has been the most widely used positron emission tomography (PET) imaging agent for brain amyloid. Several longitudinal studies evaluating the progression of Alzheimer's disease (AD), and numerous therapeutic intervention studies are underway using [(11)C]PIB PET as an AD biomarker. Quantitative analysis of [(11)C]PIB data requires the definition of regional volumes of interest. This investigation systematically compared two data analysis routes both using a probabilistic brain atlas with 11 bilateral regions. Route 1 used individually segmented structural magnetic resonance images (MRI) for each subject while Route 2 used a standardised [(11)C]PIB PET template.

Methods: A total of 54 subjects, 20 with probable Alzheimer's disease (AD), 14 with amnestic Mild Cognitive Impairment (MCI) and 20 age-matched healthy controls, were scanned at two imaging centres either in London (UK) or in Turku (Finland). For all subjects structural volumetric MRI and [(11)C]PIB PET scans were acquired. Target-to-cerebellum ratios 40 min to 60 min post injection were used as outcome measures. Regional read outs for grey matter target regions were generated for both routes. Based on a composite neocortical, frontal, posterior cingulate, combined posterior cingulate and frontal cortical regions, scans were categorised into either 'PIB negative' (PIB-) or 'PIB positive' (PIB+) using previously reported cut-off target-to-cerebellar ratios of 1.41, 1.5 and 1.6, respectively.

Results: Target-to-cerebellum ratios were greater when defined with a [(11)C]PIB PET template than with individual MRIs for all cortical regions regardless of diagnosis. This difference was highly significant for controls (p<0.001, paired samples t-test), less significant for MCIs and borderline for ADs. Assignment of subjects to raised or normal categories was the same with both routes with a 1.6 cut-off while with lower cut off using frontal cortex, and combined frontal cortex and posterior cingulate demonstrated similar results, while posterior cingulate alone demonstrated significantly higher proportion of controls as amyloid positive by Route 2.

Conclusions: Definition of cortical grey matter regions is more accurate when individually segmented MRIs (Route 1) were used rather than a population-based PET template (Route 2). The impact of this difference depends on the grey-to-white matter contrast in the PET images; specifically seen in healthy controls with high white matter and low grey matter uptake. When classifying AD, MCI and control subjects as normal or abnormal using large cortical regions; discordance was found between the MRI and template approach for those few subjects who presented with cortex-to-cerebellum ratios very close to the pre-assigned cut-off. However, posterior cingulate alone demonstrated significant discordance in healthy controls using template based approach. This study, therefore, demonstrates that the use of a [(11)C]PIB PET template (Route 2) is adequate for clinical diagnostic purposes, while MRI based analysis (Route 1) remains more appropriate for clinical research.