Neurological soft signs and structural network changes: a longitudinal analysis in first-episode schizophrenia

For the present study, we used data from a previous publication, in which we examined the longitudinal relationships between NSS and grey matter volume changes in first-episode schizophrenia [14]. Detailed information on the subjects’ characteristics has been given before [14]. They are exhibited in Table 1. Twenty participants (7 males) with first-episode schizophrenia were recruited from the Department of Psychiatry, University of Heidelberg, Germany. Patients’ diagnoses were established using the German version of the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [21]. Psychopathological symptoms were rated on the Positive and Negative Syndrome Scale (PANSS) [22]. None of the patients had a history of severe substance abuse or neurological disorder. All patients were again examined after a mean duration of 13.8 months (Standard Deviation, SD = 1.6). At baseline patients received atypical antipsychotics with a mean dose of 549.3 mg/day (SD = 271.3) of chlorpromazine equivalents (CPZ; [23]). Sixteen patients (80%) adhered to psychiatric treatment regularly with a mean dose of 443.8 mg/day CPZ (SD = 137.7) during the follow-up period. Twenty healthy individuals were recruited through advertisements and included in the study. The group comprised 10 males and 10 females with a mean age of 24.1 years (SD = 3.5) and 11.9 years (SD = 1.5) of education. None of the participants had any major psychotic disorders, history of neurological or medical illness, head injury or substance abuse. The study was approved by the ethics committee of the Medical Faculty, University of Heidelberg, Germany. All participants provided written informed consent after full explanation of the proceedings.

NSS were assessed at baseline and follow-up with the Heidelberg Scale [4], which is well established and widely used in neuropsychiatric diseases [24‐28]. It includes 5 subscales comprising 16 items. Ratings were given on a 0–3-point scale (no/slight/moderate/marked abnormality). In order to further investigate the changes of NSS over time, the whole patients group (20 patients) were divided into 2 subgroups based on the mean change of NSS total scores. There were 10 patients with a notable decrease in total NSS scores (NSS-decreasing subgroup: 3 males, 7 females; mean age: 26.1 years (SD = 6.5); change of NSS: 5 ~ 18) and 10 patients with persisting or increasing NSS level (NSS-persisting subgroup: 4 males, 6 females; mean age: 25.0 years (SD = 8.2); change of NSS: − 10 ~ 4). The changes of PANSS scores in both subgroups were also summarized in Table 1.

Structural MRI data of all participants were obtained at the German Cancer Research Centre with a 1.5 T Magnetom Vision MR scanner (Siemens Medical Solutions, Erlangen, Germany) with the following parameters: magnetization-prepared rapid gradient echo (MP-RAGE), 126 coronal slices, image matrix = 256 × 256, voxel size = 0.98 × 0.98 × 1.8 mm³, repetition time (TR) = 10 ms, echo time (TE) = 4 ms, flip angle = 12°.

In order to compare the present network results with our earlier grey matter volume findings from the same sample [14], we kept using the same preprocessing tools with VBM8 (http://​dbm.​neuro.​uni-jena.​de/​vbm) in SPM8 (http://​www.​fil.​ion.​ucl.​ac.​uk/​spm) on Matlab 2013 platform (http://​www.​mathworks.​com/​products/​matlab). Detailed steps of VBM8 have been described in previous studies (Ashburner and Friston, 2000; Good et al., 2001). Briefly, it included (1) normalization of all images to a Montreal Neurological Institute (MNI) template; (2) normalized brain images were segmented into grey matter, white matter, and cerebrospinal fluid compartments; and (3) normalized grey matter images were modulated.

The WFU PickAtlas Toolbox was used to extract 116 regions of interest (ROIs) using the Automated Anatomical Labeling (AAL) atlas [29]. These cerebral ROIs were re-sliced based on the normalized grey matter images. The REX toolbox (https://​web.​mit.​edu/​swg/​software.​htm) was then used to extract ROI volumes from the modulated and normalized grey matter images.

In the present study, the 116 extracted AAL ROIs were defined as nodes, and the Pearson correlation coefficients between ROIs represented the edges. Structural correlation matrices were then constructed by calculating the correlation coefficients between regions for each group. Age and gender were entered as covariates to control for potential confounding effects. Binary adjacency matrices were then derived based on the association matrices by thresholding a range of densities. The lower bound of the range was determined in which all nodes were fully connected. The upper bound of the range was 0.5 according to previous results, as grey matter structural networks became less biological above this bound [30, 31].

