Center for Bioelectric Interfaces participates on the BIOMAG-2018 conference
Dmitrii Altukhov and Aleksandra Kuznetsova present two posters on the leading international conference on biomagnetism BIOMAG-2018, August 26-30, Philadelphia.
Abstract:
For years now connectivity analysis has proven to be a valuable tool in the analysis of MEG and EEG data and is widely reported in the literature. Brain networks are known to exhibit various types of coupling scenarios. Zero- phase coupling, a type of within frequency synchrony, reliably detected with invasive electrophysiology, has remained elusive to the non-invasive functional imaging techniques such as EEG and MEG due to the spatial leakage effect. Most of the existing approaches for within frequency coupling detection either follow the path suggested in Nolte’s seminal imaginary coherency paper or don’t address the spatial leakage issue at all. The latter is acceptable in connectivity estimation using intracranial recordings but leads to a large number of falsely detected networks when used with non-invasive measurements such as EEG or MEG.
Recently, we proposed a framework for correction of the spatial leakage effect based on the observation that the sensor space cross-spectrum can be represented as an additive mixture of the contributions from the spatial leakage and the genuine zero-phase coupling subspaces. We have shown that while these two subspaces overlap it is possible to devise a projection operator to significantly reduce the contribution of the spatial leakage to the sensor-space cross-spectrum and at the same time largely preserve the genuine zero-phase coupling subspace components.
Here we suggest an improvement to the technique proposed earlier based on the optimal unmixing of the spatial leakage subspace and the subspace spanned by zero-phase-coupled networks before the projection. Such unmixing is achieved via the preliminary whitening of the sensor space cross-spectrum against the genuine zero-phase coupling contributions followed by the projection away from the spatial leakage subspace. This oblique projection approach dubbed as ObPSIICOS, allows to achieve the best possible suppression of the power from the spatial leakage subspace with minimal distortion to the genuine zero-phase coupling components.
Our analysis shows that the proposed optimal scheme allows to achieve more than 3-fold improvement in the mean signal-to-noise ratio for the genuine zero- phase coupling signal in contrast to the original PSIICOS technique. As our realistic MEG Monte-Carlo simulations show thus increased SNR leads to a significantly improved detector characteristics of the novel method as compared to the PSIICOS formulation.
The improvement brought about by this new procedure depends on the mutual configuration of the spatial leakage and the genuine zero-phase coupling subspaces which is determined by the MEG probe design and can be predicted by the Riemannian distance between the two subspaces. We show that this quantity significantly varies across the MEG probes of different manufacturers of the modern MEG systems and, given the growing interest of the community to MEG as a technique in general and to the non-invasive connectivity studies in particular, we speculate that his new metrics can serve as an additional MEG probe optimization criterion used to create future generations of MEG systems.
Aleksandra Kuznetsova has a poster devoted to the new method of MEG inverse problem solution "MEG based functional microscopy using traveling wave priors".
Abstract:
Localization of epileptogenic regions in patients with pharmacoresistant epilepsy is one of the major clinical applications of magnetoencephalography. Advances in neurosurgery, including the development of non-invasive or minimally invasive techniques near the day when epilepsy surgery will become an ambulatory non-invasive procedure. To complete the loop, approaches for non-invasive high fidelity localization of epileptogenic regions have to be developed. Not only the location of the epileptogenic zone, but also the fine spatial-temporal dynamics, including the dominant direction and speed of local activity propagation is of interest. In the future, this information may help to profile maximally sparing resection of the neural tissue to minimize the post-surgical deficits and enable the operations in close to eloquent cortex locations.
To address this need we employed local cortical wave model in order to regularize the MEG inverse problem and developed a novel technique for localization of interictal spikes and unraveling their subtle spatial-temporal dynamics. We describe spike cortical activity as a linear combination of multiple traveling waves characterized by propagation speed and direction. Using LASSO technique we seek to find a minimal configuration of such waves to describe the non-invasively observed interictal spike MEG data. Analysis of the obtained solutions allows to establish the dominant propagation direction, explore various local propagation scenarios and perform more accurate localization.
We have explored the localization accuracy of the proposed technique in comparison to more conventional methods such as dipole fitting and MNE estimation. Application of the proposed methodology to analysis of real interictal spikes from a patient with multi-focal epilepsy allowed to determine physiologically plausible value for propagation speed of 0. 1m/s and establish local propagation direction explaining spike topography rotation observed in the sensor-space data.
In combination with the recent development of MEG instrumentation and advances in forward modeling we believe that the proposed approach may serve as a significant step towards dominantly non-invasive diagnostics and treatment of patients with pharmacologically intractable epilepsy.
Date
20 July
2018