Objectives:Amyloid PET quantification can be affected by differences in image processing. For example, brain tissue segmentations can be performed in the subject- or standard-space (e.g. MNI), which might have an impact in regional and parametric analysis of the images. The aim of this study was to assess the effects of using different non-rigid image deformations (to standard-space) on quantitative PET metrics. Additionally, effects of using different grey matter masks, derived from population- or subject-based segmentations, were compared.
Methods:Alzheimer’s Disease (AD, n=12) and healthy (HC, n=16) subjects underwent 70 min dynamic [11C]PiB PET scans and T1-weighted MRI scan. The 40–60 min of the [11C]PiB image were averaged and used for the registration to the individual MR scan (subject-space). Following, the MRI was used to calculate the transformation from subject-to-standard space using three different methods: preserving total amount (AMOUNT), preserving concentrations (CONCENTRATION), and with tissue probability maps (TPM). Grey matter probability maps were defined in subject-space (derived from the MRI) and in standard-space (population-based).Metrics were extracted for all the combinations of space, tissue segmentation, and transformation matrix, within grey matter. [11C]PiB SUV ratios (SUVR) were calculated using the cerebellum as reference. Two intensity metrics were extracted, one using the uptake in all grey matter (SUVRmeanall) and another including only amyloid-beta positive (Aβ+) voxels, with a SUVR≥1.5 threshold (SUVRmeanAβ+). Moreover, two volume-based metrics were defined: amyloid fractional volume (AFV, percentage of Aβ+ volume) and total amyloid burden (TAB, SUVRmeanAβ+ times Aβ+ volume). All image data were processed with SPM12, and the analysis was performed in R Studio (v1.1.456; R v3.5.1). Mean differences and standard deviations are shown.
Results:When using the same spatial normalisation methodology and grey matter definition, small variations in the intensity metrics were found (0.4%±0.7%) between the subject- and standard-space results. However, within a specific space, grey matter definition had a greater impact in SUVRmeanall (5.67%±2.8%), but not on SUVRmeanAβ+ (0.7%±1.5%). Overall, normalisation methods AMOUNT and CONCENTRATION provided similar intensity metrics (0.7%±0.6%). Meanwhile, TPM was different than the other methods (7.5%±7.4%).Volumetric measurements showed larger differences between spaces when the spatial normalisation and grey matter were the same (AFV: 4.6%±6.8%; TAB:15%±10%). Within the same space and normalisation, large differences were observed for AFV and TAB depending of the grey matter segmentation method (4.6%±15% and 33%±14%, respectively). For spatial normalisation methods, CONCENTRATION and AMOUNT provided similar volume-based metrics (12%±9.5%), while TPM showed larger differences to the other methods (27%±21%).
Conclusions:Intensity metrics were less affected by differences in PET image processing procedures while volumetric measurements depended more strongly on both space and spatial normalisation method applied and may thus only be valid when derived in subject-space.Therefore, only when SUVR are quantitative image analysis may be performed in either subject- or template-space, while volumetric measures should be performed in subject-space only.(This project is funded by European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 764458.)
Originele taal-2English
Pagina's (van-tot)111-112
Aantal pagina's2
TijdschriftJournal of Cerebral Blood Flow and Metabolism
StatusPublished - 2-jul-2019

Citeer dit