To investigate the global topological properties of a network, clustering coefficient (C) and characteristic path length (L) were calculated [31‐33]. The clustering coefficient of a node is a measurement of the number of connections between its directly neighboring regions, and the clustering coefficient of a network is the average of the clustering coefficient, which reflects the segregation of the network [31]. The characteristic path length is the average shortest passing through the node, which reflects the integration of the network. The small-world index of the network was defined as (C/C_rand)/ (L/L_rand), where C_rand and L_rand are the mean clustering coefficient and the characteristic path length of the random network [34]. A network with small-world index means that it has a significantly higher clustering coefficient than in random networks, and comparable characteristics path length to that in random networks [33].

Regional network characteristics of the constructed structural networks were assessed based on nodal betweenness centrality for each of the anatomical ROIs at minimum density with full connectivity for the network. Nodal betweenness represents the fraction of all shortest paths in the network that pass through, which can defect important anatomical or functional connections [35].

The hubs represent the importance of the nodes and are crucial regulators for efficient information communication in brain networks [31]. A node was set as a hub if its betweenness was at least two SDs above the mean betweenness [33, 36].

The clinical and demographic data were analyzed using Student’s t tests for continuous variables and Chi-squared tests for categorical variables in SPSS 22. Repeated measures ANOVA were performed to examine time effects and group effects for NSS and PANSS scores.

To investigate the differences of the global and regional network measures between groups, parametric t-tests and nonparametric tests with 1000 repetitions were performed using GAT software [33, 36]. After calculating group differences at each density, areas under curve (AUC) were then performed for each network measures so that the between-group differences were less sensitive to the thresholding values. In an exploratory fashion, we assessed changes in network characteristics over one-year follow-up between the two subgroups. Two-tailed p values were obtained based on its percentile position [33, 37], and false discovery rate (FDR) corrected p values were also reported to correct for multiple comparisons with p < 0.05.

The clinical and demographic characteristics of the sample are summarized in Table 1. There were no significant differences in age and sex between healthy controls and patients with schizophrenia. NSS scores in patients are significantly higher than in healthy controls (p < 0.001). Repeated measures ANOVA analysis showed a significant reduction of NSS in the decreasing-NSS subgroup compared to that in the persisting-NSS subgroup (p < 0.001). In addition, compared with the persisting-NSS subgroup, the decreasing-NSS subgroup tended to have a more favorable one-year outcome as indicated by scores on the PANSS; however, the differences between the two subgroups were not statistically significant.

There were no significant differences with respect to global network measures between patients with schizophrenia and healthy controls at baseline. AUC analysis showed that patients with schizophrenia demonstrated significant changes in local network properties and network hubs compared to controls mainly involving regions of cortical-subcortical-cerebellar circuits. Detailed information is described in the supplementary file (Figure S1 for local network properties and Figure S2 for network hubs).

Both NSS-persisting and NSS-decreasing subgroups followed small-world characteristics at baseline and follow-up. After the one-year follow-up, both subgroups did not show significant global network differences including the small-world index, clustering coefficient and path length.

AUC analysis showed that the NSS-persisting subgroup developed longitudinally significant reductions of local betweenness mainly involving right opercular inferior frontal cortex, left medial superior frontal cortex, left superior temporal cortex, right putamen and cerebellum vermis, while betweenness became larger in right lingual cortex (Fig. 1a). For the NSS-decreasing subgroup, patients demonstrated a significant smaller betweenness at follow-up in right middle temporal cortex, right insula and right fusiform, while betweenness became larger in left lingual cortex. However, none of the involved regions survived the correction for multiple comparisons (p < 0.05, FDR corrected) (Fig. 1b).

The network hubs of the NSS-persisting subgroup at baseline were identified in left orbital inferior frontal cortex, left middle temporal cortex, right putamen, and cerebellum vermis; after one-year follow-up, the network hubs were observed in left and right orbital inferior frontal cortices, left triangular inferior frontal cortex, left postcentral cortex and left and right middle temporal cortices (Fig. 2a). The network hubs of the NSS-decreasing subgroup at baseline were identified in left opercular and triangular inferior frontal cortices, right medial superior frontal cortex, left inferior temporal cortex, right fusiform and right cerebellum; after one-year of follow-up, the network hubs located in left middle and inferior temporal cortices, left angular and left cerebellum (Fig. 2b).