Premium
IC‐P‐048: Estimating and Accounting for The Effect of MRI Scanner Hardware Changes on Longitudinal Whole‐Brain Atrophy Measurements
Author(s) -
Lee Hyunwoo,
Nakamura Kunio,
Narayanan Sridar,
Brown Robert,
Arnold Douglas L.
Publication year - 2016
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2016.06.058
Subject(s) - scanner , atrophy , neuroimaging , nuclear medicine , brain size , computer science , psychology , medicine , artificial intelligence , magnetic resonance imaging , neuroscience , radiology , pathology
Background:Longitudinal MRI studies are often subjected to scanner hardware changes or upgrades, which may alter image characteristics such as contrast, signal-to-noise ratio, intensity nonuniformity and geometric distortion. This renders measurements of brain atrophy potentially unreliable as it may introduce nonphysiological brain volume fluctuation across the point of hardware change. Mixed-effects modelling is one way to estimate rates of brain atrophy while identifying and correcting this bias. Methods: We analyzed 680 control, mild cognitive impairment (MCI), and Alzheimer’s disease (AD) subjects who were scanned at 1.5T for the Alzheimer’s Disease Neuroimaging Initiative (ADNI) phases 1, GO, and 2. The subjects were initially scanned on 10 different scanner hardware models from GE, Siemens, and Philips. 442 subjects continued with the consistent scanner model throughout the follow-up up to 8 years, but 238 subjects were subject to intraor inter-vendor MRI scanner hardware upgrades or changes during the study. A total of 3411 T1-weighted scans were pre-processed using a cross-sectional pipeline including nonparametric non-uniform intensity normalization. Then, the percentage brain volume changes (PBVCs) between the follow-up scans and the baseline “screening” scans were calculated using FSL-SIENA.Amixed-effects model with subject-specific random slopes and intercepts was applied to estimate the fixed effect of scanner hardware changes on the PBVC measures. The same model also included a term to estimate the fixed effect of a scanning sequence change from MPRAGE to IR-SPGR in some subjects. Results: Significant fluctuations in PBVC were found across the following hardware upgrade or change combinations (SE; p): Siemens Symphony to SymphonyTIM -0.5% (0.1; p<0.0001); Philips Intera to Siemens Avanto -1.6% (0.5; p1⁄40.001); GE Excite to GE HDx 0.2% (0.07; p1⁄40.002); GE Excite to GE HDxt 0.3% (0.1; p1⁄40.02). The change of sequence from MP-RAGE to IR-SPGR was associated with an average -1.5% (0.1; p<0.0001) change. Different scanner combinations also showed different biases. Conclusions: Scanner hardware and pulse sequence changes have significant effects on estimation of brain atrophy rates. If suitable data are available, it may be possible to explicitly account for these during the analysis. References: [1] Nakamura K. et al. NeuroImage Clin. 2013 [2] Smith S.M. et al. NeuroImage. 2002.