Characterization Of White Matter Network Topology And Its Correlation With Clinical Features In Patients With Depressive Disorder Based On Diffusion Tensor Imaging

Objective: In order to explore the pathogenesis of depressive disorders, we investigated the alterations of the global and regional topological characteristics of the brain white matter networks of unipolar depression patients. Futher, we explored the differences of the topological organization of the brain white matter networks between the unipolar depression(UD) patients and the healthy control(HC) subjects, between the first-episode(FD) and recurrent(RD) depressed patients, and between the UD patients and bipolar depression(BD) patients. Then, we also explored the relationships between the abnormal brain topological features and the clinical characteristics of patients.Methods: Depressive disorder patients and age-, sex- and education-matched healthy controls were recruited to complete this study. All subjects were canned with 3.0 T structural magetic resonance imaging(s MRI) scanner and diffusion tensor imaging(DTI) scanner, and the behavioral data and clinical data of all subjects were collected on the day of scanning. 17-item Hamilton Rating Scale for Depression(HAMD-17) were used to evaluate the patient’s condition. Firstly, all imaging data of subjects were pre-processed with the FSL software package, including the correction of head motion and eddy current. Then, each entire brain was parcellated into 90 cortical regions by using the automated anatomical labeling(AAL) template and the fiber tracking was performed in the whole cerebral cortex of each subject to reconstruct white matter tracts using the fiber assignment by continuous tracking algorithm(FACT). Lastly, the brain white matter networks were constructed using the complex network theory. We use the following methods to explore:(1) From the perspective of the global network properties, the differences of the small-worldness, average clustering coefficient, average characteristic path length, betweenness centrality between the UD patients and HC subjects were examined by two sample t-tests and then the correlation between the metrics of the significant node and the course of the disease was explored by the Pearson correlation analysis.(2) From the perspective of the regional network properties, the differences of the nodal strength, clustering coefficient and local efficiency between the UD patients and HC subjects were examined by two sample t-tests and then the correlation between the metrics of the significant node and the severity of the disease was explored by the Pearson correlation analysis.(3) From the perspective of the the efficiency, the differences of the efficiency of the networks and the nodes between the UD patients and HC subjects were examined by two sample t-tests and then the correlation between the metrics of the significant node and the severity of the disease was explored by the Pearson correlation analysis.(4) Further, all patients were divided into FD and RD patients groups. Then, the differences of the nodal degree among the brain white matter networks of FD, RD patients and HC subjects were examined by two sample t-tests respectively and then the correlations between the metrics of the significant node and the course and the number of episodes of the disease were explored by the Pearson correlation analysis.(5) In addition, the differences of the betweenness centrality of the brain white matter networks between the BD patients and HC subjects were examined by two sample t-tests and then the correlation between the metrics of the significant node and the severity of the disease was explored by the Pearson correlation analysis.(6) Lastly, the patients were divided into UD and BD patients groups. Then, the differences of the nodal global efficiency of the brain networks of UD, BD and HC were compared by the one-way anova and two sample t-tests, and then the correlations between the global efficiency of the significant nodes and the clinical features of patients were explored by the Pearson correlation analysis.To control for multiple comparisons, a significance threshold of p<0.05 after false discovery rate(FDR) correction was used.Results:(1) The small-worldness and the distribution of hubs of UD and its relationship with the courseThe brain structural networks had small-world properties in both groups. When compared with HC, the betweenness centrality of the nodes of the networks in UD descent significantly in right superior frontal gyrus(orbital part)(P = 0.03, FDR corrected), and left putamen(P = 0.02, FDR corrected). Significant negative correlation was found between the betweenness centrality of left hippocampus and the course in UD.