Methods for artifact detection and removal from scalp EEG: A review

Deep Convolutional Neural Networks for Feature-Less Automatic Classification of Independent Components in Multi-Channel Electrophysiological Brain Recordings

OBJECTIVE

Interpretation of the electroencephalographic (EEG) and magnetoencephalographic (MEG) signals requires off-line artifacts removal. Since artifacts share frequencies with brain activity, filtering is insufficient. Blind source separation, mainly through independent component analysis (ICA), is the gold-standard procedure for the identification of artifacts in multi-dimensional recordings. However, a classification of brain and artifactual independent components (ICs) is still required. Since ICs exhibit recognizable patterns, classification is usually performed by experts' visual inspection. This procedure is time consuming and prone to errors. Automatic ICs classification has been explored, often through complex ICs features extraction prior to classification. Relying on deep-learning ability of self-extracting the features of interest, we investigated the capabilities of convolutional neural networks (CNNs) for off-line, automatic artifact identification through ICs without feature selection.

METHODS

A CNN was applied to spectrum and topography of a large dataset of few thousand samples of ICs obtained from multi-channel EEG and MEG recordings acquired during heterogeneous experimental settings and on different subjects. CNN performances, when applied to EEG, MEG, and combined EEG and MEG ICs, were explored and compared with state-of-the-art feature-based automatic classification.

RESULTS

Beyond state-of-the-art automatic classification accuracies were demonstrated through cross validation (92.4% EEG, 95.4% MEG, 95.6% EEG+MEG).

CONCLUSION

High CNN classification performances were achieved through heuristical selection of machinery hyperparameters and through the CNN self-selection of the features of interest.

SIGNIFICANCE

Considering the large data availability of multi-channel EEG and MEG recordings, CNNs may be suited for classification of ICs of multi-channel brain electrophysiological recordings.

Parental neural responsivity to infants' visual attention: How mature brains influence immature brains during social interaction

Almost all attention and learning-in particular, most early learning-take place in social settings. But little is known of how our brains support dynamic social interactions. We recorded dual electroencephalography (EEG) from 12-month-old infants and parents during solo play and joint play. During solo play, fluctuations in infants' theta power significantly forward-predicted their subsequent attentional behaviours. However, this forward-predictiveness was lower during joint play than solo play, suggesting that infants' endogenous neural control over attention is greater during solo play. Overall, however, infants were more attentive to the objects during joint play. To understand why, we examined how adult brain activity related to infant attention. We found that parents' theta power closely tracked and responded to changes in their infants' attention. Further, instances in which parents showed greater neural responsivity were associated with longer sustained attention by infants. Our results offer new insights into how one partner influences another during social interaction.

Within- and Between-Session Prefrontal Cortex Response to Virtual Reality Exposure Therapy for Acrophobia

Exposure Therapy (ET) has demonstrated its efficacy in the treatment of phobias, anxiety and Post-traumatic Stress Disorder (PTSD), however, it suffers a high drop-out rate because of too low or too high patient engagement in treatment. Virtual Reality Exposure Therapy (VRET) is comparably effective regarding symptom reduction and offers an alternative tool to facilitate engagement for avoidant participants. Neuroimaging studies have demonstrated that both ET and VRET normalize brain activity within a fear circuit. However, previous studies have employed brain imaging technology which restricts people's movement and hides their body, surroundings and therapist from view. This is at odds with the way engagement is typically controlled. We used a novel combination of neural imaging and VR technology-Functional Near-Infrared Spectroscopy (fNIRS) and Immersive Projection Technology (IPT), to avoid these limitations. Although there are a few studies that have investigated the effect of VRET on a brain function after the treatment, the present study utilized technologies which promote ecological validity to measure brain changes after VRET treatment. Furthermore, there are no studies that have measured brain activity within VRET session. In this study brain activity within the prefrontal cortex (PFC) was measured during three consecutive exposure sessions. N = 13 acrophobic volunteers were asked to walk on a virtual plank with a 6 m drop below. Changes in oxygenated (HbO) hemoglobin concentrations in the PFC were measured in three blocks using fNIRS. Consistent with previous functional magnetic resonance imaging (fMRI) studies, the analysis showed decreased activity in the DLPFC and MPFC during first exposure. The activity increased toward normal across three sessions. The study demonstrates potential efficacy of a method for measuring within-session neural response to virtual stimuli that could be replicated within clinics and research institutes, with equipment better suited to an ET session and at fraction of the cost, when compared to fMRI. This has application in widening access to, and increasing ecological validity of, immersive neuroimaging across understanding, diagnosis, assessment and treatment of, a range of mental disorders such as phobia, anxiety and PTSD or addictions.

