Effect of magnetic resonance image-based motion correction on the centiloid scale: a comparison with and without correction.

OBJECTIVE: Owing to cognitive impairment, patients with dementia often struggle to understand the importance of remaining still during imaging procedures, which can lead to motion-related artifacts. To address this issue, magnetic resonance (MR) image-based motion correction (MC) techniques have been developed. However, the impact of MRMC on the quantification of amyloid-β (Aβ) positron emission tomography (PET) using the centiloid (CL) method remains unclear. This study aimed to investigate the correlations and differences in the CL values between scans acquired with and without MRMC. METHODS: Among individuals who underwent Aβ PET/MRI at our institution between 2015 and 2024, those who received scans incorporating MRMC (Brain COMPASS, Siemens Healthineers) were selected for analysis. In the MRMC process, a rapid BOLD 3D volume MRI was acquired simultaneously during the 20-min PET scan, enabling deformation estimation and generation of a motion-corrected average PET image. In the conventional protocol, list-mode PET data were divided into four 5-min frames. These frames were then averaged using one of two methods: (1) Rigid MC, in which the second to fourth frames were rigidly registered to the first frame, followed by the averaging of the aligned frames; or (2) Without MC, in which all four frames were averaged without any transformation. CL was analyzed using Amyquant software (version 1.00.0013) and performed on three types of averaged Aβ PET images, namely (1) MRMC, (2) Rigid MC, and (3) Without MC. RESULTS: CL from the three different MC methods showed a strong linear correlation between MRMC and Rigid MC (R2 = 0.998), between MRMC and Without MC (0.978), and between Rigid MC and Without MC (0.974). Bland–Altman analysis demonstrated that larger CL values were associated with greater differences between the methods. Significantly higher CL values were observed with Rigid MC than with MRMC, Without MC than with MRMC, and Without MC than with Rigid MC. However, these biases were small (1.3, 0.47, −0.78), and CLs were strongly correlated. CONCLUSIONS: CL is a semi-quantitative indicator that varies depending on the analytical method, including MC, and is not a fixed value. Caution should be exercised when applying this method in clinical practice.