Low consistency of four brain connectivity measures derived from intracranial electrode measurements.

Stephen E Jones,Imad Najm,Erik B Beall,Ken E Sakaie,Jorge A Gonzalez-Martinez,Michael D Phillips,Myron Zhang

doi:10.3389/fneur.2014.00272

Abstract

Measures of brain connectivity are currently subject to intense scientific and clinical interest. Multiple measures are available, each with advantages and disadvantages. Here, we study epilepsy patients with intracranial electrodes, and compare four different measures of connectivity. Perhaps the most direct measure derives from intracranial electrodes; however, this is invasive and spatial coverage is incomplete. These electrodes can be actively stimulated to trigger electrophysical responses to provide the first measure of connectivity. A second measure is the recent development of simultaneous BOLD fMRI and intracranial electrode stimulation. The resulting BOLD maps form a measure of effective connectivity. A third measure uses low frequency BOLD fluctuations measured by MRI, with functional connectivity defined as the temporal correlation coefficient between their BOLD waveforms. A fourth measure is structural, derived from diffusion MRI, with connectivity defined as an integrated diffusivity measure along a connecting pathway. This method addresses the difficult requirement to measure connectivity between any two points in the brain, reflecting the relatively arbitrary location of the surgical placement of intracranial electrodes. Using a group of eight epilepsy patients with intracranial electrodes, the connectivity from one method is compared to another method using all paired data points that are in common, yielding an overall correlation coefficient. This method is performed for all six paired-comparisons between the four methods. While these show statistically significant correlations, the magnitudes of the correlation are relatively modest (r2 between 0.20 and 0.001). In summary, there are many pairs of points in the brain that correlate well using one measure yet correlate poorly using another measure. These experimental findings present a complicated picture regarding the measure or meaning of brain connectivity.

Highlights

The importance of brain connectivity is self-evident given the underlying network structure of the brain
Examples of functional connectivity include measures derived from scalp EEG, intracranial EEG, PET studies, cortical thickness studies, task-related fMRI, and resting state fMRI
Right orbito-frontal stimulation and BOLD fMRI. These four modalities of connectivity are compared on a pairwise basis, and we show that the comparisons reveal statistically significant correlations, the correlation values are modest

Summary

Introduction

The importance of brain connectivity is self-evident given the underlying network structure of the brain. Examples of functional connectivity include measures derived from scalp EEG, intracranial EEG, PET studies, cortical thickness studies, task-related fMRI, and resting state fMRI (rsfMRI). All these methods have advantages and disadvantages, related to spatial and temporal resolution, coverage, effectiveness, and invasiveness. Given numerous methodologies and metrics of connectivity [1, 2], it is natural to compare them with the hypothesis that if the metrics are sensitive to the underlying network architecture of the brain, the connectivity measures should strongly correlate with each other This inquiry raises the question about the exact definition of connectivity, and what could be considered the “gold standard.”

Methods

Results

Discussion

Conclusion