Automatic Removal of Physiological Artifacts in EEG: The Optimized Fingerprint Method for Sports Science Applications

Data contamination due to physiological artifacts such as those generated by eyeblinks, eye movements, and muscle activity continues to be a central concern in the acquisition and analysis of electroencephalographic (EEG) data. This issue is further compounded in EEG sports science applications where the presence of artifacts is notoriously difficult to control because behaviors that generate these interferences are often the behaviors under investigation. Therefore, there is a need to develop effective and efficient methods to identify physiological artifacts in EEG recordings during sports applications so that they can be isolated from cerebral activity related to the activities of interest. We have developed an EEG artifact detection model, the Fingerprint Method, which identifies different spatial, temporal, spectral, and statistical features indicative of physiological artifacts and uses these features to automatically classify artifactual independent components in EEG based on a machine leaning approach. Here, we optimized our method using artifact-rich training data and a procedure to determine which features were best suited to identify eyeblinks, eye movements, and muscle artifacts. We then applied our model to an experimental dataset collected during endurance cycling. Results reveal that unique sets of features are suitable for the detection of distinct types of artifacts and that the Optimized Fingerprint Method was able to correctly identify over 90% of the artifactual components with physiological origin present in the experimental data. These results represent a significant advancement in the search for effective means to address artifact contamination in EEG sports science applications.

Wearable brain imaging with multimodal physiological monitoring

The brain is a central component of cognitive and physical human performance. Measures, including functional brain activation, cerebral perfusion, cerebral oxygenation, evoked electrical responses, and resting hemodynamic and electrical activity are all related to, or can predict, health status or performance decrements. However, measuring brain physiology typically requires large, stationary machines that are not suitable for mobile or self-monitoring. Moreover, when individuals are ambulatory, systemic physiological fluctuations—e.g., in heart rate, blood pressure, skin perfusion, and more—can interfere with noninvasive brain measurements. In efforts to address the physiological monitoring and performance assessment needs for astronauts during spaceflight, we have developed easy-to-use, wearable prototypes, such as NINscan, for near-infrared scanning, which can collect synchronized multimodal physiology data, including hemodynamic deep-tissue imaging (including brain and muscles), electroencephalography, electrocardiography, electromyography, electrooculography, accelerometry, gyroscopy, pressure, respiration, and temperature measurements. Given their self-contained and portable nature, these devices can be deployed in a much broader range of settings—including austere environments—thereby, enabling a wider range of novel medical and research physiology applications. We review these, including high-altitude assessments, self-deployable multimodal e.g., (polysomnographic) recordings in remote or low-resource environments, fluid shifts in variable-gravity, or spaceflight analog environments, intracranial brain motion during high-impact sports, and long-duration monitoring for clinical symptom-capture in various clinical conditions. In addition to further enhancing sensitivity and miniaturization, advanced computational algorithms could help support real-time feedback and alerts regarding performance and health.

EEG-Based Brain-Computer Interfaces

Brain-Computer Interfaces (BCIs) are real-time computer-based systems that translate brain signals into useful commands. To date most applications have been demonstrations of proof-of-principle; widespread use by people who could benefit from this technology requires further development. Improvements in current EEG recording technology are needed. Better sensors would be easier to apply, more confortable for the user, and produce higher quality and more stable signals. Although considerable effort has been devoted to evaluating classifiers using public datasets, more attention to real-time signal processing issues and to optimizing the mutually adaptive interaction between the brain and the BCI are essential for improving BCI performance. Further development of applications is also needed, particularly applications of BCI technology to rehabilitation. The design of rehabilitation applications hinges on the nature of BCI control and how it might be used to induce and guide beneficial plasticity in the brain.

