Major depressive disorder (MDD) is the most common mood disorder in the world and the most important precursor of suicide. Despite decades of research, the pathophysiology of MDD remains not well understood. Recently, several MRI studies have focused on structural and functional connectivity evaluation and suggested that alterations of some specific regions of the brain, in both gray and white matter structures and some specific cortical–subcortical neuronal circuits, may play important roles of MDD [1, 2]. Generalized q-sampling imaging (GQI) is a more accurate and sophisticated diffusion MR approach compared to diffusion tensor imaging (DTI), which can extract additional information about the altered diffusion environments to resolve the complicated neural structure changes of neural disease [3]. In this study, we used GQI and graph theoretical analysis to evaluate brain structure and connectivity change of MDD compared to healthy controls and correlation with symptom severity.

Sixteen MDD patients (aged 44.81 ± 2.2 y/o) and 30 healthy controls (aged 45.03 ± 1.88 y/o) participated in accordance with Institutional Review Board of Chung Shan Medical University Hospital and underwent Hamilton Depression Rating Scale (HAMD), Hospital Anxiety and Depression Scale (HADS) evaluations and 1.5T MRI scans (Signa HDxt, GE Medical System, USA) with an 8-channel head coil. Diffusion imaging were obtained with TR/TE = 10,500/120 ms, resolution (voxel size) = 2 × 2 x 4 mm³, 76 non-collinear diffusion gradient directions with b-values from 0 to 2,000 s/mm².

In voxel-based statistical analysis (VBA), FSL-eddy correction (FMRIB, Oxford, UK) was performed first, followed by registration of the corrected images to the b0 (null) image then mapped the images to the standard T2 template using Statistical Parametric Mapping (SPM8). After the preprocessing procedure, GQI indices mappings, including generalized fractional anisotropy (GFA), normalized quantitative anisotropy (NQA) and isotropic value of the orientation distribution function (ISO), were reconstructed from multiple shells diffusion data using DSI studio. The differences of the GQI indices between MDD patients and healthy controls were evaluated with two-sample t-test, and the correlation between the GOI indices and the HAMD/HADS scores were examined with multiple regression using SPM8.

In graph theoretical analysis (GTA) and network based statistical analysis (NBS), the whole brain network was divided into 90 ROIs (based on the AAL template) and a 90x90 connectivity matrix was calculated with GQI tractography for each individual using DSI studio. Several topological measures, including cluster coefficient, local efficiency, characteristic path length and small-worldness, were calculated using Graph Analysis Toolbox (GAT) with different correlation thresholds (0.15 - 0.3 in 0.01 increment). The differences in network topology and regional network between groups were evaluated with two-sample t-test and non-parametric permutation test (1,000 repetitions).

In VBA, we revealed a decrease of GFA and NQA in left superior longitudinal fasciculus (SLF), a decrease of GFA in bilateral inferior fronto-occipital fasciculus (IFOF) and an increase of ISO in the bilateral frontal lobes of the MDD patients compared to the healthy controls (p < 0.01) (Fig. 1). Furthermore, we found a significant negative correlation between the HAMD score and NQA, as well as between the depression (of HADS) score and GFA in the corpus callosum (p < 0.01) (Fig. 2). The findings were consisted with previous studies and indicated that disruption of cortical–subcortical circuit integrity may be involved in the etiology and reflect severity of MDD [2, 4, 5].

In GTA, the MDD patients had trends of higher cluster coefficient (Fig. 3A), local efficiency (Fig. 3B) and characteristic path length (Fig. 3C) but no significant difference of small-worldness compared to the healthy subjects (Fig. 3D). Furthermore, in NBS, we revealed several disrupted connected sub-networks in the MDD patients, majorly in the bilateral frontal lobes (Fig. 4). The results suggested decreased global connective efficiency and bilateral frontal sub-networks connection but compensatory increased general local connectivity in MDD patients.

GQI indices can help to detect structural and connective abnormalities of MDD patients and these alterations are correlated with depressive severity. Meaningfully altered topological organization of structural connectivity network could be achieved based on GQI tractography, indicating disturbance of the optimal balance between local segregation and global integration in MDD patients. Meanwhile, decreased structural connectivity regarding regions of bilateral frontal lobes might facilitate understanding of the underlying mechanism in the disorder.