Impact of motion correction on longitudinal [18F]-MK6240 tau PET imaging

Introduction: PET imaging of tau deposition using [¹⁸F]-MK6240 often involves long acquisitions in older subjects, many of whom exhibit dementia symptoms. The resulting unavoidable head motion can greatly degrade image quality thus requiring motion correction (MC). Motion increases variability of PET quantitation across subjects which in turn requires larger sample sizes in clinical trials. Here, we apply an ultra-fast list-mode based motion estimation and correction method to longitudinal [¹⁸F]-MK6240 studies in a large cohort of subjects and evaluate its effect on estimated standardized uptake value ratio changes (ΔSUVR) in key brain regions involved in Alzheimer’s disease (AD). We also evaluate the impact of MC on the statistical power of a hypothetical clinical trial.

Methods: Fifty-one subjects (45 Cognitively Normal (CN), 4 with Mild Cognitive Impairment (MCI), 2 AD; age: 68 ± 11 years) were scanned on a GE Discovery MI PET/CT scanner for 20 min, 90 min after administration of ~185 MBq [¹⁸F]-MK6240. All subjects underwent a repeat scan 8 ± 4 months after their baseline scan using the same protocol. For MC, list-mode PET data were first divided into multiple short (~20 secs) frames, which were separately reconstructed into low-resolution images using the ultra-fast list-mode reconstruction provided by GE Healthcare [Spangler-Bickell et al, JNM, 2021]. Rigid registrations were then used to generate motion transformations needed to align all the frames to a selected reference image, itself registered to the attenuation map. Finally, a list-mode reconstruction incorporating the motion parameters was applied to generate a motion-corrected image. A motion metric was developed using the average displacement of voxels in the brain across all frames. A histogram-based threshold allowed us to identify the studies with important motion. We applied our approach to all 51 longitudinal [¹⁸F]-MK6240 studies and computed ΔSUVR in the entorhinal cortex, the inferior temporal gyrus, the precuneus, and the amygdala using the cerebellar gray matter as the reference region. We calculated the standard deviation of ΔSUVR across all the subjects for data reconstructed with MC and with no MC (NMC). Assuming an annualized rate of change of 10% in early-AD regions, we calculated the sample sizes (for MC and NMC) to achieve the same statistical power for differentiating between two groups (therapy vs control) for varying degrees of reduction in ΔSUVR (one-sided test, confidence level: 95%, power: 80%), in a hypothetical clinical trial.

Results: Individually, 26% of the scans exhibited notable motion, affecting 39% of the longitudinal datasets (motion in one or both time points). The motion metric was significantly higher in MCI/AD subjects than in CN subjects (p<0.05). The mean age of the moving subjects is statistically equivalent to that of the non-moving subjects (p<0.05). Visually, MC images showed substantial reduction in motion-induced blurring, especially in scans with larger motion. Images from one subject with and without MC are shown in Figure 1. MC reduced the standard deviation of ΔSUVR across subjects by -44%, -17%, -13%, and -19% in the entorhinal, inferior temporal, precuneus, and amygdala regions, respectively. In Figure 2, the sample sizes needed for rejecting the null hypothesis (i.e., no significant ΔSUVR difference between therapy and control groups) are plotted as a function of the difference in group means. MC yields lower sample sizes across all regions and its impact is more pronounced for detecting a smaller difference (e.g., resulting from shorter inter-scan duration).

Conclusions: Motion is a major confounding factor in [¹⁸F]-MK6240 tau PET imaging. Motion correction can reduce the variance in estimated ΔSUVR, potentially allowing for shorter inter-scan time and smaller sample size in clinical trials evaluating the effect of a candidate drug against AD.

• Download figure
• Open in new tab
• Download powerpoint