2. Materials and Methods
2.1 Subjects
All procedures and ethical aspects of this study were approved by the institutional review board of Beijing Normal University Imaging Center for Brain Research. Written informed consent was obtained from each subject.
Subjects in this study were from the Beijing Aging Brain Rejuvenation Initiative (BABRI), which is an ongoing cohort project to investigate the cognitive ageing and impairment of urban residents in Beijing and to identify neuroimaging biomarkers for normal ageing and AD (Yang et al., 2021). In this cross-sectional study, one hundred and sixty-eight (168) subjects from Wave 1 and 2 of the BABRI cohort were included per the following criteria: (a) aged 55 years old and above; (b) had at least 6 years of education; (c) scored 24 or higher on the Chinese version Mini-Mental State Examination (MMSE); (d) had valid N-back task performance data and N-back task fMRI data; and (e) had no history of neurological, psychiatric, or systemic illness known to influence cerebral function.
2.2 fMRI Experimental Paradigm
A blocked periodic design that incorporated alternating 0-, 1- and 2-back conditions was used during the N-back WM task. Each condition contained three blocks and were pseudo-randomly shown to subjects. During the 0-back condition, subjects were asked to press a button when a preassigned digit (e.g., 1) appeared on the screen. In the 1- or 2-back condition, subjects pressed a button when the digit on the screen matched the digit presented one or two items prior, respectively. Every block started with a 10-s cue presentation that indicated the 0-, 1- or 2-back, which was followed by 20 consecutive trials of single-digit stimuli (1000 ms duration and 1000 ms interstimulus interval). The whole experiment ended with 20-s cue presentation that the task was over, and he or she could have a rest. The responses and reaction time from each subject were recorded by an MRI-compatible response button box. The stimuli were presented using E-Prime (version 1.0, Psychology Software Tools Inc., Pittsburgh, PA).
To ensure that subjects understood the instructions and performed the tasks correctly, they were asked to practice a simplified version of the tasks for 10-15 min before the experiment.
2.3 MRI Data Acquisition and Preprocessing
The MRI data were acquired by scanning on a 3.0 T Siemens scanner at the Imaging Center for Brain Research at Beijing Normal University. With the head snugly fixed by straps and foam pads, subjects were asked to refrain from head movement. The functional images were acquired using an echo-planar imaging (EPI) sequence as follows: 33 axial slices, repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, slice thickness = 3.5 mm, flip angle = 90°, field of view = 200 mm×200 mm, acquisition matrix = 64×64, 235 volumes. The T1-weighted structural data were also collected for spatial normalization using three-dimensional (3D) magnetization prepared rapid gradient echo (MP-RAGE) sequences as follows: 176 sagittal slices, repetition time = 1900 ms, echo time = 3.44 ms, slice thickness = 1 mm, flip angle = 9°, field of view = 200 mm×200 mm, acquisition matrix = 256×256.
Functional imaging data were preprocessed using the Statistical Parametric Mapping toolbox (SPM12, http://www.fil.ion.ucl.ac.uk/spm) run within MATLAB software (Mathwork, Inc., Natick, MA). The images were spatially realigned to the iteratively generated (over the realignment procedure) mean image from the series and then resliced, followed by slice-time correction for acquisition order (referenced to the first slice). Based on the transformation parameters from the segmentation of co-registered structural images, the functional images were spatially normalized to the Montreal Neurological Institute (MNI) space at 3 mm isotropic voxel resolution. Together with the six realignment paraments, the mean signals of the brain ventricle and white matter were extracted in the naive space and regressed out from the data. The functional images were then smoothed with a 6-mm full-width half-maximum Gaussian kernel. After preprocessing, six subjects were excluded due to excessive head motion (larger than 3 mm or degree), resulting in 162 subjects included in the final analyses.
For each subject, the preprocessed time series was modelled using GLMs. Task-evoked neural effects were estimated using separate block regressors representing 4 task conditions (fixation/cues, 0-back, 1-back, and 2-back) convolved with the canonical haemodynamic response function (HRF) plus its temporal and dispersion derivatives, and contrast images of interest (i.e., “1back - 0back” and “2-back - 0-back”) were attained by subtracting the 0-back from the 1-/2-back. To acquire task-based background brain activity images, residual images from the GLM were high-pass filtered with a cut-off frequency of 0.01 Hz.
2.4 Voxel-wise Task-evoked Activity Analyses
Group-level one-sample t tests were first conducted to identify regions with task-evoked positive and negative activity in the contrasts of 1back - 0back and 2-back - 0-back, and binary masks of positive and negative regions during both contrasts were generated, under which age-related changes in task-evoked activity were then separately calculated for positive and negative regions in each contrast via multiple regression models in SPM12 toolbox, with gender and years of education as covariates.
For the above analyses, the Gaussian random field (GRF) correction was used for multiple comparison corrections on the results with voxel-level p < 0.001 and cluster level p < 0.05. Brain regions showing significant age-related effects were identified as seed regions for further functional connectivity analyses.
2.5 Task-based Background Functional Connectivity Analyses
The background functional connectivity examined the functional correlations among brain regions that occurred in the background of stimulus-locked changes, that is, independent of the evoked responses to individual stimuli during the tasks. To assess such correlations, for each subject, based on the acquired background brain activity images, the strength of connectivity was evaluated through Pearson’s correlations between the averaged time series of voxels in each seed region (significant regions in 2.4 Voxel-wise Task-evoked Activity Analyses ) and the time series of each voxel in the rest of the brain. Then, Fisher’s r to z transformation was applied to normalize the original correlation maps, followed by whole-brain z score standardization to rescale the functional connectivity using Data Processing & Analysis for Brain Imaging (DPABI) (Yan et al., 2016).
Next, multiple regression models were performed on the standardized FC maps using the SPM12 toolbox to determine regions where FC showed significant age-related changes, with gender, years of education, and mean task-evoked activity of the seed region included as covariates, and all the results were corrected for multiple comparisons using the same method in 2.4 Voxel-wise Task-evoked Activity Analyses . The mean functional connectivity of the resulting areas was then extracted for subsequent analyses.
2.6 Statistical Analyses
Relationships between N-back task performance and task-evoked activity, as well as task-based background FCs, were separately estimated via linear regression models using SPSS (version 22.0, IBM Inc., New York, NY), with gender, years of education, age, and task-evoked activity as covariates, and a statistical significance level of p < 0.05 was applied.
To test how task-evoked activity and task-based background FC mediated age effects on N-back task performance, mediation analyses were further conducted using Mplus 7.11 (Muthén and Muthén, 2012), and bootstrapped 95% confidence intervals (CIs) were reported alongside parameter estimates (Preacher et al., 2007; MacKinnon, 2008).