Advertisement

INCREASED SENSITIVITY OF AV45-PET FOR THE DETECTION OF EARLY STAGE AMYLOIDOSIS AFTER CORRECTION OF WHITE MATTER SPILL-IN EFFECTS

      Background

      Amyloid-sensitive PET is an increasingly used biomarker for the detection of cerebral amyloid pathology but its sensitivity may critically depend on the way the PET scans are analyzed. Here we explored the effect of different image processing strategies on the concordance of amyloid-PET findings with cerebrospinal fluid (CSF) Aβ42 levels; an alternative biomarker of cerebral amyloidosis which typically shows only weak to modest associations with amyloid-PET findings in early phases of amyloid accumulation.

      Methods

      We investigated the effects of 2- and 3-compartment models of partial volume correction (PVC-2/-3) and choice of reference region on the correlation between cortical AV45-PET uptake ratios (AV45-SUVR) and CSF-Aβ42 levels using data from 603 subjects enrolled in ADNI-2. Furthermore, in a subset of 152 cognitively normal subjects the ability to detect regional AV45-SUVR increases in groups with decreased CSF-Aβ42 levels was compared between the different processing approaches using voxel-wise analyses.

      Results

      When using a whole cerebellar reference region, PVC-3, which also controls for spill-in effects of white matter (WM) signal, resulted in a significantly increased correlation of AV45-SUVRs with CSF-Aβ42 levels. This effect was most pronounced for the lower range of AV45-SUVRs (Figure-1), and was not observed for the simpler PVC-2 model that only controls for CSF dilution. Using PVC-3, cognitively normal subjects within the 3rd quintile of CSF-Aβ42 values (mean = 202 pg/ml) showed significantly increased AV45-SUVR values in fronto-temporo-parietal association areas compared to subjects within the highest CSF-Aβ42 quintile (≥ 241 pg/ml), and amyloid signal further extended across the cortex in subjects within the lowest CSF-Aβ42 quintiles (Figure-2). In uncorrected data, significant AV45-SUVR increases were only detected in subjects within the two lowest CSF-Aβ42 quintiles. Use of a WM reference region increased the correlation with CSF-Aβ42 in uncorrected PET data to a similar degree as PVC-3, and these effects were non-additive.

      Conclusions

      Preprocessing techniques that account for the contamination of gray matter signal by unspecific WM binding can uncover biologically meaningful signal in AV45-PET scans that would typically be regarded as “amyloid-negative”, and thus increase their sensitivity for detecting early stage amyloidosis.
      Figure thumbnail fx1
      Figure 1The effect of PVC-3 on the relation between cortical AV45-SUVRs and CSF-Aβ42 at the low range of AV45-SUVR values
      Figure thumbnail fx2
      Figure 2Regional AV45-SUVR increases in groups of cognitively normal subjects with decreased CSF-Aβ42 levels