Dysfunction of anterior insula in the non- affected hemisphere in patients with post- stroke depression: A resting-state fMRI study

2.1Participants

In this cross-sectional study, twenty-five patients with stroke (16 females, 9 males) with unilateral left or right middle cerebral artery (MCA) infarction and 13 age-matched healthy control subjects (8 females, 5 males) were recruited. All participants were right-handed. None of the subjects had severe visual or perceptual impairments, or a previous history of affective disorders. All patients underwent rs-fMRI within one month (25.6 ± 5.75 days) from the onset of stroke, and all had severe hemiplegia with no hand or finger motion at that time. The risk factors for stroke were assessed in all patients. The risk factors included diabetes, hypertension, smoking, alcohol, and dyslipidemia. Other factors possibly associated with PSD were assessed using the National Institutes of Health Stroke Scale (NIHSS), Modified Barthel Index (MBI), and modified Rankin Scale (mRS). We used the Beck Depression Inventory (BDI)-II to evaluate the severity of depression immediately before rs-fMRI scanning. All participants had an interview by a psychiatrist and then were diagnosed as PSD according to DSM-V criteria. Five or more of the following symptoms present during the same 2-week period; depressed most of the day, subjective hopeless or observative tearful nearly every day, markedly diminished interest/pleasure in all, significant weight changes, insomnia or hypersomnia nearly every day, psychomotor agitation or retardation nearly every day, fatigue or loss of energy nearly every day, feelings of worthlessness or inappropriate guilt, diminished ability to concentrate, and recurrent thoughts of death, suicidal ideation, or a suicidal attempt. There were at least one of the symptoms is either depressed mood or loss of interest/pleasure in PSD. One group (patients without PSD group: n= 12, 8 female patients, 4 male patients, 63.08 ± 15.18 years) was composed of patients with no depression after stroke, while the other group (patients with PSD group: n= 13, 8 female patients, 5 male patients, 68.92 ± 9.81 years) was composed of individuals with diagnosed as PSD. All subjects provided written informed consent and agreed to participate in the rs-fMRI study. The study protocol was approved by the local Institutional Review Board (IRB no. 2012-05-023).

2.2Rs-fMRI data acquisition

A 3.0T GE Signa Exite (Milwaukee, WI, USA) scanner was used to obtain the rs-fMRI images and T1-weighted images. The rs-fMRI images were acquired for BOLD signals using the following parameters: echo planar imaging (EPI) sequence, echo time (TE) = 30 ms, repetition time (TR) = 2000 ms, flip angle (FA) = 90∘, field of view (FOV) = 22 cm, acquisition matrix = 64 × 64 and 4 mm thickness with no gap (voxel size = 3.43 × 3.43 × 4 mm, resolution = 0.29 pixel/mm). The total scan volumes were 240 volumes without dummy scans (six volumes), and the total scan time was 8 min 12 s, including dummy scans. During scanning participants were required to keep their eyes closed, to stay awake, and to avoid thinking. An anatomical T1-weighted image was acquired using the following parameters: 3D fast spoiled gradient echo sequence, TE = 3 ms, TR = 7.8 ms, FA = 20∘, FOV = 21 cm, acquisition matrix = 256 × 256, 1.3 mm thickness with no gap (voxel size = 0.82 × 0.82 × 1.3 mm, resolution = 1.21 pixel/mm), and scan time was 2 min 36 s.

