Assessment of attenuation correction methods for quantitative neuro-PET/MR

Sammanfattning: Hybrid PET/magnetic resonance (MR) can provide physiological, functional, and structural information simultaneously, facilitating research in neurological disorders. For quantitative PET, correction for photon attenuation (AC) is necessary. However, in contrast to dedicated PET and PET/computed tomography (CT) systems, PET/MR has no direct possibility to measure photon attenuation. As such, MR-based methods are required for AC (MRAC), and these need to be thoroughly validated before clinical implementation.The primary aim of this thesis was to evaluate two vendor-provided MRAC methods (single-atlas and zero echo time, ZTE), a previously published maximum probability (MaxProb) method, and a composite transmission scan atlas (CTR) method for a SIGNA PET/MR. This evaluation was done both in terms of absolute quantification in static scans and of outcome measures of tracer kinetic modelling based on dynamic scans. The secondary aim was to compare quantitative brain PET measurements acquired on the SIGNA PET/MR with those acquired on a dedicated PET scanner. Ten patients with parkinsonism who underwent dynamic dopamine transporter scans using 11C-PE2I in a PET/MR and dedicated PET were included. Standardized uptake values (SUV), binding potential (BPND), and relative delivery (R1) were assessed at volume of interest (VOI) and voxel level to compare the various MRAC methods with the gold-standard, a 68Ge transmission scan, and to compare quantitative outcomes between scanners.In general, ZTE provided the highest precision in SUV, R1 and BPND, showing the least inter-subject variability in bias compared to 68Ge-transmission AC, whereas MaxProb and CTR showed the lowest precision. Contrary to this, accuracy of absolute SUV values was best for CTR followed by MaxProb, with ZTE showing a homogeneous positive bias of about 10%. ZTE provided the highest accuracy in outcome measures of tracer kinetic analysis. Differences in quantitative results between stand-alone PET and PET/MR exceeded what can be explained by difference in AC alone, although they were still comparable to previously published test-retest variability of 11C-PE2I. Additionally, an activation in the auditory cortex was seen in PET data from the PET/MR because of the noise produced by the MR gradients.ZTE-MRAC appears to be the best method for dynamic scanning and tracer kinetic analysis using reference methods, while CTR- and MaxProb-MRAC appear the most appropriate for absolute quantification. Also, attention should be taken to the auditory cortex activation in R1 images when comparing data from PET/MR and other PET- systems.