Participants and procedure
This study adopted a cross-sectional design. Sixty-nine university students without psychopathology backgrounds in Hong Kong were recruited through social media, posters, or emails at local universities using the snowball sampling method. The inclusion criteria were as follows: 1. Age between 18–35 years old; 2. Full-time or part-time university student; 3. No habit of substance or alcohol use; 4. No history of, or current suffering from, any psychiatric or psychological disorders; 5. Not suffering from acute or long-term inflammation; 6. Not suffering from an infectious disease during the study; 7. Not currently pregnant (for female participants only); and, 8. MRI capable. All inclusion criteria were confirmed by a brief phone interview prior to the day of the experiment.
On the day of the experiment, participants were first briefed on the procedure of the experiment and asked to complete a written consent form. A complete MRI safety screening form was also administered by the staff in the MRI center. Then, they were instructed to undergo a 30-minute brain scan, which included 15 minutes of T1 scan, and 12 minutes of T2* while fixated on a cross with eyes open. Participants were instructed to remain calm and still during the scan and not to think of anything specifically. Lastly, participants completed a set of questionnaires regarding demographics, handiness, and three different conceptualizations of resilience right after scanning and outside of the scanning room (for details, refer to 2.2 Psychological measures). The data collection procedures were approved by the Human Research Ethics Committee of the Education University of Hong Kong (ref no. 2019-2020-0398). All methods were carried out in accordance with relevant guidelines and regulations. Informed consent was obtained from all participants.
Psychological measures
Overall self-perceived resilience
Overall self-perceived resilience was measured by CD-RISC[2], which was reported to be one of the most commonly used scales for measuring resilience[1]. It consists of 25 items rated on a 5-point Likert scale from 0 (not true at all) to 4 (true nearly all the time). This scale contains questions like “I am able to adapt to change” to target one’s overall self-perceived resilience. A higher score means better overall self-perceived resilience for the individual. Our data showed Cronbach’s alpha = 0.885 and Omega = 0.886, representing good internal consistency[43].
State and trait resilience
The SRC/TRC[4] was administered to capture both state and trait resilience separately. SRC consists of 15 items while TRC consists of 18 items. Both measures are scored with a 5-point Likert rating (from 1 = ‘strongly disagree’ to 5 = ‘strongly agree’). SRC contains statements like “I have someone who loves me”, and TRC contains statements like “I have warm positive relationships with adults”. Instruction for SRC asks participants to rate the items based on “how you feel right now”, while instruction for TRC asks “how you generally feel”. Our data showed good internal consistency for both SRC (Cronbach’s alpha = 0.805; Omega = 0.806) and TRC (Cronbach’s alpha = 0.882; Omega = 0.882)[43].
Outcome-based resilience
Outcome-based resilience was measured by the ratio of the Perceived Stress Measure (PSS) to the General Health Questionnaire (GHQ)[7]. The PSS consists of 14 items on a 5-point Likert scale from 0 (never) to 4 (always), focusing on recent potentially stressful events[44]. The PSS includes questions such as “In the last month, how often have you felt that you were on top of things?” to examine one’s perceived stress over the past month. The GHQ consists of 12 items rated on a binary method with 1 (absence of the symptom) to 2 (presence of the symptom) focusing on mental health problems[45]. The GHQ includes statements like “Felt constantly under strain” to assess mental health. For GHQ, our data showed Cronbach’s alpha = 0.843 and Omega = 0.841. For PSS, our data showed Cronbach’s alpha = 0.843 and Omega = 0.849. Both scales had good internal consistency[43].
MRI data acquisition and preprocessing
Data acquisitions were performed on a 3.0T scanner (SIGNA™ Premier) located at the University of Hong Kong. High-resolution T1-weighted images were acquired with an MPRAGE sequence using the following parameters: repetition time (TR) = 1900 ms; echo time (TE) = 31.52 ms; inversion time (TI) = 900 ms; flip angle = 8°; repetition matrix = 256 x 256; and voxel size = 0.5 x 0.5 x 0.5 mm3. Subsequently, a gradient echo-planar imaging sequence was utilized to acquire 244 volumes with the following parameters: TR = 2000 ms; slices = 45; TE = 30 ms; thickness = 3.5 mm; flip angle = 90°; field of view = 24 x 24 cm2; repetition matrix = 90 x 90; and voxel size = 1.75 x 1.875 x 3.5 mm3. Participants were instructed to stay still and relax while fixating on a cross during the resting state fMRI scans.
Data preprocessing was conducted using the DPARSF software[46]. The first 10 images were discarded due to signal equilibrium. Slice timing and head motion correction were performed. Eleven participants were excluded because their head motion exceeded ± 1.5 mm, resulting in fifty-eight participants for data analyses. Then, these images were co-registered with echo planar imaging (EPI) templates and normalized into the standard Montreal Neurological Institute (MNI) space with a 3 x 3 x 3 mm3 resolution. Next, the band-pass filter (0.01–0.1 Hz) was used to smooth the data. Some confounding factors, including white matter volume, global signal, cerebrospinal fluid, and head movements (Friston 24-parameter model), were removed from the smoothed data.
ReHo-resilience multiple regression analysis
The Regional Homogeneity (ReHo) method was used to examine the local activities. ReHo is a reliable measure that represents the local synchronization of spontaneous brain activities[47], which is commonly used in resting-state function MRI analysis. The ReHo values were extracted using the DPABI toolbox. A whole-brain atlas named Anatomical Automatic Labelling (aal)[48] with 116 regions of interest (ROIs) was adopted. The Statistical Parametric Mapping Program (SPM12, http://www.fil.ion.ucl.ac.uk/spm) was utilized in MATLAB R2019a (The Mathworks, Natick, MA) to perform the multiple regression analyses using the ReHo value of different brain regions in the whole-brain atlas to predict each resilience measure separately. Covariates were age, gender, and education levels. A corrected cluster threshold of p < 0.05 and the cluster size ≥ 10 were set.
Regression analysis of long-range functional connectivity and different operational definitions of resilience
The functional connectivity was extracted using the DPABI toolbox with the same whole-brain atlas, aal. Multivariate regression analyses were performed to assess the relationship between different resilience measures and long-range functional connectivity using R studio. The whole brain analysis consisted of 116 x 116 pairs of functional connectivity, while the seed-based analysis set predetermined ROIs (seeds) and their functional connectivity to the other 116 ROIs in the aal atlas. A total of 14 seeds were included. Seeds were the 4 regions: ACC, amygdala, OFC, and insula, bilaterally, indicated in the previous review[14], and the seven significant regions found in ReHo-resilience multiple regression analysis, which included the left hippocampus, left parahippocampal gyrus (PHG), right postcentral gyrus, right inferior occipital gyrus (IOG), right angular gyrus, right gyrus rectus, and right insula. Neuroimaging data were set as independent variables to predict the three operational definitions of resilience in separate regression models. Covariates included age, gender, and education levels. The false discovery rate (FDR) correction method was adopted to better control the balance of Type I and II error rates[49].