2.1 Participants
A total of 150 participants were recruited from the community (85 HCs) and the Beijing Anding Hospital (65 patients with BD); the sample is the same as in our previous publication29. In summary, the patients and HC were matched for age, gender, and education. Twenty-three subjects (nine patients with BD and 14 HCs) were excluded from the final analysis due to excessive head motion (details are shown in the “fMRI data preprocessing”). All patients satisfied the diagnostic criteria of the Structured Clinical Interview for DSM-IV (SCID)30. Our inclusion and exclusion criteria were described in our previous study29. In this experiment, the Young Mania Rating Scale (YMRS) 31 scores of all patients were less than 7, thereby indicating that all patients were not in a manic mood state. We focused on patients with BD in a depressive mood state or a euthymic mood state because BD patients spend more than 80% of their time in a nonmanic mood state. The Hamilton Depression Rating Scale (HAMD)32 was used to assess patients’ depressive symptoms.
The current study has been approved by the Institutional Review Board of Brain Image Center, Beijing Normal University and Beijing Anding Hospital, Capital Medical University. Written informed consents were obtained from all subjects prior to participation in this study.
2.2 Fmri Data Acquisition
All imaging data were acquired at the Brain Imaging Center of the Beijing Normal University by using a Siemens TIM Trio 3T scanner (Siemens, Erlangen, Germany). Resting state functional images (240 volumes) were acquired initially, followed by the T1 images. During scanning, subjects were asked to try not to swallow or move their body, especially their heads. During resting state, subjects were also instructed to close their eyes, relax, and remain awake. Resting state functional images were collected using echo-planar imaging (EPI) sequence, as follows: axial scanning, repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, flip angle = 90°, field of view (FOV) = 200 × 200 mm2, matrix size = 64 × 64, slices = 33, interlaced scanning, slice thickness = 3.5 mm, and gap = 0.7 mm. T1 images were collected using T1-weighted sagittal 3D magnetization-prepared rapid gradient echo (MPRAGE) sequence, as follows: TR = 2530 ms, TE = 3.39 ms, flip angle = 7°, FOV = 256 × 256 mm2, matrix size = 192 × 256, slices = 128, thickness = 1.33 mm, and voxel size = 1.33 × 1 × 1 mm3.
2.3 Fmri Data Preprocessing
Resting state fMRI date was preprocessed in the toolbox for the Data Processing and Analysis for (Resting State) Brain Imaging (DPABI v3.1)33. The preprocessing steps included removing the first five time points, slice timing, realignment, detrending, removing the head motion effect (using Friston 24-parameter model)34, regressing out nuisance covariates (the white matter signals, the cerebral spinal fluid signals, and the global mean signals), normalizing the MNI space (using T1 images), resampling (voxel size = 2 × 2 × 2 mm3), smoothing (FWHM = 4 mm, only for the fALFF and rsFC analysis), and band-pass filtering (0.01–0.1 Hz, only for the ReHo, DC, and rsFC analysis). In addition, we used volume-based framewise displacement (FD)35 to quantify head motion. FD reflects head motion from one volume to the next, and a volume with FD < 0.2 indicates a good time point. Mean FD is calculated by the average of the sum of the absolute values of the differentiated realignment estimates (by backwards differences) at every timepoint36. Subjects would be excluded if their mean FD exceeded three standard deviations beyond the mean value of the entire sample or the number of good time points was less than 120.
2.4 Fmri Data Analysis
First, three indices of the fALFF, the ReHo, and the DC were used to unravel abnormal brain activities in the patients with BD. Subsequently, using the overlapped cluster of the significant brain region found in the analysis as seed, rsFC analysis was performed to explore possible brain regions, which showed abnormal rsFC with the overlapped cluster in patients with BD. Finally, two additional strategies were used to reidentify the seed region and then confirmed our rsFC analysis.
fALFF, ReHo, and DC calculation. Using the preprocessed resting state fMRI data, we calculated the fALFF, ReHo, and DC values for each subject in DPABI v3.1. All values were calculated at voxel level. The fALFF value was calculated using the signal strength of the low frequency range (i.e., 0.01–0.1 Hz) to divide the detectable entire frequency range 22. The ReHo value, also called the Kendall’s coefficient of concordance (KCC), of the time series of a given voxel with its nearest neighbors (27 voxels were considered) was calculated to generate individual ReHo maps 23. The DC value of a certain voxel was the sum of the Pearson's correlation coefficients of all possible pairs of voxels (correlation threshold of r0 was set at 0.2) 24, 37. Lastly, the fALFF, ReHo, and DC maps for each subject were converted into Z-score maps, respectively. For the ReHo and DC maps, smoothing (FWHM = 4 mm) was performed.
Functional connectivity analysis. This analysis was performed in DPABI v3.1. The overlapped brain region, which showed significant group differences in fALFF, ReHo, and DC values between patients with BD and HC, was used as a seed in the following functional connectivity analysis. First, time series of each voxel was extracted within the seed. Second, the mean time series of the seed was calculated by averaging each voxel’s time series. Third, the Pearson correlation coefficients of the mean time series of the seed and all other voxels’ time series were calculated. They were used to construct each subject’s rsFC map. Finally, each subject’s rsFC map was converted into Z-score map.
2.6 Statistical Analysis
Two sample T test or chi-square tests were used to compared the differences between the patients with BD and the HC in demographic (including age, gender, education, and mean FD) and clinical (i.e. HAMD scores) variables in SPSS 20.0.
For each functional index (i.e., the fALFF, the ReHo, the DC, and the rsFC), analyses of the differences between the two groups were conducted in SPM12 (Wellcome Department of Cognitive Neurology, London, UK). In these analyses, two-sample T tests were used with age, gender, education, and mean FD controlled. The significance level was set at the combination of voxel-level uncorrected P < 0.001 and cluster-level Family Wise Error (FWE)-corrected P < 0.05. These statistical analyses were limited in a pre-made gray mask. The gray mask was calculated by the average of the gray mask segmented from each subjects’ T1 images.
2.7 Confirmation Analyses Of The Rsfc
To validate the findings obtained in the rsFC analysis, we used two strategies to confirm the findings. (i) We used the significant cluster obtained in the group analyses of the fALFF and the DC, but not the ReHo (due to the negative finding in a strict statistical threshold) to calculate the overlapped cluster. The overlapped cluster was then used as a seed in the rsFC analysis. (ii). We attempted to map our interested brain regions identified in the analysis of the fALFF, ReHo, and DC to the Brainnetome atlas 38. Then, we selected the corresponding brain region of the atlas as seeds for further rsFC analysis. This method reduces the subjective bias in selecting the seed region and makes our findings more likely to be replicated by future studies.