Contextual MEG and EEG Source Estimates Using Spatiotemporal LSTM Networks.

Most magneto- and electroencephalography (M/EEG) based source estimation techniques derive their estimates sample wise, independently across time. However, neuronal assemblies are intricately interconnected, constraining the temporal evolution of neural activity that is detected by MEG and EEG; the observed neural currents must thus be highly context dependent. Here, we use a network of Long Short-Term Memory (LSTM) cells where the input is a sequence of past source estimates and the output is a prediction of the following estimate. This prediction is then used to correct the estimate. In this study, we applied this technique on noise-normalized minimum norm estimates (MNE). Because the correction is found by using past activity (context), we call this implementation Contextual MNE (CMNE), although this technique can be used in conjunction with any source estimation method. We test CMNE on simulated epileptiform activity and recorded auditory steady state response (ASSR) data, showing that the CMNE estimates exhibit a higher degree of spatial fidelity than the unfiltered estimates in the tested cases.

MNE Scan: Software for real-time processing of electrophysiological data.

BACKGROUND: Magnetoencephalography (MEG) and Electroencephalography (EEG) are noninvasive techniques to study the electrophysiological activity of the human brain. Thus, they are well suited for real-time monitoring and analysis of neuronal activity. Real-time MEG/EEG data processing allows adjustment of the stimuli to the subject's responses for optimizing the acquired information especially by providing dynamically changing displays to enable neurofeedback. NEW METHOD: We introduce MNE Scan, an acquisition and real-time analysis software based on the multipurpose software library MNE-CPP. MNE Scan allows the development and application of acquisition and novel real-time processing methods in both research and clinical studies. The MNE Scan development follows a strict software engineering process to enable approvals required for clinical software. RESULTS: We tested the performance of MNE Scan in several device-independent use cases, including, a clinical epilepsy study, real-time source estimation, and Brain Computer Interface (BCI) application. COMPARISON WITH EXISTING METHOD(S): Compared to existing tools we propose a modular software considering clinical software requirements expected by certification authorities. At the same time the software is extendable and freely accessible. CONCLUSION: We conclude that MNE Scan is the first step in creating a device-independent open-source software to facilitate the transition from basic neuroscience research to both applied sciences and clinical applications.

Real-Time Clustered Multiple Signal Classification (RTC-MUSIC).

Magnetoencephalography (MEG) and electroencephalography provide a high temporal resolution, which allows estimation of the detailed time courses of neuronal activity. However, in real-time analysis of these data two major challenges must be handled: the low signal-to-noise ratio (SNR) and the limited time available for computations. In this work, we present real-time clustered multiple signal classification (RTC-MUSIC) a real-time source localization algorithm, which can handle low SNRs and can reduce the computational effort. It provides correlation information together with sparse source estimation results, which can, e.g., be used to identify evoked responses with high sensitivity. RTC-MUSIC clusters the forward solution based on an anatomical brain atlas and optimizes the scanning process inherent to MUSIC approaches. We evaluated RTC-MUSIC by analyzing MEG auditory and somatosensory data. The results demonstrate that the proposed method localizes sources reliably. For the auditory experiment the most dominant correlated source pair was located bilaterally in the superior temporal gyri. The highest activation in the somatosensory experiment was found in the contra-lateral primary somatosensory cortex.

Versatile synchronized real-time MEG hardware controller for large-scale fast data acquisition.

Versatile controllers for accurate, fast, and real-time synchronized acquisition of large-scale data are useful in many areas of science, engineering, and technology. Here, we describe the development of a controller software based on a technique called queued state machine for controlling the data acquisition (DAQ) hardware, continuously acquiring a large amount of data synchronized across a large number of channels (>400) at a fast rate (up to 20 kHz/channel) in real time, and interfacing with applications for real-time data analysis and display of electrophysiological data. This DAQ controller was developed specifically for a 384-channel pediatric whole-head magnetoencephalography (MEG) system, but its architecture is useful for wide applications. This controller running in a LabVIEW environment interfaces with microprocessors in the MEG sensor electronics to control their real-time operation. It also interfaces with a real-time MEG analysis software via transmission control protocol/internet protocol, to control the synchronous acquisition and transfer of the data in real time from >400 channels to acquisition and analysis workstations. The successful implementation of this controller for an MEG system with a large number of channels demonstrates the feasibility of employing the present architecture in several other applications.

BabyMEG: A whole-head pediatric magnetoencephalography system for human brain development research.

We developed a 375-channel, whole-head magnetoencephalography (MEG) system ("BabyMEG") for studying the electrophysiological development of human brain during the first years of life. The helmet accommodates heads up to 95% of 36-month old boys in the USA. The unique two-layer sensor array consists of: (1) 270 magnetometers (10 mm diameter, ∼15 mm coil-to-coil spacing) in the inner layer, (2) thirty-five three-axis magnetometers (20 mm × 20 mm) in the outer layer 4 cm away from the inner layer. Additionally, there are three three-axis reference magnetometers. With the help of a remotely operated position adjustment mechanism, the sensor array can be positioned to provide a uniform short spacing (mean 8.5 mm) between the sensor array and room temperature surface of the dewar. The sensors are connected to superconducting quantum interference devices (SQUIDs) operating at 4.2 K with median sensitivity levels of 7.5 fT/√Hz for the inner and 4 fT/√Hz for the outer layer sensors. SQUID outputs are digitized by a 24-bit acquisition system. A closed-cycle helium recycler provides maintenance-free continuous operation, eliminating the need for helium, with no interruption needed during MEG measurements. BabyMEG with the recycler has been fully operational from March, 2015. Ongoing spontaneous brain activity can be monitored in real time without interference from external magnetic noise sources including the recycler, using a combination of a lightly shielded two-layer magnetically shielded room, an external active shielding, a signal-space projection method, and a synthetic gradiometer approach. Evoked responses in the cortex can be clearly detected without averaging. These new design features and capabilities represent several advances in MEG, increasing the utility of this technique in basic neuroscience as well as in clinical research and patient studies.

Real-Time MEG Source Localization Using Regional Clustering.

With its millisecond temporal resolution, Magnetoencephalography (MEG) is well suited for real-time monitoring of brain activity. Real-time feedback allows the adaption of the experiment to the subject's reaction and increases time efficiency by shortening acquisition and off-line analysis. Two formidable challenges exist in real-time analysis: the low signal-to-noise ratio (SNR) and the limited time available for computations. Since the low SNR reduces the number of distinguishable sources, we propose an approach which downsizes the source space based on a cortical atlas and allows to discern the sources in the presence of noise. Each cortical region is represented by a small set of dipoles, which is obtained by a clustering algorithm. Using this approach, we adapted dynamic statistical parametric mapping for real-time source localization. In terms of point spread and crosstalk between regions the proposed clustering technique performs better than selecting spatially evenly distributed dipoles. We conducted real-time source localization on MEG data from an auditory experiment. The results demonstrate that the proposed real-time method localizes sources reliably in the superior temporal gyrus. We conclude that real-time source estimation based on MEG is a feasible, useful addition to the standard on-line processing methods, and enables feedback based on neural activity during the measurements.