SSVEP detection assessment by combining visual stimuli paradigms and no-training detection methods.

Introduction: Brain-Computer Interfaces (BCI) based on Steady-State Visually Evoked Potentials (SSVEP) have great potential for use in communication applications because of their relatively simple assembly and in some cases the possibility of bypassing the time-consuming training stage. However, among multiple factors, the efficient performance of this technology is highly dependent on the stimulation paradigm applied in combination with the SSVEP detection algorithm employed. This paper proposes the performance assessment of the classification of target events with respect to non-target events by applying four types of visual paradigms, rectangular modulated On-Off (OOR), sinusoidal modulated On-Off (OOS), rectangular modulated Checkerboard (CBR), and sinusoidal modulated Checkerboard (CBS), with three types of SSVEP detection methods, Canonical Correlation Analysis (CCA), Filter-Bank CCA (FBCCA), and Minimum Energy Combination (MEC). Methods: We set up an experimental protocol in which the four types of visual stimuli were presented randomly to twenty-seven participants and after acquiring their electroencephalographic responses to five stimulation frequencies (8.57, 10.909, 15, 20, and 24 Hz), the three detection methods were applied to the collected data. Results: The results are conclusive, obtaining the best performance with the combination of either OOR or OOS visual stimulus and the FBCCA as a detection method, however, this finding contrasts with the opinion of almost half of the participants in terms of visual comfort, where the 51.9% of the subjects felt more comfortable and focused with CBR or CBS stimulation. Discussion: Finally, the EEG recordings correspond to the SSVEP response of 27 subjects to four visual paradigms when selecting five items on a screen, which is useful in BCI navigation applications. The dataset is available to anyone interested in studying and evaluating signal processing and machine-learning algorithms for SSVEP-BCI systems.

Decoding EEG rhythms offline and online during motor imagery for standing and sitting based on a brain-computer interface.

Motor imagery (MI)-based brain-computer interface (BCI) systems have shown promising advances for lower limb motor rehabilitation. The purpose of this study was to develop an MI-based BCI for the actions of standing and sitting. Thirty-two healthy subjects participated in the study using 17 active EEG electrodes. We used a combination of the filter bank common spatial pattern (FBCSP) method and the regularized linear discriminant analysis (RLDA) technique for decoding EEG rhythms offline and online during motor imagery for standing and sitting. The offline analysis indicated the classification of motor imagery and idle state provided a mean accuracy of 88.51 ± 1.43% and 85.29 ± 1.83% for the sit-to-stand and stand-to-sit transitions, respectively. The mean accuracies of the sit-to-stand and stand-to-sit online experiments were 94.69 ± 1.29% and 96.56 ± 0.83%, respectively. From these results, we believe that the MI-based BCI may be useful to future brain-controlled standing systems.

Comparative analysis of spectral and temporal combinations in CSP-based methods for decoding hand motor imagery tasks.

BACKGROUND: A widely used paradigm for brain-computer interfaces (BCI) is based on the detection of event-related (des)synchronization (ERD/S) in response to hand motor imagery (MI) tasks. The common spatial pattern (CSP) has been recognized as a powerful algorithm to design spatial filters for ERD/ERS detection. However, a limitation of CSP focus on identification only of discriminative spatial information but not the spectral one. NEW METHOD: An open problem remains in literature related to extracting the most discriminative brain patterns in MI-based BCIs using an optimal time segment and spectral information that accounts for intersubject variability. In recent years, different variants of CSP-based methods have been proposed to address the problem of decoding motor imagery tasks under the intersubject variability of frequency bands related to ERD/ERS events, including Filter Bank Common Spatial Patterns (FBCSP) and Filter Bank Common Spatio-Spectral Patterns (FBCSSP). COMPARISON WITH EXISTING METHODS: We performed a comparative study of different combinations of time segments and filter banks for three methods (CSP, FBCSP, and FBCSSP) to decode hand (right and left) motor imagery tasks using two different EEG datasets (Gigascience and BCI IVa competition). RESULTS: The best configuration corresponds to a filter bank with 3 filters (8-15 Hz, 15-22 Hz and 22-29 Hz) using a time window of 1.5 s after the trigger, which provide accuracies of approximately 74% and an estimated ITRs of approximately 7 bits/min. CONCLUSION: Discriminative information in time and spectral domains could be obtained using a convenient filter bank and a time segment configuration, to enhance the classification rate and ITR for detection of hand motor imagery tasks with CSP-related methods, to be used in the implementation of a real-time BCI system.

