MRI researchers and clinicians who are interested in the field of diffusion imaging and brain connectivity.Decreased white matter integrity in patients with systemic lupus erythematosus (SLE) has been revealed in several previous researches [1, 2]. However, little is known about the changes of the brain structural connectivity in this patient group, especially those without neuropsychiatric symptoms (non-NPSLE). Diffusion tensor imaging (DTI) enables us to construct comprehensive structural networks throughout the entire brain by utilizing the white matter fiber tractography. The aim of this study was to investigate the alterations of the topological metrics in the structural networks of non-NPSLE patients based on a graph theoretical method.

Thirty-two adults diagnosed with non-NPSLE and 31 age- and gender-comparable healthy control subjects were recruited. DTI data acquisition was performed on a Philips 3.0-T Achieva MRI scanner with an 8-channel head coil for receiving by employing a single-shot spin-echo EPI sequence. Diffusion tensor imaging parameters included TR=9737 ms, TE=88 ms, 112×112 matrix, 224 mm×224 mm FOV, 75 axial contiguous 2.0 mm slices, 32 diffusion orientations with b-value=1000 s/mm2. 3D-T1WI data were acquired using gradient echo FFE sequence. The acquisition parameters included TR=9 ms, TE=4 ms, 256×256 matrix, 256 mm×256 mm FOV, a total of 176 sagittal 1.0 mm slices.

All acquired MRI images were processed using Pipeline for Analyzing braiN Diffusion imAges (PANDA) ( http://www.nitrc.org/projects/panda/) [3]. The raw DTI data were preprocessed by correcting the head movement and eddy current distortions, and the images were nonlinearly normalized into standard Montreal Neurological Institute (MNI) space. The 3D-T1MR images of each participant were segmented into 116 cortical and subcortical brain regions according to the automatic anatomical labeling (AAL) template. Then whole-brain deterministic tractography was performed to calculate the number of streamlines linking each pair of regions. This resulted in a 116×116 structural connectivity matrix for each subject.

Graph theoretical analysis was applied to investigate the alteration of the network connectivity between patient group and healthy controls. For all the generated connectivity matrixes, a series of network topological metrics including normalized clustering coefficient, normalized characteristic path length, small-world index, nodal betweenness centrality and nodal degree were calculated across a user-defined density range of 0.09-0.125 with an interval of 0.001 using graph analysis toolbox (GAT) [4]. After this, functional data analysis (FDA) in GAT was used to alleviate the sensitivity of the comparison to the thresholding process. A nonparametric permutation test was then conducted on the FDA results to reveal statistically significant differences. Lastly, the network hubs as the crucial components for efficient communication were quantified from FDA analysis of the betweenness and degree curves.

Figure 1 shows the global network measurements. The non-NPSLE group exhibited a tendency of decreased normalized clustering coefficient, decreased normalized path length and decreased small-world index across the final selected density range, though the results were not statistically significant. Figure 2 shows the regional network measurements using FDA analysis. Several brain regions with changed degree and betweenness were revealed. Specifically, Precuneus_L, Temporal_Pole_Sup_R, Cerebelum_Crus_L presented significantly higher betweenness in non-NPSLE group, while the Occipital_Inf_L and Temporal_Mid_R showed significantly lower betweenness in non-NPSLE group. Meanwhile, frontal_Mid_Orb_L, Precuneus_L, Putamen_R, Temporal_Pole_Sup_R presented significantly higher nodal degree in non-NPLSE group while Cingulum_Post_R, Temporal_Mid_R, Vermis_9 showed lower nodal degree in non-NPSLE group (Table1). However, note that none of these regions survived after the false discovery rate (FDR) correction. Seven and six hub nodes using betweenness, and four and five hub nodes using degree were identified for the non-NPSLE and HC groups, respectively (Table 2).In this study, we investigated the topological metric alteration in non-NPSLE patients through a graph theoretical analysis and three main findings were presented. The first major finding was the trend of decreased global network properties. This suggested a more segregated configuration of the structural networks in non-NPLSE patients. The second was the significantly changed betweenness centrality and degree in several brain regions of the patient group. This might facilitate our understanding of the neuropathological mechanism of non-NPSLE. The third was the identification of the hubs of the two groups. Overall, the graph theoretical analysis can be a promising method for analyzing the brain networks of systemic lupus erythematosus.