2.3Rs-fMRI data analysis

The statistical parametric mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm/) program on MATLAB (MathWorks, Inc., Natick, MA, USA) was used for pre-processing and statistical analysis of rs-fMRI images and T1-weighted images. The rs-fMRI images and T1-weighted images from patients with a left lesion were flipped to change it from a left lesion to a right lesion because the whole brain analysis was performed on contralesional and ipsilesional hemispheres. All rs-fMRI data, including flipped data, were preprocessed using a slice timing correction, spatial realignment, normalization, and smoothing. Rs-MRI data were obtained by an interleaved method, and slice timing was performed to correct the time between slices. Then, corrected rs-MRI data were realigned to the first image to correct the head movement between scans, and the translation parameters for the x (left/right), y (anterior/posterior), and z (superior/ inferior) directions were obtained for each brain volume compared to the first volume. The realigned rs-fMRI data were used for outlier detection using the Artifact Detection Tool (ART). The realigned rs-fMRI data were mean adjusted by proportional scaling and spatially normalized in a Montreal Neurological Institute (MNI) template (voxel size = 2 × 2 × 2 mm, resolution = 0.5 pixel/mm). Finally, the full-width at half maximum (FWHM) of the Gaussian kernel in 8 mm was used to smooth the normalized rs-fMRI data. The mean displacement (displacement = square root [x2+ y2+ z2]) was calculated to estimate the head motion. No experiment had a mean displacement over 2 mm. There was no mean displacement difference (P= 0.985) between the groups without PSD (mean displacement, 0.43 ± 0.06 mm), with PSD (mean displacement, 0.46 ± 0.07 mm) or healthy controls (mean displacement, 0.51 ± 0.12 mm).

Pre-processed data were entered the CONN tool (http://web.mit.edu/swg/software.htm, version 15.b) to determine FC using a seed-based method. The coordinates of the anterior insula in the non-affected hemisphere [9, 16] was identified using MarsBaR software (http://marsbar.sourceforge.net/) and the meta data of the Neurosynth site (http://www.neurosynth.org). Noise due to cerebrospinal fluids, white matters, or head motions, and sources of artifacts in time series data from the ART were corrected for as nuisance covariates. Individual FC maps from the anterior insula in the non-affected hemisphere were made for the seed-to-voxel connectivity maps in the first-level analysis. Then, the results of the first-level analysis were used for within-group, analysis of variance (ANOVA), post-hoc, and multiple regression analyses for the seed-to-voxel connectivity maps in the second-level analysis.

Lesion boundaries were determined using T1-weighted images from each patient with the aid of MRIcron software (http://people.cas.sc.edu/rorden/mricron/index.html). To mask the regions of damage, normalized T1-weighted images were used as a reference, and masking areas were manually drawn by a rehabilitation doctor who was blinded to the study. Masked images were combined using the SPM12 program and displayed using MRIcroGL software (http://www.cabiatl.com/mricrogl/). In addition, the lesion volume was calculated to compare the total lesion size between patients without PSD group and with PSD group, and the comparison was tested using the non-parametric Mann-Whitney test.

2.4Thresholding criteria & statistical analysis

To compensate for multiple comparisons across the brain, spatial extent thresholds were calculated using AlphaSim with Monte Carlo simulations from the SPM REST toolbox (http://restfmri.net/). Input parameters for minimum cluster size were determined by performing a 5000 Monte Carlo simulation with a cluster connection radius of 5 mm and a FWHM of 8 mm to satisfy the Family Wise Error (FWE) rate correction of PFWE< 0.001 for the T-contrast.

All statistical analyses on clinical data were performed using SPSS software (SPSS Inc., Chicago, IL, USA, version 18), and differences with a P< 0.05 were defined as statistically significant. Where data were on the ratio scale, age and FC score, normal distribution was assessed using the Shapiro-Wilk normality test on the groups. Then, where the data were normal, a one-way ANOVA and two sample t-test were performed to determine differences between the groups. The Kruskal-Wallis test was performed where data were not normally distributed. Where data were on the nominal scale, i.e., sex, the chi-square test was used to identify differences between the groups.

3.1General characteristics

The mean age of the 25 patients with stroke was 66.12 ± 12.75 years (range, 32–85 years) and that for 13 healthy control subjects was 64.61 ± 6.13 years (range, 55–72 years). No differences in age were observed between patients without PSD, patients with PSD, or healthy controls (P= 0.385). The BDI-II scores of patients without PSD, patients with PSD, and healthy controls were 8.33 ± 3.95, 23.23 ± 9.65, and 5.38 ± 3.04, respectively. There was no statistical difference in BDI-II scores between the patients without PSD group and the healthy control group (P= 0.483). The PSD group had significantly different BDI-II scores when compared to other two groups (P< 0.001). The group differences in demographics, risk factors of stroke, type of stroke, and severity of impairment after stroke are summarized in Table 1. The comparison of lesion volume showed no statistical difference between the without PSD group and with PSD group (P= 0.06).

