Electrocorticographic Signals Research Articles

High-frequency band temporal dynamics in response to a grasp force task

Objective. Brain–computer interfaces (BCIs) are being developed to restore reach and grasping movements of paralyzed individuals. Recent studies have shown that the kinetics of grasping movement, such as grasp force, can be successfully decoded from electrocorticography (ECoG) signals, and that the high-frequency band (HFB) power changes provide discriminative information that contribute to an accurate decoding of grasp force profiles. However, as the models used in these studies contained simultaneous information from multiple spectral features over multiple areas in the brain, it remains unclear what parameters of movement and force are encoded by the HFB signals and how these are represented temporally and spatially in the SMC. Approach. To investigate this, and to gain insight in the temporal dynamics of the HFB during grasping, we continuously modelled the ECoG HFB response recorded from nine individuals with epilepsy temporarily implanted with ECoG grids, who performed three different grasp force tasks. Main results. We show that a model based on the force onset and offset consistently provides a better fit to the HFB power responses when compared with a model based on the force magnitude, irrespective of electrode location. Significance. Our results suggest that HFB power, although potentially useful for continuous decoding, is more closely related to the changes in movement. This finding may potentially contribute to the more natural decoding of grasping movement in neural prosthetics.

Dynamic Brain Interactions during Picture Naming.

Brain computations involve multiple processes by which sensory information is encoded and transformed to drive behavior. These computations are thought to be mediated by dynamic interactions between populations of neurons. Here, we demonstrate that human brains exhibit a reliable sequence of neural interactions during speech production. We use an autoregressive Hidden Markov Model (ARHMM) to identify dynamical network states exhibited by electrocorticographic signals recorded from human neurosurgical patients. Our method resolves dynamic latent network states on a trial-by-trial basis. We characterize individual network states according to the patterns of directional information flow between cortical regions of interest. These network states occur consistently and in a specific, interpretable sequence across trials and subjects: the data support the hypothesis of a fixed-length visual processing state, followed by a variable-length language state, and then by a terminal articulation state. This empirical evidence validates classical psycholinguistic theories that have posited such intermediate states during speaking. It further reveals these state dynamics are not localized to one brain area or one sequence of areas, but are instead a network phenomenon.

Context-dependent relationship in high-resolution micro-ECoG studies during finger movements.

The activation of the sensorimotor cortex as measured by electrocorticographic (ECoG) signals has been correlated with contralateral hand movements in humans, as precisely as the level of individual digits. However, the relationship between individual and multiple synergistic finger movements and the neural signal as detected by ECoG has not been fully explored. The authors used intraoperative high-resolution micro-ECoG (µECoG) on the sensorimotor cortex to link neural signals to finger movements across several context-specific motor tasks. Three neurosurgical patients with cortical lesions over eloquent regions participated. During awake craniotomy, a sensorimotor cortex area of hand movement was localized by high-frequency responses measured by an 8 × 8 µECoG grid of 3-mm interelectrode spacing. Patients performed a flexion movement of the thumb or index finger, or a pinch movement of both, based on a visual cue. High-gamma (HG; 70-230 Hz) filtered µECoG was used to identify dominant electrodes associated with thumb and index movement. Hand movements were recorded by a dataglove simultaneously with µECoG recording. In all 3 patients, the electrodes controlling thumb and index finger movements were identifiable approximately 3-6-mm apart by the HG-filtered µECoG signal. For HG power of cortical activation measured with µECoG, the thumb and index signals in the pinch movement were similar to those observed during thumb-only and index-only movement, respectively (all p > 0.05). Index finger movements, measured by the dataglove joint angles, were similar in both the index-only and pinch movements (p > 0.05). However, despite similar activation across the conditions, markedly decreased thumb movement was observed in pinch relative to independent thumb-only movement (all p < 0.05). HG-filtered µECoG signals effectively identify dominant regions associated with thumb and index finger movement. For pinch, the µECoG signal comprises a combination of the signals from individual thumb and index movements. However, while the relationship between the index finger joint angle and HG-filtered signal remains consistent between conditions, there is not a fixed relationship for thumb movement. Although the HG-filtered µECoG signal is similar in both thumb-only and pinch conditions, the actual thumb movement is markedly smaller in the pinch condition than in the thumb-only condition. This implies a nonlinear relationship between the cortical signal and the motor output for some, but importantly not all, movement types. This analysis provides insight into the tuning of the motor cortex toward specific types of motor behaviors.

