P4-289: BOLD SIGNAL VARIABILITY CHANGES IN THE DEFAULT MODE AND SALIENCE NETWORKS IN AMNESTIC MILD COGNITIVE IMPAIRMENT AND ALZHEIMER'S DISEASE AND THEIR ASSOCIATIONS WITH COGNITIVE DECLINE
Liwen Zhang,Xi‐Nian Zuo,Kwun Kei Ng,Joanna Su Xian Chong,Hee Youn Shim,Yng Miin,Boon Linn Choo,Eddie Chong,Zi Xuen Wong,Saima Hilal,Narayanaswamy Venketasubramanian,Boon Yeow Tan,Christopher Chen,Juan Zhou,Fang Ji
DOI: https://doi.org/10.1016/j.jalz.2019.06.3958
2019-01-01
Abstract:Optimal level of Blood-Oxygenation-Level-Dependent (BOLD) signal variability (variability hereafter) in the grey-matter plays an important role in healthy brain functioning, which is compromised during ageing (Fox and Raichle, 2007; Grady and Garrett, 2014). Nonetheless, network- and frequency-specific variability changes remain largely unknown along the disease continuum of Alzheimer's disease (AD). Given previous findings of network disruptions in AD and amnestic mild cognitive impairment (aMCI), we hypothesized that AD and/or aMCI showed disturbed variability compared with the healthy elderly, especially in the default mode network (DMN). We investigated resting-state variability patterns in 98 AD, 103 aMCI, and 48 age-matched healthy controls (HC), focusing on the two characteristic slow frequency-bands of slow4 (0.027−0.073 Hz) and slow5 (0.01−0.027 Hz) (Cordes et al., 2001). Variability was defined as standard deviation of the BOLD signal following previous publication (e.g., Garrett et al., 2010). Since no gold standard on fMRI data denoising exists, we repeated analyses using two commonly used pre-processing approaches: global signal regression (GSR) and independent component analysis (ICA). We also explored the associations of variability changes with baseline cognitive performance and cognitive decline, as well as hippocampal atrophy. Both overlapping and distinct patterns were found between frequency bands and data pre-processing methods (Figure 1 +Source ). Specifically, using GSR, aMCI had higher variability in the posterior DMN but lower salience network (SN) variability compared with AD and HC for both frequency bands. Using ICA, we replicated this divergent DMN-SN variability changes in aMCI at slow5, and similar greater DMN variability in aMCI compared with AD at slow4. However, the slow4 DMN-SN differences between aMCI and HC were missing or even reversed. Furthermore, ICA-based denoising revealed lower posterior DMN variability in AD compared with HC for both frequency bands. Importantly, lower slow4 posterior DMN variability was related to poorer cognition at baseline and smaller hippocampal volume, and correlated with faster cognitive decline over time using both data denoising approaches.