Abstracts

Rationale: Magnetoencephalography (MEG) is used for presurgical language lateralization. An auditory word recognition task is the most employed although a battery of tasks has been proposed similar to fMRI. Functional connectivity (FC) measures relationships between cerebral signals over time and potentially allows assumptions to be made regarding functional interactions between brain regions. Despite its promise, FC still has little utility in clinical practice largely because of unfamiliarity of clinicians of its computational basis and lack of an easily applied application that can be easily employed. We propose a GUI network connectivity analysis toolbox that is designed to study language network connectivity

Methods: GUI Analysis pipeline has the following features: 1) Data preprocessing allows band-pass, notch filtering and artefact rejection on MEG sensor data and exporting isotropic MRI volume in freesurfer format. 2) Cortical sheets extraction using MNE-C packages helps to establish number of cortical sources at baseline and MRI-MEG coregistration. A subject-specific head model, channel positions and MEG cortical sheets are used to compute the forward solution. 3) source reconstruction using Linearly Constrained Minimum Variance (LCMV) beamformer to estimate the magnitude of magnetic activity at different locations in the cortex. ROIs related to brain language circuit (inferior frontal, superior temporal and supramarginal gyri) are extracted and activation time series from these regions are calculated. 4) Quantitative FC using partial directed coherence (PDC) is calculated to infer the intensity of information flow over the selected ROIs in the time and frequency domains. PDC puts more emphasis on the receiving node hence might be more adapted to study language networks. Laterality Index (LI) is used to determine the dominant language lateralization based on ROIs total inflow and outflow (fig 1).

Results: We tested the algorithm on five patients with intractable epilepsy who underwent presurgical MEG language mapping (four adults and one child) using an auditory word recognition task. Data was band pass filtered at 5-25 Hz and PDC was calculated using sources in the 200-600 milliseconds time post stimulus onset. Laterality Index (LI) results of all 5 subjects showed left sided language dominance confirmed by fMRI results in all five. The results are displayed in the figures showing connection maps, the maximum strength of connections between the sources, maximum total inflow and total outflow on 3D cortical surface. Different colors are used to display connectivity strength for interactions and information flow (fig 1) and laterality indices (fig 2).

Conclusions: We propose a new toolbox to visualize brain language networks. In addition, this tool has the potential to investigate and compare FC measures in different cognitive networks. Additional functionalities and quantitative measures can be easily incorporated. The advantages of this tool are: (a) simple and easy to interact graphical user interface (b) optimized time usage.

Funding: Please list any funding that was received in support of this abstract.: The study was funded through Pediatric KPRI grant and NSF-EPSCOR # 1632891.