Quantification and accuracy of clinical [11C]-PiB PET/MRI: The effect of MR-based attenuation correction

Objectives The objective of this study was to measure the regional and absolute bias of the standard Dixon-Water-Fat segmentation (DWFS) (1) attenuation correction (AC) method in combined PET/MR images of the brain using [11C]-PiB for cerebral amyloid binding and evaluate the impact on the clinical reading of these images.

Methods In an ongoing dementia study 8 healthy volunteers, and 1 patient with Alzheimers disease (6 m, 3 f, age: 51-77 y, median: 68 y) underwent a simultaneous PET/MRI (mMR, Siemens) acquisition 40 min pi of (201-426) MBq [11C]-PiB. The same single 30 min frame was reconstructed using identical parameters and software (OSEM-3D, 21 it, 4 subs, 5 mm Gaussian), the only difference being AC reconstruction using either DWFS or a co-registered cerebral CT scan (minctracc). PET images were fused with T1w-MRI and activity conc. (Bq/mL) was sampled from ROI’s in frontal, parietal, temporal, occipital cortical areas, white matter, cerebellar grey matter (GM), and the caudate nuclei. Ratio of region-to-cerebellar GM (RGM) was calculated. PET PiB scans were categorized visually as amyloid positive according to standard clinical criteria. Wilcoxon matched-pairs signed rank test was used for statistical evaluation (p < 0.05).

Results The activity conc. of all ROI’s was significantly reduced (p < 0.005) by 19% in DWFS-AC compared to CT-AC. The RGM were spatially variable from the center of the brain to the periphery. RGM using DWFS-AC were on average 8% larger ((0-13)%, p < 0.01) in the central ROI’s and -1% ((-9-6)%, N.S.) in lateral cortical ROI’s compared to CT-AC. The categorization of both, amyloid-positive (n=3) and -negative (n=6) scans was affected neither by the differences in RGM nor by visual interpretation.

Conclusions Following DWFS-AC [11C]-PiB images exhibit a noticeable radially variable bias, which, however, did not affect the clinical interpretation.