Centiloid harmonisation strategies across longitudinal studies: Evaluation on AIBL, ADNI and OASIS3. (31st December 2021)

• Record Type: Journal Article
• Authors: Bourgeat, Pierrick; Li, Shenpeng; Tosun, Duygu; Benzinger, Tammie L.S.; Burnham, Samantha C; Weiner, Mike W.; Morris, John C.; Masters, Colin L; Rowe, Christopher C; Fripp, Jurgen; Dore, Vincent; Villemagne, Victor L
• Is Part Of: Alzheimer's & dementia. Volume 17(2021)Supplement 1
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12-31
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.053660
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: The Centiloid (CL) was proposed to harmonise the quantification of Amyloid PET images across tracers, scanners, and processing pipelines. However, several groups have reported differences across tracers and scanners. In this study, we aim to evaluate the impact of different pre/post‐processing harmonisation steps on the consistency of longitudinal data across multiple large studies. Method: All Amyloid PET data in AIBL (N=3830), ADNI (N=3453) and OASIS3 (N=1398) were quantified using the Centiloid SPM pipeline. SUVR were converted into Centiloids using each tracer's respective transform. All images were smoothed to a uniform 8mm FWHM PSF. Both Raw and 8mm FWHM smoothed images were quantified. For Florbetapir, we evaluated using both the standard Whole Cerebellum (WC) and a Composite WM+WC reference region, which has been previously shown to improve longitudinal consistency. Additionally, our recently proposed Non‐negative Matrix Factorisation quantification (NMF: Neuroimage 2021) was applied to all spatially and SUVR normalised images. Longitudinal consistency was evaluated using: fitting error when fitting a regression line to each subject; percentage of longitudinal changes exceeding 95 th percentile of changes seen with PiB when using a single scanner (6.56CL/y in the negative and 17.06CL/y in the positive); Spearman rank correlation ρ between the baseline Centiloid and the rate of change. Result: Figure 1 illustrates the longitudinal consistency, while the fitting error and percentage of outliers are presented in Table 1. These results indicate that the 8mm FWHM smoothing reduced some of the variability, but its impact was limited. Using the Composite reference region on Florbetapir had a bigger impact in reducing the variability. The best results were however obtained when using the NMF method, compared to SPM. Figure 2 shows the baseline vs rate of CL changes. These plots also show a moderate influence of the FWHM smoothing, and a stronger influence of the Composite reference region. They also show that the NMF method produced the highest Spearman and peak CL/y regardless of the pre‐processing or normalisation method used. Conclusion: FWHM smoothing has moderate impact on longitudinal consistency or outliers. A Composite reference region including subcortical WM should be used for Florbetapir longitudinal Centiloids. NMF improves consistency over SPM.