Detection of Spikes With Defined Parameters in the ECoG Signal

Patients undergoing presurgical evaluation for resective epilepsy surgery require the localization of the epileptogenic zone. The detection and analysis of interictal and ictal epileptiform spikes are of major importance for identifying this area. The “irritative zone” of the cortex with interictal spikes is usually revealed intraoperatively during acute electrocorticography (ECoG). Since ECoG recordings cannot be completely visually reviewed in a reasonable amount of time, computer algorithms for the automatic detection of seizures and spikes were developed. We proposed the spike detection algorithm based on the definition of the physical properties of the sought spikes (duration and amplitude). The parameters are easily interpreted by neurophysiologists. The detection sensitivity was 93%, and the precision was 99%. In addition, based on the number of spikes detected on individual electrodes, it is possible to draw the “frequency of spike occurrence” map. The algorithm proved to be so effective in analyzing ECoG records, as it is allowed to be used in practice.

Cross-Frequency Coupling Based Neuromodulation for Treating Neurological Disorders.

Synchronous, rhythmic changes in the membrane polarization of neurons form oscillations in local field potentials. It is hypothesized that high-frequency brain oscillations reflect local cortical information processing, and low-frequency brain oscillations project information flow across larger cortical networks. This provides complex forms of information transmission due to interactions between oscillations at different frequency bands, which can be rendered with cross-frequency coupling (CFC) metrics. Phase-amplitude coupling (PAC) is one of the most common representations of the CFC. PAC reflects the coupling of the phase of oscillations in a specific frequency band to the amplitude of oscillations in another frequency band. In a normal brain, PAC accompanies multi-item working memory in the hippocampus, and changes in PAC have been associated with diseases such as schizophrenia, obsessive-compulsive disorder (OCD), Alzheimer disease (AD), epilepsy, and Parkinson’s disease (PD). The purpose of this article is to explore CFC across the central nervous system and demonstrate its correlation to neurological disorders. Results from previously published studies are reviewed to explore the significant role of CFC in large neuronal network communication and its abnormal behavior in neurological disease. Specifically, the association of effective treatment in PD such as dopaminergic medication and deep brain stimulation with PAC changes is described. Lastly, CFC analysis of the electrocorticographic (ECoG) signals recorded from the motor cortex of a Parkinson’s disease patient and the parahippocampal gyrus of an epilepsy patient are demonstrated. This information taken together illuminates possible roles of CFC in the nervous system and its potential as a therapeutic target in disease states. This will require new neural interface technologies such as phase-dependent stimulation triggered by PAC changes, for the accurate recording, monitoring, and modulation of the CFC signal.

Keyword Spotting Using Human Electrocorticographic Recordings.

Neural keyword spotting could form the basis of a speech brain-computer-interface for menu-navigation if it can be done with low latency and high specificity comparable to the “wake-word” functionality of modern voice-activated AI assistant technologies. This study investigated neural keyword spotting using motor representations of speech via invasively-recorded electrocorticographic signals as a proof-of-concept. Neural matched filters were created from monosyllabic consonant-vowel utterances: one keyword utterance, and 11 similar non-keyword utterances. These filters were used in an analog to the acoustic keyword spotting problem, applied for the first time to neural data. The filter templates were cross-correlated with the neural signal, capturing temporal dynamics of neural activation across cortical sites. Neural vocal activity detection (VAD) was used to identify utterance times and a discriminative classifier was used to determine if these utterances were the keyword or non-keyword speech. Model performance appeared to be highly related to electrode placement and spatial density. Vowel height (/a/ vs /i/) was poorly discriminated in recordings from sensorimotor cortex, but was highly discriminable using neural features from superior temporal gyrus during self-monitoring. The best performing neural keyword detection (5 keyword detections with two false-positives across 60 utterances) and neural VAD (100% sensitivity, ~1 false detection per 10 utterances) came from high-density (2 mm electrode diameter and 5 mm pitch) recordings from ventral sensorimotor cortex, suggesting the spatial fidelity and extent of high-density ECoG arrays may be sufficient for the purpose of speech brain-computer-interfaces.

