Genetic and neuroanatomical support for functional brain network dynamics in epilepsy

Abstract

Focal epilepsy is a devastating neurological disorder that affects an overwhelming number of patients worldwide, many of whom prove resistant to medication. The efficacy of current innovative technologies for the treatment of these patients has been stalled by the lack of accurate and effective methods to fuse multimodal neuroimaging data to map anatomical targets driving seizure dynamics. Here we propose a parsimonious model that explains how large-scale anatomical networks and shared genetic constraints shape inter-regional communication in focal epilepsy. In extensive ECoG recordings acquired from a group of patients with medically refractory focal-onset epilepsy, we find that ictal and preictal functional brain network dynamics can be accurately predicted from features of brain anatomy and geometry, patterns of white matter connectivity, and constraints complicit in patterns of gene coexpression, all of which are conserved across healthy adult populations. Moreover, we uncover evidence that markers of non-conserved architecture, potentially driven by idiosyncratic pathology of single subjects, are most prevalent in high frequency ictal dynamics and low frequency preictal dynamics. Finally, we find that ictal dynamics are better predicted by white matter features and more poorly predicted by geometry and genetic constraints than preictal dynamics, suggesting that the functional brain network dynamics manifest in seizures rely on -and may directly propagate alongunderlying white matter structure that is largely conserved across humans. Broadly, our work offers insights into the generic architectural principles of the human brain that impact seizure dynamics, and could be extended to further our understanding, models, and predictions of subject-level pathology and response to intervention.

Pairwise maximum entropy model explains the role of white matter structure in shaping emergent co-activation states

“…Since iEEG datasets from healthy controls do not exist, we cannot directly address these concerns. Nevertheless, we draw on recent work demonstrating that the patterns of iEEG functional connectivity in patients show statistical similarities to structural connectivity estimated in healthy volunteers 52 , and that this statistical similarity is upheld and even strengthened during ictal epochs 59,60 . Future work using data from patients with other pathologies, or using source-localized MEG in healthy patients, could be helpful in further understanding the nature of the structure-function correspondence accessible to the MEM.…”

A pairwise maximum entropy model uncovers the white matter scaffold underlying emergent dynamics in intracranial EEG

“…These findings are consistent with the notion that the pairwise MEM may be particularly sensitive to structurally-driven functional relations while conventional functional connectivity methods may be particularly sensitive to non-structurally-driven functional relations that might vary appreciably over short time intervals. Finally, the observed high degree of structure-function coupling suggests that structural connectivity is a useful proxy for time-invariant functional relationships Reddy et al, 2018). This observation could be useful in the treatment of epilepsy patients, where access to the brain is traditionally limited to recording loci but could be augmented with non-invasive measurements of structural connectivity for more informed surgical planning .…”

“…Our use of the power amplitude envelope to define the activation states is motivated by several factors. First, it has been shown that the BOLD fMRI signal echoes the envelope of high frequency neural activity (Logothetis et al, 2001;Winder et al, 2017), specifically when measured by iEEG (Lachaux et al, 2007;Jacques et al, 2016;Nir et al, 2008;Ojemann et al, 2013;Reddy et al, 2018). Thus, in light of studies linking BOLD fMRI FC to white matter SC (Honey et al, 2009;Watanabe et al, 2013), we hypothesize that the power of iEEG recordings should also exhibit a clear relationship with the underlying SC.…”

“…Since iEEG datasets from healthy controls do not exist, we cannot directly address these concerns. Nevertheless, we draw on recent work demonstrating that the patterns of ECoG functional connectivity in patients show statistical similarities to structural connectivity estimated in healthy volunteers , and that this statistical similarity is upheld and even strengthened during ictal epochs (Reddy et al, 2018;Shah 10/18 et al, 2018). Future work using data from patients with other pathologies, or using source-localized MEG in healthy patients, could be helpful in further understanding the nature of the structure-function correspondence accessible to the MEM.…”

How sensitive is functional connectivity to electrode resampling on intracranial EEG? Implications for personalized network models in drug-resistant epilepsy

“…This result is practically concerning in the case of metrics with lower reliability to spatial resampling, such as control centrality and synchronizability, because placement of electrodes in the network periphery away from clinical zones of interest is often variable across patients and across epilepsy centers. To increase the clinical confidence in the results of these network measures, the incomplete network may be supplemented using structural connectivity data and atlas-based approaches Greicius et al 2009;Liao et al 2011;Fan et al 2016;Betzel et al 2017;Reddy et al 2018) . Network theory also proposes several methods of predicting missing links (Lü et al 2015;Pan et al 2016;Lü and Zhou 2011;Guimerà and Sales-Pardo 2009) .…”