Functional Connectivity and Frequency Power Alterations during P300 Task as a Result of Amyotrophic Lateral Sclerosis.

Amyotrophic Lateral Sclerosis (ALS) is one of the most aggressive neurodegenerative diseases and is now recognized as a multisystem network disorder with impaired connectivity. Further research for the understanding of the nature of its cognitive affections is necessary to monitor and detect the disease, so this work provides insight into the neural alterations occurring in ALS patients during a cognitive task (P300 oddball paradigm) by measuring connectivity and the power and latency of the frequency-specific EEG activity of 12 ALS patients and 16 healthy subjects recorded during the use of a P300-based BCI to command a robotic arm. For ALS patients, in comparison to Controls, the results (p < 0.05) were: an increment in latency of the peak ERP in the Delta range (OZ) and Alpha range (PO7), and a decreased power in the Beta band among most electrodes; connectivity alterations among all bands, especially in the Alpha band between PO7 and the channels above the motor cortex. The evolution observed over months of an advanced-state patient backs up these findings. These results were used to compute connectivity- and power-based features to discriminate between ALS and Control groups using Support Vector Machine (SVM). Cross-validation achieved a 100% in specificity and 75% in sensitivity, with an overall 89% success.

Deep Learning for Gravitational-Wave Data Analysis: A Resampling White-Box Approach.

In this work, we apply Convolutional Neural Networks (CNNs) to detect gravitational wave (GW) signals of compact binary coalescences, using single-interferometer data from real LIGO detectors. Here, we adopted a resampling white-box approach to advance towards a statistical understanding of uncertainties intrinsic to CNNs in GW data analysis. We used Morlet wavelets to convert strain time series to time-frequency images. Moreover, we only worked with data of non-Gaussian noise and hardware injections, removing freedom to set signal-to-noise ratio (SNR) values in GW templates by hand, in order to reproduce more realistic experimental conditions. After hyperparameter adjustments, we found that resampling through repeated k-fold cross-validation smooths the stochasticity of mini-batch stochastic gradient descent present in accuracy perturbations by a factor of 3.6. CNNs are quite precise to detect noise, 0.952 for H1 data and 0.932 for L1 data; but, not sensitive enough to recall GW signals, 0.858 for H1 data and 0.768 for L1 data-although recall values are dependent on expected SNR. Our predictions are transparently understood by exploring tthe distribution of probabilistic scores outputted by the softmax layer, and they are strengthened by a receiving operating characteristic analysis and a paired-sample t-test to compare with a random classifier.

Evaluation of a P300-Based Brain-Machine Interface for a Robotic Hand-Orthosis Control.

This work presents the design, implementation, and evaluation of a P300-based brain-machine interface (BMI) developed to control a robotic hand-orthosis. The purpose of this system is to assist patients with amyotrophic lateral sclerosis (ALS) who cannot open and close their hands by themselves. The user of this interface can select one of six targets, which represent the flexion-extension of one finger independently or the movement of the five fingers simultaneously. We tested offline and online our BMI on eighteen healthy subjects (HS) and eight ALS patients. In the offline test, we used the calibration data of each participant recorded in the experimental sessions to estimate the accuracy of the BMI to classify correctly single epochs as target or non-target trials. On average, the system accuracy was 78.7% for target epochs and 85.7% for non-target trials. Additionally, we observed significant P300 responses in the calibration recordings of all the participants, including the ALS patients. For the BMI online test, each subject performed from 6 to 36 attempts of target selections using the interface. In this case, around 46% of the participants obtained 100% of accuracy, and the average online accuracy was 89.83%. The maximum information transfer rate (ITR) observed in the experiments was 52.83 bit/min, whereas that the average ITR was 18.13 bit/min. The contributions of this work are the following. First, we report the development and evaluation of a mind-controlled robotic hand-orthosis for patients with ALS. To our knowledge, this BMI is one of the first P300-based assistive robotic devices with multiple targets evaluated on people with ALS. Second, we provide a database with calibration data and online EEG recordings obtained in the evaluation of our BMI. This data is useful to develop and compare other BMI systems and test the processing pipelines of similar applications.

Single-Option P300-BCI Performance Is Affected by Visual Stimulation Conditions.

