Tracking cortical activity from M/EEG using graph cuts with spatiotemporal constraints

Encoding cortical dynamics in sparse features

Distributed cortical solutions of magnetoencephalography (MEG) and electroencephalography (EEG) exhibit complex spatial and temporal dynamics. The extraction of patterns of interest and dynamic features from these cortical signals has so far relied on the expertise of investigators. There is a definite need in both clinical and neuroscience research for a method that will extract critical features from high-dimensional neuroimaging data in an automatic fashion. We have previously demonstrated the use of optical flow techniques for evaluating the kinematic properties of motion field projected on non-flat manifolds like in a cortical surface. We have further extended this framework to automatically detect features in the optical flow vector field by using the modified and extended 2-Riemannian Helmholtz–Hodge decomposition (HHD). Here, we applied these mathematical models on simulation and MEG data recorded from a healthy individual during a somatosensory experiment and an epilepsy pediatric patient during sleep. We tested whether our technique can automatically extract salient dynamical features of cortical activity. Simulation results indicated that we can precisely reproduce the simulated cortical dynamics with HHD; encode them in sparse features and represent the propagation of brain activity between distinct cortical areas. Using HHD, we decoded the somatosensory N20 component into two HHD features and represented the dynamics of brain activity as a traveling source between two primary somatosensory regions. In the epilepsy patient, we displayed the propagation of the epileptic activity around the margins of a brain lesion. Our findings indicate that HHD measures computed from cortical dynamics can: (i) quantitatively access the cortical dynamics in both healthy and disease brain in terms of sparse features and dynamic brain activity propagation between distinct cortical areas, and (ii) facilitate a reproducible, automated analysis of experimental and clinical MEG/EEG source imaging data.

Sparse MEG source imaging for reconstructing dynamic sources of interictal spikes in partial epilepsy

PURPOSE

The present study aimed to test the feasibility of a novel neuroimaging technique, that is, variation-based sparse cortical current density (VB-SCCD) imaging algorithm, in noninvasively estimating location and extent of epileptic sources from interictal magnetoencephalography (MEG) data.

METHODS

A total of 108 interictal spikes from 3 partial epilepsy patients were selected to perform VB-SCCD source analysis. Cortical sources were identified at spike peaks, rising phases, and entire spikes, respectively, from all interictal spikes in each patient, to estimate source locations and extents, and validated using presurgical evaluation data. Other source analysis methods, that is, minimum norm estimate and sparse source imaging were also performed for comparison.

RESULTS

Cortical sources reconstructed by VB-SCCD that are consistent with clinical presurgical evaluation outcomes have detection rates of 65.8% at spike peaks, 85.1% during rising phases, and 92.6% in entire spikes. Stable spatiotemporal patterns of reconstructed cortical sources were also obtained using VB-SCCD, which provide more insights about the formation and propagation of interictal epileptic activity.

CONCLUSIONS

Our present results suggest that the VB-SCCD technique has the capability in estimating location and extent of epileptic sources of interictal spikes and is promising to become a valuable noninvasive tool in assisting presurgical planning for partial epilepsy patients.

Three-dimensional interactive and stereotactic human brain atlas of white matter tracts

We present a human brain atlas of white matter (WM) tracts containing 40 major tracts, which is three-dimensional (3D), segmented, labeled, interactive, stereotactic and correlated to structure and vasculature. We consider: (1) WM accuracy by correlating WM tracts to underlying neuroanatomy and quantifying them; (2) balance between realism and completeness by processing a sequence of track volumes generated for various parameters with the increasing track number to enable a tract "shape convergence". MPRAGE and DTI in 64 directions of the same subject were acquired on 3 Tesla. The method has three steps: DTI-MPRAGE registration, 3D tract generation from DTI, to WM reconstruction from MPRAGE to parcellation into 17 components. 82 track volumes were generated for a wide spectrum of parameter values: Fractional Anisotropy threshold in [0.0125, 0.55] and trajectory angle lower than 45°, 60°, 65°, 70°, 75°, 80°, 85°, 90°. For each tract, a sequence of track volumes was processed to create/edit contours delineating this tract to achieve its shape convergence. The parcellated tracts were grouped into commissures, associations, projections and posterior fossa tracts, and labeled following Terminologia Anatomica. To facilitate that, a dedicated tract editor is developed which processes multiple track volumes, handles tracts in three representations (tracks, contours, envelopes); provides editing/visualization simultaneously on axial, coronal, sagittal planes; enables tract labeling and coloring; and provides numerous tools (track counting, smoothing and length thresholding; representation conversion and saving; structural atlas support). A stereotactic tract atlas along with parcellated WM was developed to explore in real-time any individual tract or their groups along with surrounding neuroanatomy.

Mixed-norm estimates for the M/EEG inverse problem using accelerated gradient methods

Magneto- and electroencephalography (M/EEG) measure the electromagnetic fields produced by the neural electrical currents. Given a conductor model for the head, and the distribution of source currents in the brain, Maxwell's equations allow one to compute the ensuing M/EEG signals. Given the actual M/EEG measurements and the solution of this forward problem, one can localize, in space and in time, the brain regions that have produced the recorded data. However, due to the physics of the problem, the limited number of sensors compared to the number of possible source locations, and measurement noise, this inverse problem is ill-posed. Consequently, additional constraints are needed. Classical inverse solvers, often called minimum norm estimates (MNE), promote source estimates with a small ℓ₂ norm. Here, we consider a more general class of priors based on mixed norms. Such norms have the ability to structure the prior in order to incorporate some additional assumptions about the sources. We refer to such solvers as mixed-norm estimates (MxNE). In the context of M/EEG, MxNE can promote spatially focal sources with smooth temporal estimates with a two-level ℓ₁/ℓ₂ mixed-norm, while a three-level mixed-norm can be used to promote spatially non-overlapping sources between different experimental conditions. In order to efficiently solve the optimization problems of MxNE, we introduce fast first-order iterative schemes that for the ℓ₁/ℓ₂ norm give solutions in a few seconds making such a prior as convenient as the simple MNE. Furthermore, thanks to the convexity of the optimization problem, we can provide optimality conditions that guarantee global convergence. The utility of the methods is demonstrated both with simulations and experimental MEG data.