Inferring Cortical Connectivity from ECoG Signals Using Graph Signal Processing.

A novel method to characterize connectivity between sites in the cerebral cortex of primates is proposed in this paper. Connectivity graphs for two macaque monkeys are inferred from Electrocorticographic (ECoG) activity recorded while the animals were alert. The locations of ECoG electrodes are considered as nodes of the graph, the coefficients of the auto-regressive (AR) representation of the signals measured at each node are considered as the signal on the graph and the connectivity strengths between the nodes are considered as the edges of the graph. Maximization of the graph smoothness defined from the Laplacian quadratic form is used to infer the connectivity map (adjacency matrix of the graph). The cortical evoked potential (CEP) map was obtained by stimulating different electrodes and recording the evoked potentials at the other electrodes. The maps obtained by the graph inference and the traditional method of spectral coherence are compared with the CEP map. The results show that the proposed method provides a description of cortical connectivity that is more similar to the stimulation-based measures than spectral coherence. The results are also tested by the surrogate map analysis in which the CEP map is randomly permuted and the distribution of the errors is obtained. It is shown that error between the two maps is comfortably outside the surrogate map error distribution. This indicates that the similarity between the map calculated by the graph inference and the CEP map is statistically significant.

Multimodal Modeling of Neural Network Activity: Computing LFP, ECoG, EEG, and MEG Signals With LFPy 2.0.

Recordings of extracellular electrical, and later also magnetic, brain signals have been the dominant technique for measuring brain activity for decades. The interpretation of such signals is however nontrivial, as the measured signals result from both local and distant neuronal activity. In volume-conductor theory the extracellular potentials can be calculated from a distance-weighted sum of contributions from transmembrane currents of neurons. Given the same transmembrane currents, the contributions to the magnetic field recorded both inside and outside the brain can also be computed. This allows for the development of computational tools implementing forward models grounded in the biophysics underlying electrical and magnetic measurement modalities. LFPy (LFPy.readthedocs.io) incorporated a well-established scheme for predicting extracellular potentials of individual neurons with arbitrary levels of biological detail. It relies on NEURON (neuron.yale.edu) to compute transmembrane currents of multicompartment neurons which is then used in combination with an electrostatic forward model. Its functionality is now extended to allow for modeling of networks of multicompartment neurons with concurrent calculations of extracellular potentials and current dipole moments. The current dipole moments are then, in combination with suitable volume-conductor head models, used to compute non-invasive measures of neuronal activity, like scalp potentials (electroencephalographic recordings; EEG) and magnetic fields outside the head (magnetoencephalographic recordings; MEG). One such built-in head model is the four-sphere head model incorporating the different electric conductivities of brain, cerebrospinal fluid, skull and scalp. We demonstrate the new functionality of the software by constructing a network of biophysically detailed multicompartment neuron models from the Neocortical Microcircuit Collaboration (NMC) Portal (bbp.epfl.ch/nmc-portal) with corresponding statistics of connections and synapses, and compute in vivo-like extracellular potentials (local field potentials, LFP; electrocorticographical signals, ECoG) and corresponding current dipole moments. From the current dipole moments we estimate corresponding EEG and MEG signals using the four-sphere head model. We also show strong scaling performance of LFPy with different numbers of message-passing interface (MPI) processes, and for different network sizes with different density of connections. The open-source software LFPy is equally suitable for execution on laptops and in parallel on high-performance computing (HPC) facilities and is publicly available on GitHub.com.

