Increased functional connectivity between default mode network and visual network potentially correlates with duration of residual dizziness in patients with benign paroxysmal positional vertigo

Abstract

Objective: To assess changes in static and dynamic functional network connectivity (sFNC and dFNC) and explore their correlations with clinical features in benign paroxysmal positional vertigo (BPPV) patients with residual dizziness (RD) after successful canalith repositioning maneuvers (CRM) using resting-state fMRI.

Methods: We studied resting-state fMRI data from 39 BPPV patients with RD compared to 38 BPPV patients without RD after successful CRM. Independent component analysis and methods of sliding window and k-means clustering were adopted to investigate the changes in dFNC and sFNC between the two groups. Additionally, temporal features and meta-states were compared between the two groups. Furthermore, the associations between fMRI results and clinical characteristics were analyzed using Pearson's partial correlation analysis.

Results: Compared with BPPV patients without RD, patients with RD had longer duration of BPPV and higher scores of dizziness handicap inventory (DHI) before successful CRM. BPPV patients with RD displayed no obvious abnormal sFNC compared to patients without RD. In the dFNC analysis, patients with RD showed increased FNC between default mode network (DMN) and visual network (VN) in state 4, the FNC between DMN and VN was positively correlated with the duration of RD. Furthermore, we found increased mean dwell time (MDT) and fractional windows (FW) in state 1 but decreased MDT and FW in state 3 in BPPV patients with RD. The FW of state 1 was positively correlated with DHI score before CRM, the MDT and FW of state 3 were negatively correlated with the duration of BPPV before CRM in patients with RD. Additionally, compared with patients without RD, patients with RD showed decreased number of states and state span.

Conclusion: The occurrence of RD might be associated with increased FNC between DMN and VN, and the increased FNC between DMN and VN might potentially correlate with the duration of RD symptoms. In addition, we found BPPV patients with RD showed altered global meta-states and temporal features. These findings are helpful for us to better understand the underlying neural mechanisms of RD and potentially contribute to intervention development for BPPV patients with RD.

Keywords: benign paroxysmal positional vertigo; dynamic functional network connectivity; independent component analysis; residual dizziness; resting-state fMRI; static functional network connectivity.

Figure 1

Spatial maps of the 13 independent components (ICs) identified as resting state networks (RSNs) from the 25 ICs. DMN, default mode network; lFPN, left frontoparietal network; rFPN, right frontoparietal network; SN, salience network; AN, attention network; VN, visual network; SMN, sensorimotor network; AuN, auditory network.

Figure 2

Results of static functional network connectivity (sFNC; one sample t-test, p < 0.05, FDR corrected). (A) The mean sFNC matrices of all subjects in the 13 independent components. (B) The averaged sFNC matrices of all subjects in eight networks. DMN, default mode network; lFPN, left frontoparietal network; rFPN, right frontoparietal network; SN, salience network; AN, attention network; VN, visual network; SMN, sensorimotor network; AuN, auditory network.

Figure 3

Results of the clustering analysis for each state. Percentage corresponds to the time that all subjects dwell in a certain state, and the order of the state corresponds to the order of the introduction of k-means algorithm. DMN, default mode network; lFPN, left frontoparietal network; rFPN, right frontoparietal network; SN, salience network; AN, attention network; VN, visual network; SMN, sensorimotor network; AuN, auditory network.

Figure 4

Differences in dynamic functional network connectivity (dFNC) between patients with and without residual dizziness (RD; p < 0.05, FDR corrected). (A,B) represent median FNC matrices, in RD and non-RD, respectively. The number of participants displayed in median FNC matrices were subjects who showed correlation for at least 10 windows and were included in the data to determine group differences. (C) represents difference between A and B. Red means RD > non-RD, blue indicates RD < non-RD. (D) shows difference in z value of DMN (IC 1)-VN (IC 7) between RD and non-RD patients in state 4. DMN, default mode network; lFPN, left frontoparietal network; rFPN, right frontoparietal network; SN, salience network; AN, attention network; VN, visual network; SMN, sensorimotor network; AuN, auditory network.

Figure 5

Temporal properties and meta-states of dynamic functional network connectivity for patients with and without residual dizziness (RD). (A) Percentage of total time patients spent in each state; (B) Mean dwell time; (C) Number of transitions between states; (D) Number of states; (E) Change between states; (F) State span; (G) Total distance. Red five-pointed star represents significant difference between two groups (p < 0.05, FDR corrected).

Figure 6

Results of the correlations between dFNC and clinical characteristics in patients with RD (all p < 0.05, FDR corrected). RD, residual dizziness; DMN, default mode network; VN, visual network; IC, independent component; DHI, dizziness handicap inventory; CRM, canalith repositioning maneuvers. (A) The z-value between DMN (IC 1) and VN (IC 7) was positively correlated with the duration of RD (p = 0.007, r = 0.649, FDR corrected); (B) The fractional windows of state 1 was positively correlated with the DHI score before CRM (p = 0.021, r = 0.369, FDR corrected); (C) The mean dwell time of state 3 was negatively correlated with the duration of vertigo before CRM (p = 0.044, r = −0.370, FDR corrected); (D) The fractional windows of state 3 was negatively correlated with the duration of vertigo before CRM (p = 0.026, r = −0.405, FDR corrected).