Quantitative Measurement of Tau Burden in a Dual-Time-Window Dynamic PET Imaging Protocol with [18F]MK6240

Abstract

This study aimed to test and validate a dual-time-window (DTW) protocol for 6-(fluoro-¹⁸F)-3-(¹H-pyrrolo[2,3-c]pyridin-1-yl)isoquinolin-5-amine ([¹⁸F]MK6240) dynamic PET imaging in experimental datasets acquired in human subjects. Methods: DTW protocols were tested and validated in datasets previously collected in 25 participants: 13 were cognitively normal, 10 had mild cognitive impairment, and 2 had Alzheimer disease. Participants underwent full 120-min [¹⁸F]MK6240 dynamic PET scans as well as structural MRI. Intermediary 3-dimensional volumes were removed from the acquired dynamic PET images to emulate DTW acquisitions consisting of an early phase and a late phase. Five break durations (30, 40, 50, 60, and 70 min) were investigated to determine the optimal break for 2 study durations (120 and 110 min). Regional brain time-activity curves were extracted using atlases available in the Montreal Neurologic Institute template space and using the FreeSurfer parcellation. Interpolation strategies were tested to recover the missing time points. Distribution volume ratio (DVR) estimates obtained from the DTW time-activity curves were compared with those obtained from the full time-activity curves as reference. Parametric maps were generated for the selected protocol and evaluated. Results: The correlation and agreement between DVR values obtained from the DTW method and the full time-activity curves were overall very good. The DTW protocol with a 60-min break using a biexponential model fit as the interpolation method provided the best compromise between practicality and quantitative accuracy. The mean differences between this DTW and the full acquisition, averaged across brain regions and all subjects, were less than 1% with a corresponding SD of less than 4%, and DVR estimates were not statistically different from those obtained from the full acquisition (P > 0.05). DVR parametric images were visually and quantitatively consistent with those obtained from the full acquisition. Conclusion: This study presents strong support for the use of a DTW protocol with [¹⁸F]MK6240. Such a protocol would be well suited to allow for both quantification of tau and derivation of an index of cerebral perfusion while reducing patient discomfort and increasing scanning efficiency in comparison to a full dynamic acquisition.

Keywords: Alzheimer disease; PET; [18F]MK6240; dual-time-window imaging; tau imaging.

Graphical abstract

FIGURE 1.

Comparison of DVR estimates computed for 60-min break in 120-min scan analysis with biexponential fit with DVR estimates obtained from full 120-min scan duration. (A and B) Correlation and modified Bland–Altman plots of DVR estimates for analysis performed using atlases from MNI template space. (C and D) Correlation and modified Bland–Altman plots of DVR estimates for analysis performed using FreeSurfer parcellation. In correlation plots (A and C), solid red lines represent total least squares regressions and black dashed lines are lines of identity. In Bland–Altman plots (B and D), solid red lines represent mean differences, dashed red lines show mean differences ± 1.96 SDs, and black dashed lines are zero lines for references. Blue dots represent CN subjects, and orange dots represent MCI/AD subjects. Data points are for all subjects and regional brain time–activity curves. FS = FreeSurfer.

FIGURE 2.

Comparison of DVR estimates computed for 60-min break in 110-min scan analysis with biexponential fit with DVR estimates obtained from full 110-min scan duration. (A and B) Correlation and modified Bland–Altman plots of DVR estimates for analysis performed using atlases from MNI template space. (C and D) Correlation and modified Bland–Altman plots of DVR estimates for analysis performed using FreeSurfer parcellation. In correlation plots (A and C), solid red lines represent total least squares regressions and black dashed lines are lines of identity. In Bland–Altman plots (B and D), solid red lines represent mean differences, dashed red lines show mean differences ± 1.96 SDs, and black dashed lines are zero lines for references. Blue dots represent CN subjects, and orange dots represent MCI/AD subjects. Data points are for all subjects and regional brain time–activity curves. FS = FreeSurfer.

FIGURE 3.

T-RT variability in 2 CN subjects for different break durations (30, 40, 50, 60, and 70 min) of DTW protocol with biexponential fit, compared with full scan acquisition (no break). (A) T-RT performance for 120-min scan analysis. (B) T-RT performance for 110-min scan analysis. (A and B) Analyses performed using FreeSurfer parcellation. Analyses performed using atlases from MNI template space are in Supplemental Figure 10. Each data point represents a brain region surveyed in this study. Yellow and orange dots correspond to regional data for 2 different CN subjects. In panel B, 2 data points for 70-min break condition do not appear on graph as they lie outside graph limits. FS = FreeSurfer.

FIGURE 4.

DVR maps computed for 60-min break with biexponential fit and 120-min scan analysis, and corresponding MR structural images in AD subject and CN subject. First row shows magnetization-prepared rapid gradient echo MRI for anatomic reference. Second row shows parametric DVR images generated from DTW images. Third row shows parametric DVR images generated from full 120-min dynamic images for comparison. Fourth row shows images of difference (DVR DTW – DVR full scan).

FIGURE 5.

DVR maps computed for 60-min break with biexponential fit and 110-min scan analysis, and corresponding MR structural images for same AD and CN subjects as shown in Figure 4. First row shows magnetization-prepared rapid gradient echo as anatomic reference. Second row shows parametric DVR images generated from DTW images. Third row shows parametric DVR images generated from full 110-min dynamic images for comparison. Fourth row shows images of difference (DVR DTW – DVR full scan).

FIGURE 6.

Model-based SUVR_70-90 maps computed for 60-min break with biexponential fit and 120-min scan analysis as well as corresponding MR structural images for same AD and CN subjects as shown in Figure 4. First row shows magnetization-prepared rapid gradient echo as anatomic reference. Second row shows parametric SUVR_70-90 images generated from DTW images. Third row shows original (i.e., from acquired data) SUVR_70-90 images for comparison. Fourth row shows images of difference (SUVR_70-90 DTW – SUVR_70-90 original).

FIGURE 7.

Model-based SUVR_70-90 maps computed for 60-min break with biexponential fit and 110-min scan analysis as well as corresponding MR structural images for same AD and CN subjects as shown in Figure 4. First row shows magnetization-prepared rapid gradient echo as anatomic reference. Second row shows parametric SUVR_70-90 images generated from DTW images. Third row shows original (i.e., from acquired data) SUVR_70-90 images for comparison. Fourth row shows images of difference (SUVR_70-90 DTW – SUVR_70-90 original).