A Polyimide-Based 3-D Ultrathin Bioelectrode With Elastic Sites for Neural Recording

Flexible sheet electrode can record electrocorticography signals without penetrating the blood-brain barrier. However, the attachment between the cortex and the electrode influences the quality of the signal directly. Here, we present a 3-D ultrathin bioelectrode with elastic electrode sites to achieve conformal coverage on brain tissue and increase the flexibility of the electrode. Inspired by kirigami, the micro-warping elastic structures on the electrode can increase the conformality significantly when compared with the conventional planar electrodes. It has been proved that the warping structure has an excellent springback curve and maintains at least one week under the maximum stress of 7.5 kPa. Furthermore, the electrode sites were modified with platinum black to improve the electrochemical performance, and electrochemical characterization was performed. The electrode was compared with a plane one in vitro , which demonstrated the prominent attachment performance from the surface with sulci. Finally, the local field potential was recorded in vivo . [2018-0088]

Concurrent control of a brain–computer interface and natural overt movements

Objective. A primary control signal in brain–computer interfaces (BCIs) have been cortical signals related to movement. However, in cases where natural motor function remains, BCI control signals may interfere with other possibly simultaneous activity for useful ongoing movement. We sought to determine if the brain could learn to control both a BCI and concurrent overt movement execution in such cases. Approach. We designed experiments where BCI and overt movements must be used concurrently and in coordination to achieve a 2D centre out control. Power in the 70–90 Hz band of human electrocorticography (ECoG) signals, was used to generate BCI control commands for vertical movement of the cursor. These signals were deliberately recorded from the same human cortical site that produced the strongest movement related activity associated with the concurrent overt finger movements required for the horizontal movement of the cursor. Main results. We demonstrate that three subjects were able to perform the concurrent BCI task, controlling BCI and natural movements simultaneously and to a large extent independently. We conclude that the brain is capable of dissociating the original control signal dependency on movement, producing specific BCI control signals in the presence of motor related responses from the ongoing overt behaviour with which the BCI signal was initially correlated. Significance. We demonstrate a novel human brain–computer interface (BCI) which can be used to control movement concurrently and in coordination with movements of the natural limbs. This demonstrates the dissociation of cortical activity from the behaviour with which it was originally associated despite the ongoing behaviour and shows the feasibility of achieving simultaneous BCI control of devices with natural movements.

Unilateral, 3D Arm Movement Kinematics Are Encoded in Ipsilateral Human Cortex.

There is increasing evidence that the hemisphere ipsilateral to a moving limb plays a role in planning and executing movements. However, the exact relationship between cortical activity and ipsilateral limb movements is uncertain. We sought to determine whether 3D arm movement kinematics (speed, velocity, and position) could be decoded from cortical signals recorded from the hemisphere ipsilateral to the moving limb. By having invasively monitored patients perform unilateral reaches with each arm, we also compared the encoding of contralateral and ipsilateral limb kinematics from a single cortical hemisphere. In four motor-intact human patients (three male, one female) implanted with electrocorticography electrodes for localization of their epileptic foci, we decoded 3D movement kinematics of both arms with accuracies above chance. Surprisingly, the spatial and spectral encoding of contralateral and ipsilateral limb kinematics was similar, enabling cross-prediction of kinematics between arms. These results clarify our understanding that the ipsilateral hemisphere robustly contributes to motor execution and supports that the information of complex movements is more bihemispherically represented in humans than has been previously understood.SIGNIFICANCE STATEMENT Although limb movements are traditionally understood to be driven by the cortical hemisphere contralateral to a moving limb, movement-related neural activity has also been found in the ipsilateral hemisphere. This study provides the first demonstration that 3D arm movement kinematics can be decoded from human electrocorticographic signals ipsilateral to the moving limb. Surprisingly, the spatial and spectral encoding of contralateral and ipsilateral limb kinematics was similar. The finding that specific kinematics are encoded in the ipsilateral hemisphere demonstrates that the ipsilateral hemisphere contributes to the execution of unilateral limb movements, improving our understanding of motor control. Additionally, the bihemisheric representation of voluntary movements has implications for the development of neuroprosthetic systems for reaching and for neurorehabilitation strategies following cortical injuries.