3.2FC maps

A one sample within-group analysis showed the FC map between the anterior insula and the other brain areas in the non-affected hemisphere for the three groups (Fig. 1). The detailed brain areas showing FC with the anterior insula in the non-affected hemisphere are summarized in the Supplement (Tables S1–S3). While a one sample within-group analysis also showed FC between the anterior insula in the non-affected hemisphere and brain areas in the affected hemisphere, the brain areas in the affected hemisphere were not included in Supplementary Tables S1–S3. The main reason for excluding brain areas in the affected hemisphere was that the affected hemisphere is the damaged side of the brain, and thus only brain areas in the intact hemisphere were considered for analysis since damage-related physiological factors, such as a microbleeding, can be possible confounders for analysis.

Figure 1.

The functional connection with the anterior insula from the within subject analysis with a Family Wise Error (FWE)-corrected threshold of P< 0.001 for multiple comparisons. In the maps, left part corresponds to the non-affected hemisphere and right part corresponds to the affected hemisphere.

Figure 2.

Significant functional connectivity differences at the whole brain level were revealed by ANOVA and post-hoc two sample test. The SPM{F}s and SPM{t}s with a Family Wise Error (FWE)-corrected threshold of P< 0.001 for multiple comparisons.

Figure 3.

The functional score between the non-affected hemispheric anterior insula and the superior frontal, middle frontal, and orbitofrontal gyrus regions showed significantly statistical differences between patients with PSD group and each of the other groups (patients without PSD group and healthy control group). P*< 0.05, P**< 0.01 indicated.

The ANOVA of the FC maps revealed that the superior frontal, middle frontal, and orbitofrontal regions of the non-affected hemisphere were significantly different between the three groups (Fig. 2A and Supplementary Table S4). A post-hoc two-sample t-test was then performed for further comparison between groups. First, comparison of patients with PSD and healthy control groups showed FC differences in the superior frontal, middle frontal, and orbitofrontal regions (Fig. 2B and Supplementary Table S5). There was no difference in FC between patients with PSD group and healthy control groups. Second, the comparison of patients with PSD group and without PSD group showed a difference in FC in the superior frontal region (Fig. 2B and Supplementary Table S5). Third, the comparison between patient without PSD group and healthy control groups showed no statistical difference in FC. All post-hoc analyses were performed at a FWE rate correction of PFWE< 0.001 performed using the Monte Carlo simulation in the REST Tool box.

3.3FC score

FC scores in the superior frontal, middle frontal, and orbitofrontal areas of the non-affected hemisphere are represented in Fig. 3. The FC scores in the superior frontal area were as follows: -0.17 ± 0.13 (healthy group), -0.03 ± 0.12 (patients without PSD group), and 0.12 ± 0.10 (patients with PSD group). In the middle frontal area, the FC scores were as follows: -0.15 ± 0.08 (healthy group), -0.03 ± 0.18 (patients without PSD group), and 0.17 ± 0.20 (patients with PSD group). The FC scores in the orbitofrontal area were as follows: -0.16 ± 0.13 (healthy group), -0.02 ± 0.12 (patients without PSD group), and 0.12 ± 0.11 (patients with PSD group). In these three brain areas, the FC scores were significantly different among three groups.

Figure 4.

Correlation map showed that FC scores in the superior frontal, middle frontal, and orbitofrontal regions in the non-affected hemisphere strongly correlates with BDI-II scores.

3.4Correlation between FC and BDI-II scores

The FC scores in the superior frontal (r= 0.585, P< 0.001), middle frontal (r= 0.545, P< 0.001), and orbitofrontal (r= 0.597, P< 0.001) areas of the non-affected hemisphere showed positive correlations with BDI-II scores (Fig. 4). These positive correlations thresholded to a family-wise error (FWE) rate correction of PFWE< 0.001.