Frequency-Specific Analysis of the Dynamic Reconfiguration of the Brain in Patients with Schizophrenia

Abstract

The analysis of resting-state fMRI signals usually focuses on the low-frequency range/band (0.01−0.1 Hz), which does not cover all aspects of brain activity. Studies have shown that distinct frequency bands can capture unique fluctuations in brain activity, with high-frequency signals (>0.1 Hz) providing valuable information for the diagnosis of schizophrenia. We hypothesized that it is meaningful to study the dynamic reconfiguration of schizophrenia through different frequencies. Therefore, this study used resting-state functional magnetic resonance (RS-fMRI) data from 42 schizophrenia and 40 normal controls to investigate dynamic network reconfiguration in multiple frequency bands (0.01−0.25 Hz, 0.01−0.027 Hz, 0.027−0.073 Hz, 0.073−0.198 Hz, 0.198−0.25 Hz). Based on the time-varying dynamic network constructed for each frequency band, we compared the dynamic reconfiguration of schizophrenia and normal controls by calculating the recruitment and integration. The experimental results showed that the differences between schizophrenia and normal controls are observed in the full frequency, which is more significant in slow3. In addition, as visual network, attention network, and default mode network differ a lot from each other, they can show a high degree of connectivity, which indicates that the functional network of schizophrenia is affected by the abnormal brain state in these areas. These shreds of evidence provide a new perspective and promote the current understanding of the characteristics of dynamic brain networks in schizophrenia.

Keywords: dynamic reconfiguration; frequency-specific; multilayer network.

Figure 1

Schematic overview of analysis strategy. The resting-state fMRI data preprocessing results were first divided into 90 brain regions using the existing standard brain profile in the AAL template. Each brain region represented a node in the network. The brain’s physiological signals are decomposed into five frequency bands. The sliding window technique divides the time series into shorter time intervals. The functional connections within each layer are estimated using Pearson correlation. They connect the same nodes in adjacent periods and build a multilayer network for each participant. The dynamic community structure is detected by maximizing the multilayer modular quality function. We calculated the module allegiance matrix, recruitment, and integration to analyze the difference between NC and SZ.

Figure 2

The difference of recruitment and integration from full frequency (0.01–0.25 Hz), slow5 (0.01–0.027 Hz), slow4 (0.027–0.073 Hz), slow3 (0.073–0.198 Hz), and slow2 (0.198–0.25 Hz) in the whole brain level between NC and SZ. (A) Integration. (B) Recruitment. Asterisk indicates pairwise group differences; * represents p < 0.05; ** represents p < 0.01.

Figure 3

The difference of integration and recruitment from full frequency (0.01–0.25 Hz), slow5 (0.01–0.027 Hz), slow4 (0.027–0.073 Hz), slow3 (0.073–0.198 Hz), and slow2 (0.198–0.25 Hz) in the RSN level between NC and SZ. (A) Integration. (B) Recruitment. Asterisk indicates pairwise group differences; *** represents p < 0.001, ** represents p < 0.01, * represents p < 0.05.

Figure 4

Matrix of module allegiance in NC and SZ. (A) full frequency (0.01–0.25 Hz). (B) slow3 (0.073–0.198 Hz). P represents the mean value of module allegiance. The main diagonal represents the recruitment coefficient, and the upper/lower triangle represents the integration coefficient.

Figure 5

Group differences of integration for each pair of RSN to RSN in slow3 (0.073–0.198 Hz). (A) the integration between Visual network(VN) and Attention network(AN). (B) the integration between Visual network(VN) and Default mode network(DMN). Asterisk indicates pairwise group differences; *** denotes p < 0.001, ** denotes p < 0.01.

Figure 6

Difference between the NC and SZ in node vulnerability at full frequency (0.01–0.25 Hz) and slow3 (0.073–0.198 Hz). (A) Integration. (B) Recruitment.

Figure 7

Spearman correlations between dynamic properties and ASRS scores in slow3 (0.073–0.198 Hz). (A) Recruitment (B) Recruitment in the attention network (AN). (C) Recruitment in the inferior frontal gyrus, triangular part (IFGtriang) and (D) middle front gyrus (MFG).