Passive functional mapping of receptive language areas using electrocorticographic signals

ObjectiveTo validate the use of passive functional mapping using electrocorticographic (ECoG) broadband gamma signals for identifying receptive language cortex. MethodsWe mapped language function in 23 patients using ECoG and using electrical cortical stimulation (ECS) in a subset of 15 subjects. ResultsThe qualitative comparison between cortical sites identified by ECoG and ECS show a high concordance. A quantitative comparison indicates a high level of sensitivity (95%) and a lower level of specificity (59%). Detailed analysis reveals that 82% of all cortical sites identified by ECoG were within one contact of a site identified by ECS. ConclusionsThese results show that passive functional mapping reliably localizes receptive language areas, and that there is a substantial concordance between the ECoG- and ECS-based methods. They also point to a more refined understanding of the differences between ECoG- and ECS-based mappings. This refined understanding helps to clarify the instances in which the two methods disagree and can explain why neurosurgical practice has established the concept of a “safety margin.” SignificancePassive functional mapping using ECoG signals provides a fast, robust, and reliable method for identifying receptive language areas without many of the risks and limitations associated with ECS.

Rapid Decoding of Hand Gestures in Electrocorticography Using Recurrent Neural Networks.

Brain-computer interface (BCI) is a direct communication pathway between brain and external devices, and BCI-based prosthetic devices are promising to provide new rehabilitation options for people with motor disabilities. Electrocorticography (ECoG) signals contain rich information correlated with motor activities, and have great potential in hand gesture decoding. However, most existing decoders use long time windows, thus ignore the temporal dynamics within the period. In this study, we propose to use recurrent neural networks (RNNs) to exploit the temporal information in ECoG signals for robust hand gesture decoding. With RNN's high nonlinearity modeling ability, our method can effectively capture the temporal information in ECoG time series for robust gesture recognition. In the experiments, we decode three hand gestures using ECoG signals of two participants, and achieve an accuracy of 90%. Specially, we investigate the possibility of recognizing the gestures in a time interval as short as possible after motion onsets. Our method rapidly recognizes gestures within 0.5 s after motion onsets with an accuracy of about 80%. Experimental results also indicate that the temporal dynamics is especially informative for effective and rapid decoding of hand gestures.

Pallidal stimulation in Parkinson disease differentially modulates local and network β activity

β hypersynchrony within the basal ganglia-thalamocortical (BGTC) network has been suggested as a hallmark of Parkinson disease (PD) pathophysiology. Subthalamic nucleus (STN)-DBS has been shown to alter cortical-subcortical synchronization. It is unclear whether this is a generalizable phenomenon of therapeutic stimulation across targets. Objectives. We aimed to evaluate whether DBS of the globus pallidus internus (GPi) results in cortical-subcortical desynchronization, despite the lack of monosynaptic connections between GPi and sensorimotor cortex. Approach. We recorded local field potentials from the GPi and electrocorticographic signals from the ipsilateral sensorimotor cortex, off medications in nine PD patients, undergoing DBS implantation. We analyzed both local oscillatory power and functional connectivity (coherence and debiased weighted phase lag index (dWPLI)) with and without stimulation while subjects were resting with eyes open. Main results. DBS significantly suppressed low β power within the GPi (−26.98% ± 15.14%), p < 0.05) without modulation of sensorimotor cortical β power (low or high). In contrast, stimulation suppressed pallidocortical high β coherence (−38.89% ± 6.19%, p = 0.02) and dWPLI (−61.40% ± 8.75%, p = 0.02). Changes in cortical-subcortical functional connectivity were spatially specific to the motor cortex. Significance. We highlight the role of DBS in desynchronizing network activity, particularly in the high β band. The current study of GPi-DBS suggests these network-level effects are not necessarily dependent and potentially may be independent of the hyperdirect pathway. Importantly, these results draw a sharp distinction between the potential significance of low β oscillations locally within the basal ganglia and high β oscillations across the BGTC motor circuit.