Noise Robustness Analysis of Performance for EEG-Based Driver Fatigue Detection Using Different Entropy Feature Sets

Driver fatigue is an important factor in traffic accidents, and the development of a detection system for driver fatigue is of great significance. To estimate and prevent driver fatigue, various classifiers based on electroencephalogram (EEG) signals have been developed; however, as EEG signals have inherent non-stationary characteristics, their detection performance is often deteriorated by background noise. To investigate the effects of noise on detection performance, simulated Gaussian noise, spike noise, and electromyogram (EMG) noise were added into a raw EEG signal. Four types of entropies, including sample entropy (SE), fuzzy entropy (FE), approximate entropy (AE), and spectral entropy (PE), were deployed for feature sets. Three base classifiers (K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Decision Tree (DT)) and two ensemble methods (Bootstrap Aggregating (Bagging) and Boosting) were employed and compared. Results showed that: (1) the simulated Gaussian noise and EMG noise had an impact on accuracy, while simulated spike noise did not, which is of great significance for the future application of driver fatigue detection; (2) the influence on noise performance was different based on each classifier, for example, the robust effect of classifier DT was the best and classifier SVM was the weakest; (3) the influence on noise performance was also different with each feature set where the robustness of feature set FE and the combined feature set were the best; and (4) while the Bagging method could not significantly improve performance against noise addition, the Boosting method may significantly improve performance against superimposed Gaussian and EMG noise. The entropy feature extraction method could not only identify driver fatigue, but also effectively resist noise, which is of great significance in future applications of an EEG-based driver fatigue detection system.

Bidirectional Frontoparietal Oscillatory Systems Support Working Memory

The ability to represent and select information in working memory provides the neurobiological infrastructure for human cognition. For 80 years, dominant views of working memory have focused on the key role of prefrontal cortex (PFC) [1-8]. However, more recent work has implicated posterior cortical regions [9-12], suggesting that PFC engagement during working memory is dependent on the degree of executive demand. We provide evidence from neurological patients with discrete PFC damage that challenges the dominant models attributing working memory to PFC-dependent systems. We show that neural oscillations, which provide a mechanism for PFC to communicate with posterior cortical regions [13], independently subserve communications both to and from PFC-uncovering parallel oscillatory mechanisms for working memory. Fourteen PFC patients and 20 healthy, age-matched controls performed a working memory task where they encoded, maintained, and actively processed information about pairs of common shapes. In controls, the electroencephalogram (EEG) exhibited oscillatory activity in the low-theta range over PFC and directional connectivity from PFC to parieto-occipital regions commensurate with executive processing demands. Concurrent alpha-beta oscillations were observed over parieto-occipital regions, with directional connectivity from parieto-occipital regions to PFC, regardless of processing demands. Accuracy, PFC low-theta activity, and PFC → parieto-occipital connectivity were attenuated in patients, revealing a PFC-independent, alpha-beta system. The PFC patients still demonstrated task proficiency, which indicates that the posterior alpha-beta system provides sufficient resources for working memory. Taken together, our findings reveal neurologically dissociable PFC and parieto-occipital systems and suggest that parallel, bidirectional oscillatory systems form the basis of working memory.

Multiscale entropy analysis of biological signals

Traditional approaches to measuring the complexity of biological signals fail to account for the multiple time scales inherent in such time series. These algorithms have yielded contradictory findings when applied to real-world datasets obtained in health and disease states. We describe in detail the basis and implementation of the multiscale entropy (MSE) method. We extend and elaborate previous findings showing its applicability to the fluctuations of the human heartbeat under physiologic and pathologic conditions. The method consistently indicates a loss of complexity with aging, with an erratic cardiac arrhythmia (atrial fibrillation), and with a life-threatening syndrome (congestive heart failure). Further, these different conditions have distinct MSE curve profiles, suggesting diagnostic uses. The results support a general "complexity-loss" theory of aging and disease. We also apply the method to the analysis of coding and noncoding DNA sequences and find that the latter have higher multiscale entropy, consistent with the emerging view that so-called "junk DNA" sequences contain important biological information.