Modulation of limbic resting-state networks by subthalamic nucleus deep brain stimulation

Abstract

Beyond the established effects of subthalamic nucleus deep brain stimulation (STN-DBS) in reducing motor symptoms in Parkinson’s disease, recent evidence has highlighted the effect on non-motor symptoms. However, the impact of STN-DBS on disseminated networks remains unclear. This study aimed to perform a quantitative evaluation of network-specific modulation induced by STN-DBS using Leading Eigenvector Dynamics Analysis (LEiDA). We calculated the occupancy of resting-state networks (RSNs) in functional MRI data from 10 patients with Parkinson’s disease implanted with STN-DBS and statistically compared between ON and OFF conditions. STN-DBS was found to specifically modulate the occupancy of networks overlapping with limbic RSNs. STN-DBS significantly increased the occupancy of an orbitofrontal limbic subsystem with respect to both DBS OFF(p = 0.0057) and 49 age-matched healthy controls (p = 0.0033). Occupancy of a diffuse limbic RSN was increased with STN-DBS OFF when compared to healthy controls (p = 0.021), but not when STN-DBS was ON, which indicates rebalancing of this network. These results highlight the modulatory effect of STN-DBS on components of the limbic system, particularly within the orbitofrontal cortex, a structure associated with reward processing. These results reinforce the value of quantitative biomarkers of RSN activity in evaluating the disseminated impact of brain stimulation techniques and the personalization of therapeutic strategies.

Metastability as a candidate neuromechanistic biomarker of schizophrenia pathology

“…In order to identify recurrent spatiotemporal patterns of phase-locking-henceforth called 'LEiDA modes'-we performed k-means clustering on the phase-locked time-series of each of the datasets that were analyzed (HCPEP CONx4, HCPEP NAPx4, Cobre CONx1, Cobre SCHZx1, see Materials and methods). This is similar to a previous study [47], but different from other studies that used LEiDA where k-means clustering was either performed on concatenated datasets across groups [54][55][56] or where the centroids extracted from one group were used to seed the clustering of other groups [57][58][59]. The approach in this study considers each dataset as a unique observation of brain activity with associated variability in the spatiotemporal modes and avoids data leakage which occurs when dimensionality reduction is performed on the dataset as a whole [60].…”

“…To identify recurring spatiotemporal modes ψ or phase-locking patterns, we clustered the leading eigenvectors for each of the 10 phase-locked time-series datasets (HCPEP:CON x 4 runs, HCPEP:NAP x 4 runs, Cobre:CON x 1 run, Cobre:SCHZ x 1 run) with K-means clustering with 300 replications and up to 400 iterations for 2-10 centroids. This approach is similar to a previous study [47] but different from other studies that used LEiDA where k-means clustering was either performed on concatenated datasets across groups [54][55][56], or where the centroids extracted from one group were used to cluster other groups [57][58][59]. This approach considers each dataset as a unique observation of brain activity with associated variability in the spatiotemporal modes and avoids data leakage [60].…”

A new causal centrality measure reveals the prominent role of subcortical structures in the causal architecture of the extended default mode network

“…The advantage of using causal centrality -as a proxy for controllability -would be in the computational efficiency of the former by virtue of Bayesian model reduction. Further research is required to formalize the relationship between causal centrality and controllability metrics in neuronal networks, and to study the implications for identifying therapeutic stimulation targets (Alosaimi et al 2022;Eraifej et al 2023;Ezzyat et al 2018;Georgiev et al 2021;Wang et al 2022;Zangen et al 2023).…”