Decoding unconstrained arm movements in primates using high-density electrocorticography signals for brain-machine interface use

Motor deficit is among the most debilitating aspects of injury to the central nervous system. Despite ongoing progress in brain-machine interface (BMI) development and in the functional electrical stimulation of muscles and nerves, little is understood about how neural signals in the brain may be used to potentially control movement in one’s own unconstrained paralyzed limb. We recorded from high-density electrocorticography (ECoG) electrode arrays in the ventral premotor cortex (PMv) of a rhesus macaque and used real-time motion tracking techniques to correlate spatial-temporal changes in neural activity with arm movements made towards objects in three-dimensional space at millisecond precision. We found that neural activity from a small number of electrodes within the PMv can be used to accurately predict reach-return movement onset and directionality. Also, whereas higher gamma frequency field activity was more predictive about movement direction during performance, mid-band (beta and low gamma) activity was more predictive of movement prior to onset. We speculate these dual spatiotemporal signals may be used to optimize both planning and execution of movement during natural reaching, with prospective relevance to the future development of neural prosthetics aimed at restoring motor control over one’s own paralyzed limb.

Delayed high-frequency suppression after automated single-pulse electrical stimulation identifies the seizure onset zone in patients with refractory epilepsy

ObjectiveSingle-pulse electrical stimulation (SPES) of intracranial electrodes evokes responses that may help identify the seizure onset zone (SOZ); however, lack of automation and response variability has limited clinical adoption of this technique. We evaluated whether automated delivery of low-current SPES could evoke delayed high-frequency suppression (DHFS) of ongoing electrocorticography (ECoG) signals that, when combined with objective analytic techniques, may provide a reliable marker of this zone. MethodsLow-current SPES (1-ms, 3.5-mA biphasic pulses) was delivered to 652 electrodes across 10 patients undergoing ECoG for seizure focus localization. DHFS was measured by calculating the normalized trial-averaged time-frequency power (70–250 Hz) 0.4–1 sec post-stimulation. Electrodes that evoked suppression when stimulated or recorded suppression when stimulation was nearby were used to estimate the SOZ. ResultsThe estimated SOZ significantly identified the clinical SOZ in 6 of 10 patients (5 of 7 temporal foci) with a false-positive rate of 0–0.06. Stimulation required <2 h, was undetectable by patients, and did not induce seizures or after-discharges. ConclusionsWe show that DHFS provides accurate estimates of the clinical SOZ in patients with refractory epilepsy. SignificanceThis approach may increase the safety, speed, and reproducibility of SOZ identification while reducing cost, subjectivity, and patient discomfort.

Electrophysiological Activity Evoked by Direct Electrical Stimulation of the Human Brain: Interest of the P0 Component.

Direct electrical stimulation (DES) at 60 Hz is used clinically to perform real-time functional mapping of the brain and guide tumor resection during wide-awake neurosurgery. The electrophysiological effects of DES remain by far unknown, both locally and remotely. In this study, by lowering the DES frequency to 9 Hz and by using differential recording of electro-corticographic signals to improve the focality, we were able to observe that the amplitude of the initial P0 component of the direct cortical response increased when the inter-electrode distance was increased and the pulse width was decreased. This result strongly suggests that larger neural elements, including somas and axons of pyramidal neurons buried in deeper layers of the cortical column, are activated. Their activation produce the observed P0 component, which results from the synchronized summation of action potentials triggered by DES. Interestingly, the early P0 component was not observed during the usual 60 Hz DES. The study of the P0 component and subsequent evoked potentials may help decipher the effects of DES on the stimulated cortical column and identify the activation of underlying white matter fibers. This is crucial to better understand the electrophysiological diffusion of DES, especially at higher frequencies (e.g., 60 Hz).

