Fast transient networks in spontaneous human brain activity.

To provide an effective substrate for cognitive processes, functional brain networks should be able to reorganize and coordinate on a sub-second temporal scale. We used magnetoencephalography recordings of spontaneous activity to characterize whole-brain functional connectivity dynamics at high temporal resolution. Using a novel approach that identifies the points in time at which unique patterns of activity recur, we reveal transient (100-200 ms) brain states with spatial topographies similar to those of well-known resting state networks. By assessing temporal changes in the occurrence of these states, we demonstrate that within-network functional connectivity is underpinned by coordinated neuronal dynamics that fluctuate much more rapidly than has previously been shown. We further evaluate cross-network interactions, and show that anticorrelation between the default mode network and parietal regions of the dorsal attention network is consistent with an inability of the system to transition directly between two transient brain states. DOI: http://dx.doi.org/10.7554/eLife.01867.001.

Non-Gaussian probabilistic MEG source localisation based on kernel density estimation.

There is strong evidence to suggest that data recorded from magnetoencephalography (MEG) follows a non-Gaussian distribution. However, existing standard methods for source localisation model the data using only second order statistics, and therefore use the inherent assumption of a Gaussian distribution. In this paper, we present a new general method for non-Gaussian source estimation of stationary signals for localising brain activity from MEG data. By providing a Bayesian formulation for MEG source localisation, we show that the source probability density function (pdf), which is not necessarily Gaussian, can be estimated using multivariate kernel density estimators. In the case of Gaussian data, the solution of the method is equivalent to that of widely used linearly constrained minimum variance (LCMV) beamformer. The method is also extended to handle data with highly correlated sources using the marginal distribution of the estimated joint distribution, which, in the case of Gaussian measurements, corresponds to the null-beamformer. The proposed non-Gaussian source localisation approach is shown to give better spatial estimates than the LCMV beamformer, both in simulations incorporating non-Gaussian signals, and in real MEG measurements of auditory and visual evoked responses, where the highly correlated sources are known to be difficult to estimate.

Tremor suppression by rhythmic transcranial current stimulation.

Tremor can dominate Parkinson's disease and yet responds less well to dopaminergic medications than do other cardinal symptoms of this condition. Deep brain stimulation can provide striking tremor relief, but the introduction of stimulating electrodes deep in the substance of the brain carries significant risks, including those of hemorrhage. Here, we pioneer an alternative approach in which we noninvasively apply transcranial alternating current stimulation (TACS) over the motor cortex to induce phase cancellation of the rest tremor rhythm. We first identify the timing of cortical oscillations responsible for rest tremor in the periphery by delivering tremor-frequency stimulation over motor cortex but do not couple this stimulation to the on-going tremor-instead, the rhythms simply "drift" in and out of phase alignment with one another. Slow alternating periods of phase cancellation and reinforcement result, informing on the phase alignments that induce the greatest change in tremor amplitude. Next, we deliver stimulation at these specified phase alignments to demonstrate controlled suppression of the on-going tremor. With this technique we can achieve almost 50% average reduction in resting tremor amplitude and in so doing form the basis of a closed-loop tremor-suppression therapy that could be extended to other oscillopathies.

An adaptive spectral estimation technique to detect cavitation in HIFU with high spatial resolution.

In ultrasound-guided high-intensity focused ultrasound (HIFU) therapy, the changes observed on tissue are subtle during treatment; some ultrasound-guided HIFU protocols rely on the observation of significant brightness changes as the indicator of tissue lesions. The occurrence of a distinct hyperechogenic region ("bright-up") around the focus is often associated with acoustic cavitation resulting in microbubble formation, but it may indicate different physical events such as larger bubbles from boiling (known to alter acoustic impedance) or sometimes lesion formation. A reliable method to distinguish and spatially localize these causes within the tissue would assist the control of HIFU delivery, which is the subject of this paper. Spectral analysis of the radio frequency (RF) signal underlying the B-mode image provides more information on the physical cause, but the usual techniques that are methods on the Fourier transform require a long series for good spectral resolution and so they give poor spatial resolution. This paper introduces an active spectral cavitation detection method to attain high spatial resolution (0.15 × 0.15 mm per pixel) through a parametric statistical method (ARMA modeling) used on finite-length data sets, which enables local changes to be identified more easily. This technique uses the characteristics of the signal itself to optimize the model parameters and structure. Its performance is assessed using synthesized cavitation RF data, and it is then demonstrated in ex vivo bovine liver during and after HIFU exposure. The results suggest that good spatial and spectral resolution can be obtained by the design of suitable algorithms. In ultrasound-guided HIFU, the technique provides a useful supplement to B-mode analysis, with no additional time penalty in data acquisition.

Demand driven deep brain stimulation: regimes and autoregressive hidden Markov implementation.

Deep brain stimulation is an increasingly prevalent surgical option in the treatment of a multitude of neurological conditions, most notably Parkinson's disease. The development of a neurofeedback device is driven primarily by stimulator habituation, surgical risk factors, the cost of battery replacement, and reported neuropsychiatric side-effects under prolonged chronic administration. Here we present two distinct regimes for stimulation delivery in chronic and acute symptomatic conditions, presented in the context of Parkinsonian bradykinesias and tremor. Implementation strategies are discussed with a focus on vector-autoregressive hidden Markov models for tremor prediction. Detection of simple motor actions versus tremor are compared in a preliminary performance analysis.

Application of a null-beamformer to source localisation in MEG data of deep brain stimulation.

In this paper, we present an analysis of magnetoencephalography (MEG) signals from a patient with whole-body chronic pain in order to investigate changes in neural activity induced by DBS. The patient is one of the few cases treated using DBS of the anterior cingulate cortex (ACC). Using MEG to reconstruct the neural activity of interest is challenging because of interference to the signal from the DBS device. We demonstrate that a null-beamformer can be used to localise neural activity despite artefacts caused by the presence of DBS electrodes and stimulus pulses. We subsequently verified the accuracy of our source localisation by correlating the predicted DBS electrode positions with their actual positions, previously identified using anatomical imaging. We also demonstrated increased activity in pain-related regions including the pre-supplementary motor area, brainstem periaqueductal gray and medial prefrontal areas when the patient was in pain compared to when the patient experienced pain relief.