The P300 paradigm is one of the most promising techniques for its robustness and reliability in Brain-Computer Interface (BCI) applications, but it is not exempt from shortcomings. The present work studied single-trial classification effectiveness in distinguishing between target and non-target responses considering two conditions of visual stimulation and the variation of the number of symbols presented to the user in a single-option visual frame. In addition, we also investigated the relationship between the classification results of target and non-target events when training and testing the machine-learning model with datasets containing different stimulation conditions and different number of symbols. To this end, we designed a P300 experimental protocol considering, as conditions of stimulation: the color highlighting or the superimposing of a cartoon face and from four to nine options. These experiments were carried out with 19 healthy subjects in 3 sessions. The results showed that the Event-Related Potentials (ERP) responses and the classification accuracy are stronger with cartoon faces as stimulus type and similar irrespective of the amount of options. In addition, the classification performance is reduced when using datasets with different type of stimulus, but it is similar when using datasets with different the number of symbols. These results have a special connotation for the design of systems, in which it is intended to elicit higher levels of evoked potentials and, at the same time, optimize training time.

Spiking Neural Networks applied to the classification of motor tasks in EEG signals.

Motivated by the recent progress of Spiking Neural Network (SNN) models in pattern recognition, we report on the development and evaluation of brain signal classifiers based on SNNs. The work shows the capabilities of this type of Spiking Neurons in the recognition of motor imagery tasks from EEG signals and compares their performance with other traditional classifiers commonly used in this application. This work includes two stages: the first stage consists of comparing the performance of the SNN models against some traditional neural network models. The second stage, compares the SNN models performance in two input conditions: input features with constant values and input features with temporal information. The EEG signals employed in this work represent five motor imagery tasks: i.e. rest, left hand, right hand, foot and tongue movements. These EEG signals were obtained from a public database provided by the Technological University of Graz (Brunner et al., 2008). The feature extraction stage was performed by applying two algorithms: power spectral density and wavelet decomposition. Likewise, this work uses raw EEG signals for the second stage of the problem solution. All of the models were evaluated in the classification between two motor imagery tasks. This work demonstrates that with a smaller number of Spiking neurons, simple problems can be solved. Better results are obtained by using patterns with temporal information, thereby exploiting the capabilities of the SNNs.

EEG-Based Detection of Braking Intention Under Different Car Driving Conditions.

The anticipatory recognition of braking is essential to prevent traffic accidents. For instance, driving assistance systems can be useful to properly respond to emergency braking situations. Moreover, the response time to emergency braking situations can be affected and even increased by different driver's cognitive states caused by stress, fatigue, and extra workload. This work investigates the detection of emergency braking from driver's electroencephalographic (EEG) signals that precede the brake pedal actuation. Bioelectrical signals were recorded while participants were driving in a car simulator while avoiding potential collisions by performing emergency braking. In addition, participants were subjected to stress, workload, and fatigue. EEG signals were classified using support vector machines (SVM) and convolutional neural networks (CNN) in order to discriminate between braking intention and normal driving. Results showed significant recognition of emergency braking intention which was on average 71.1% for SVM and 71.8% CNN. In addition, the classification accuracy for the best participant was 80.1 and 88.1% for SVM and CNN, respectively. These results show the feasibility of incorporating recognizable driver's bioelectrical responses into advanced driver-assistance systems to carry out early detection of emergency braking situations which could be useful to reduce car accidents.

Decoding Upper Limb Movement Attempt From EEG Measurements of the Contralesional Motor Cortex in Chronic Stroke Patients.

GOAL: Stroke survivors usually require motor rehabilitation therapy as, due to the lesion, they completely or partially loss mobility in the limbs. Brain-computer interface technology offers the possibility of decoding the attempt to move paretic limbs in real time to improve existing motor rehabilitation. However, a major difficulty for the practical application of the BCI to stroke survivors is that the brain rhythms that encode the motor states might be diminished due to the lesion. This study investigates the continuous decoding of natural attempt to move the paralyzed upper limb in stroke survivors from electroencephalographic signals of the unaffected contralesional motor cortex. RESULTS: Experiments were carried out with the aid of six severely affected chronic stroke patients performing/attempting self-selected reaching movements of the unaffected/affected upper limb. The electroencephalographic (EEG) analysis showed significant cortical activation on the uninjured motor cortex when moving the contralateral unaffected arm and in the attempt to move the ipsilateral affected arm. Using this activity, significant continuous decoding of movement was obtained in six out of six participants in movements of the unaffected limb, and in four out of six participants in the attempt to move the affected limb. CONCLUSION: This study showed that it is possible to construct a decoder of the attempt to move the paretic arm for chronic stroke patients using the EEG activity of the healthy contralesional motor cortex. SIGNIFICANCE: This decoding model could provide to stroke survivors with a natural, easy, and intuitive way to achieve control of BCIs or robot-assisted rehabilitation devices.