Complex Modulation Method for Measuring Cross-Frequency Coupling of Neural Oscillations.

There is growing evidence from human intracranial electrocorticography (ECoG) studies that interactions between cortical frequencies are important for sensory perception, cognition and inter-regional neuronal communication. Recent studies have focused mainly on the strength of phase-amplitude coupling in cross-frequency interactions. Here, we introduce a complex modulation method based on measures of coherence to investigate cross-frequency coupling in the neural time series. This novel approach uses complex demodulation transform and coherence measures from the transformed signals. We used this method to quantify power coupling between two cortical frequency bands: theta (47 Hz) and high gamma (70-150 Hz) in ECoG signals recorded during an auditory task. We compared complex modulation results with traditional phase-amplitude coupling measures (PAC) derived from the same ECoG dataset. Our results suggest that cross-frequency coupling may involve changes in both phase-amplitude and power relationships between frequencies, reflecting the complexity of neuronal oscillatory interactions.

GridLoc: An automatic and unsupervised localization method for high-density ECoG grids

Precise localization of electrodes is essential in the field of high-density (HD) electrocorticography (ECoG) brain signal analysis in order to accurately interpret the recorded activity in relation to functional anatomy. Current localization methods for subchronically implanted HD electrode grids involve post-operative imaging. However, for situations where post-operative imaging is not available, such as during acute measurements in awake surgery, electrode localization is complicated. Intra-operative photographs may be informative, but not for electrode grids positioned partially or fully under the skull. Here we present an automatic and unsupervised method to localize HD electrode grids that does not require post-operative imaging. The localization method, named GridLoc, is based on the hypothesis that the anatomical and vascular brain structures under the ECoG electrodes have an effect on the amplitude of the recorded ECoG signal. More specifically, we hypothesize that the spatial match between resting-state high-frequency band power (45–120 Hz) patterns over the grid and the anatomical features of the brain under the electrodes, such as the presence of sulci and larger blood vessels, can be used for adequate HD grid localization. We validate this hypothesis and compare the GridLoc results with electrode locations determined with post-operative imaging and/or photographs in 8 patients implanted with HD-ECoG grids. Locations agreed with an average difference of 1.94 ± 0.11 mm, which is comparable to differences reported earlier between post-operative imaging and photograph methods. The results suggest that resting-state high-frequency band activity can be used for accurate localization of HD grid electrodes on a pre-operative MRI scan and that GridLoc provides a convenient alternative to methods that rely on post-operative imaging or intra-operative photographs.

Fast Multiway Partial Least Squares Regression.

Multiway array decomposition has been successful in providing a better understanding of the structure underlying data and in discovering potentially hidden feature dependences serving high-performance decoder applications. However, the computational cost of multiway algorithms can become prohibitive, especially when considering large datasets, rendering them unsuitable for time-critical applications. We propose a multiway regression model for large-scale tensors with optimized performance in terms of time complexity, called fast higher order partial least squares (fHOPLS). We compare fHOPLS with its native version, higher order partial least squares (HOPLS), the state-of-the-art in multilinear regression, under different noise conditions and tensor dimensionalities using synthetic data. We also compare their performance when used for predicting scalp-recorded electroencephalography signals from invasively recorded electrocorticography signals in an oddball experiment. For the sake of exposition, we evaluated the performance of standard unfolded partial least squares (PLS) and linear regression. Our results show that fHOPLS is significantly faster than HOPLS, in particular for big data. In addition, the regression performances of fHOPLS and HOPLS are comparable and outperform both unfolded PLS and linear regression. Another interesting result is that multiway array decoding yields more accurate results than epoch-based averaging procedures traditionally used in the brain computer interfacing community.