(2) The correlation between the local topological properties and the severity of the disease of UDThe local efficiency of the nodes of the networks in UD descent significantly in the left middle frontal gyrus(orbital part)(P = 0.03, FDR corrected), the left hippocampus(P = 0.02, FDR corrected), and the right parahippocampal gyrus(P = 0.02, FDR corrected) when compared with HC; and the clustering coefficient of the left middle frontal gyrus(orbital part) of the networks in UD descent significantly when compared with HC(P = 0.00, FDR corrected). Significant negative correlation was found between the local efficiency of the left middle frontal gyrus(orbital part) and the total scores of HAMD-17 in UD(r =-0.48, P = 0.02).(3) The efficiency of the brain networks and its relationship with the severity of disease in UDThe brain networks of both UD and HC exhibited a much higher local efficiency and a similar global efficiency when compared with the matched random networks. But the average global efficiency of the brain networks of UD descent significantly when compared with HC(P = 0.02). And the global efficiency of the right superior frontal gyrus(orbital part)(P = 0.04, FDR corrected) and middle temporal gyrus(temporal pole)(P = 0.03, FDR corrected) in the networks of UD descent significantly when compared with HC. Significant negative correlation was found between the global efficiency of the right superior frontal gyrus(orbital part) and the total scores of HAMD-17(r =-0.46, P = 0.02).(4) The degree of the brain networks and its relationship with the number of episodes in FD and RD patientsThe degree of the left putamen in the networks of FD descent significantly when compared with HC(P = 0.01, FDR corrected). That of RD descent significantly in the right dorsolateral superior frontal gyrus(P = 0.03, FDR corrected) and putamen(P = 0.02, FDR corrected) when compared with HC. That of RD descent significantly in the right dorsolateral superior frontal gyrus when compared with FD(P = 0.04, FDR corrected). The distribution of the hub regions in the patient and healthy was similar and the left putamen was the same hub region for both groups. Significant negative correlation was found between the degree of the right dorsolateral superior frontal gyrus and the number of episodes(r =-0.42; P = 0.00) and the course of the disease(r =-0.41; P = 0.00) in the patient.(5) The betweenness centrality of the brain networks in BD and its relationship with the severity of diseaseThe betweenness centrality of BD decreased significantly in the left superior frontal gyrus(dorsolateral)(P = 0.03, FDR corrected), inferior frontal gyrus(orbital part)(P = 0.03, FDR corrected) and anterior cingulate gyri(P = 0.03, FDR corrected), however, that of the right caudate nucleus increased significantly(P = 0.02, FDR corrected) when compared with HC. The distribution of hub regions in the left frontal and occipital lobes of BD was difference from that of HC. Significant negative correlation was found between the betweenness centrality of the left superior frontal gyrus(dorsolateral) and the total scores of 17-item Hamilton Rating Scale for Depression(r =-0.52; P = 0.03).(6) The global efficiency of the brain networks in UD and BD and its relationship with the clinical featuresThe significant nodes of the global efficiency in there groups included the left anterior cingulate gyri(P = 0.00, FDR corrected) and right medial superior frontal gyrus(P = 0.03, FDR corrected), caudate nucleus(P = 0.03, FDR corrected), pallidum(P = 0.03, FDR corrected). The global efficiency of UD descent significantly in the left medial superior frontal gyrus(P = 0.02, FDR corrected), thalamus(P = 0.03, FDR corrected) and right medial superior frontal gyrus(P = 0.03, FDR corrected), caudate nucleus(P = 0.00, FDR corrected) when compared with HC. The global efficiency of BD descent significantly in the left medial superior frontal gyrus(P = 0.00, FDR corrected) and anterior cingulate gyri(P = 0.00, FDR corrected), however, increased significantly in the right caudate nucleus(P = 0.04, FDR corrected) when compared with HC. The global efficiency of BD descent significantly in the left anterior cingulate gyri(P = 0.02, FDR corrected), however, increased significantly in the right caudate nucleus(P = 0.03, FDR corrected) and pallidum(P = 0.04, FDR corrected) when compared with UD. Significant negative correlation was found between the global efficiency of the left anterior cingulate gyri and the scores of anxiety/somatization factor(r =-0.43; P = 0.04) in UD. That were also found between the global efficiency of the right medial superior frontal gyrus and the scores of anxiety/somatization(r =-0.49; P = 0.04), and between the caudate nucleus and the retardation factor(r =-0.56; P = 0.02) in BD. However, significant positive correlation was found between the global efficiency of the right medial superior frontal gyrus and the age of the onset(r = 0.51; P = 0.04) in BD.Conclusions: Our findings suggested that both the global and regional topological characteristics of the brain white matter networks of depressive patients were impaired and these abnormal brain functional characteristics were associated with the clinical characteristics. It may be an important part of underlying mechanisms for pathogenesis of depressive disorder with different clinical characteristics and different subtypes.