Detecting the Intention to Move Upper Limbs from Electroencephalographic Brain Signals.

Early decoding of motor states directly from the brain activity is essential to develop brain-machine interfaces (BMI) for natural motor control of neuroprosthetic devices. Hence, this study aimed to investigate the detection of movement information before the actual movement occurs. This information piece could be useful to provide early control signals to drive BMI-based rehabilitation and motor assisted devices, thus providing a natural and active rehabilitation therapy. In this work, electroencephalographic (EEG) brain signals from six healthy right-handed participants were recorded during self-initiated reaching movements of the upper limbs. The analysis of these EEG traces showed that significant event-related desynchronization is present before and during the execution of the movements, predominantly in the motor-related α and ß frequency bands and in electrodes placed above the motor cortex. This oscillatory brain activity was used to continuously detect the intention to move the limbs, that is, to identify the motor phase prior to the actual execution of the reaching movement. The results showed, first, significant classification between relax and movement intention and, second, significant detection of movement intention prior to the onset of the executed movement. On the basis of these results, detection of movement intention could be used in BMI settings to reduce the gap between mental motor processes and the actual movement performed by an assisted or rehabilitation robotic device.

On the usage of linear regression models to reconstruct limb kinematics from low frequency EEG signals.

Several works have reported on the reconstruction of 2D/3D limb kinematics from low-frequency EEG signals using linear regression models based on positive correlation values between the recorded and the reconstructed trajectories. This paper describes the mathematical properties of the linear model and the correlation evaluation metric that may lead to a misinterpretation of the results of this type of decoders. Firstly, the use of a linear regression model to adjust the two temporal signals (EEG and velocity profiles) implies that the relevant component of the signal used for decoding (EEG) has to be in the same frequency range as the signal to be decoded (velocity profiles). Secondly, the use of a correlation to evaluate the fitting of two trajectories could lead to overly-optimistic results as this metric is invariant to scale. Also, the correlation has a non-linear nature that leads to higher values for sinus/cosinus-like signals at low frequencies. Analysis of these properties on the reconstruction results was carried out through an experiment performed in line with previous studies, where healthy participants executed predefined reaching movements of the hand in 3D space. While the correlations of limb velocity profiles reconstructed from low-frequency EEG were comparable to studies in this domain, a systematic statistical analysis revealed that these results were not above the chance level. The empirical chance level was estimated using random assignments of recorded velocity profiles and EEG signals, as well as combinations of randomly generated synthetic EEG with recorded velocity profiles and recorded EEG with randomly generated synthetic velocity profiles. The analysis shows that the positive correlation results in this experiment cannot be used as an indicator of successful trajectory reconstruction based on a neural correlate. Several directions are herein discussed to address the misinterpretation of results as well as the implications on previous invasive and non-invasive works.

DYNAMO: concurrent dynamic multi-model source localization method for EEG and/or MEG.

This work presents a new dipolar method to estimate the neural sources from separate or combined EEG and MEG data. The novelty lies in the simultaneous estimation and integration of neural sources from different dynamic models with different parameters, leading to a dynamic multi-model solution for the EEG/MEG source localization problem. The first key aspect of this method is defining the source model as a dipolar dynamic system, which allows for the estimation of the probability distribution of the sources within the Bayesian filter estimation framework. A second important aspect is the consideration of several banks of filters that simultaneously estimate and integrate the neural sources of different models. A third relevant aspect is that the final probability estimate is a result of the probabilistic integration of the neural sources of numerous models. Such characteristics lead to a new approach that does not require a prior definition neither of the number of sources or of the underlying temporal dynamics, allowing for the specification of multiple initial prior estimates. The method was validated by three sensor modalities with simulated data designed to impose difficult estimation situations, and with real EEG data recorded in a feedback error-related potential paradigm. On the basis of these evaluations, the method was able to localize the sources with high accuracy.

