AmyloidIQ: An advanced analytical algorithm to quantify amyloid‐PET [18F]NAV4694 scans: Neuroimaging / New imaging methods. (7th December 2020)

• Record Type: Journal Article
• Authors: Rizzo, Gaia; Whittington, Alex; Hesterman, Jacob; Gunn, Roger N
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 4
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• 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.043823
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: The global amyloid‐β (Aβ) burden in Alzheimer's disease (AD) is routinely quantified from static amyloid PET scans using SUVR. We have previously used Amyloid IQ to analyse over 3000 18 F‐Florbetapir, 18 F‐Florbetaben and 18 F‐Flutemetamol scans. These analyses demonstrated that the outcome measure amyloid load (AβL ) derived from Amyloid IQ is more powerful than SUVR in cross‐sectional and longitudinal analyses (Whittington et al. 2018 JNM) as well as having a better agreement with visual reads (Whittington et al. HAI 2019). In this work, we extend the Amyloid IQ methodology to [ 18 F]NAV4694 images downloaded from the Global Alzheimer's Association Interactive Network (GAAIN) website (http://www.gaain.org). Method: AβL is automatically calculated by the Amyloid IQ algorithm from an individual Aβ‐PET scan using voxel‐wise regression with two canonical images that represent non‐displaceable and specific binding signals. The canonical images were derived from 18 F‐Florbetapir ADNI data (Whittington et al. 2018 JNM). A composite SUVR was also calculated for each scan. 52 [ 18 F]NAV4694 scans (35 healthy controls, HC, 10 mild cognitive impaired, MCI, and 7 AD) were obtained from GAAIN. The effect sizes (Hedges' g) between HC and AD patients, and between HC and AD/MCI of AβL and SUVR were compared in [ 18 F]NAV4694 cross‐sectional data. Result: The Amyloid IQ algorithm was successfully extended to [ 18 F]NAV4694 data with accurate characterization of the data obtained with the previously derived canonical images, now demonstrating its applicability to four different Aβ tracers (Figure 1). In this small sample dataset the cross‐sectional analysis showed that, AβL had similar performance to SUVR (effect size between non‐HC patients and HC was 2.29 for AβL and 2.32 SUVR, and between HC and AD patients was 4.41 for AβL and 4.90 for SUVR). In previous works, AβL has shown better agreement with visual reads than SUVR (Figure 2) and more power in cross‐sectional analysis (Figure 3) with an average increase in effect size of 44% and 21% compared to SUVR for [ 18 F]Florbetapir and [ 18 F]Florbetaben respectively. Conclusion: Amyloid IQ is a powerful algorithm that has been successfully applied to four different Aβ tracers and that can be used for stratification. A larger sample is required to determine whether Amyloid IQ is more powerful than SUVR for [ 18 F]NAV4694.