Continuous decoding of motor attempt and motor imagery from EEG activity in spinal cord injury patients.

Spinal cord injury (SCI) associates brain reorganization with a loss of cortical representation of paralyzed limbs. This effect is more pronounced in the chronic state, which can be reached approximately 6 months after the lesion. As many of the brain-computer interfaces (BCI) developed to date rely on the user motor activity, loss of this activity hinders the application of BCI technology for rehabilitation or motor compensation in these patients. This work is a preliminary study with three quadriplegic patients close to reaching the chronic state, addressing two questions: (i) whether it is still possible to use BCI technology to detect motor intention of the paralyzed hand at this state of chronicity; and (ii) whether it is better for the BCI decoding to rely on the motor attempt or the motor imagery of the hand as mental paradigm. The results show that one of the three patients had already lost the motor programs related to the hand, so it was not possible to build a motor-related BCI for him. For the other patients it was suitable to design a BCI based on both paradigms, but the results were better using motor attempt as it has broader activation associated patterns that are easier to recognize.

Continuous decoding of intention to move from contralesional hemisphere brain oscillations in severely affected chronic stroke patients.

Decoding motor information directly from brain activity is essential in robot-assisted rehabilitation systems to promote motor relearning. However, patients who suffered a stroke in the motor cortex have lost brain activity in the injured area, and consequently, mobility in contralateral limbs. Such a loss eliminates the possibility of extracting motor information from brain activity while the patient is undergoing therapy for the affected limb. This work proposes to decode motor information from EEG activity of the contralesional hemisphere in patients who suffered a hemiparetic stroke. Four stroke patients participated in this study and the results proved the feasibility of decoding motor information while patients attempted to move the affected limb.

Detection of movements with attention or distraction to the motor task during robot-assisted passive movements of the upper limb.

Robot-assisted rehabilitation therapies usually focus on physical aspects rather than on cognitive factors. However, cognitive aspects such as attention, motivation, and engagement play a critical role in motor learning and thus influence the long-term success of rehabilitation programs. This paper studies motor-related EEG activity during the execution of robot-assisted passive movements of the upper limb, while participants either: i) focused attention exclusively on the task; or ii) simultaneously performed another task. Six healthy subjects participated in the study and results showed lower desynchronization during passive movements with another task simultaneously being carried out (compared to passive movements with exclusive attention on the task). In addition, it was proved the feasibility to distinguish between the two conditions.

DYNAMO: dynamic multi-model source localization method for EEG and/or MEG.

This paper proposes a multiple model method that addresses the estimation of the EEG/MEG neural sources as a multihypothesis, multidimensional and dynamic estimation problem. The key aspect is the probabilistic integration of several neural models to simultaneously estimate and integrate the brain activity of different dynamic neural processes that are characterized by the number of sources, the dynamic of those sources and the initial conditions. The method was validated with EEG data gathered in a protocol to elicit error-related potentials, since there is evidence of the brain region that generate those signals. The results reveal that the proposed multiple model method is able to identify the brain structure associated with error processing, which is a preliminary indicator of the validity of the proposed method.

Syllable-based speech recognition using EMG.

This paper presents a silent-speech interface based on electromyographic (EMG) signals recorded in the facial muscles. The distinctive feature of this system is that it is based on the recognition of syllables instead of phonemes or words, which is a compromise between both approaches with advantages as (a) clear delimitation and identification inside a word, and (b) reduced set of classification groups. This system transforms the EMG signals into robust-in-time feature vectors and uses them to train a boosting classifier. Experimental results demonstrated the effectiveness of our approach in three subjects, providing a mean classification rate of almost 70% (among 30 syllables).

Dynamic solution to the EEG source localization problem using Kalman filters and particle filters.

In this paper, we propose a solution to the EEG source localization problem considering its dynamic behavior. We assume a dipolar approach which makes the problem nonlinear. From the dynamic probabilistic model of the problem, we formulate the Extended Kalman Filter and Particle Filter solutions. In order to test the algorithms, we designed an experimental protocol based on error-related potentials. During the experiments, our dynamic solutions have allowed the estimation of sources which are varying in position and moment within the brain volume. Results confirm the activation of the anterior cingulate cortex which is the brain structure associated with error processing. These findings demonstrate the good performance of the dynamic solutions for estimating and tracking EEG neural generators.