Detection of event-related potentials in individual subjects using support vector machines

Neural activity related to productive vocabulary knowledge effects during second language comprehension

Second language learners and educators often believe that improving one's listening ability hinges on acquiring an extensive vocabulary and engaging in thorough listening practice. Our previous study suggested that listening comprehension is also impacted by the ability to produce vocabulary. Nevertheless, it remained uncertain whether quick comprehension could be attributed to a simple acceleration of processing or to changes in neural activity. To identify neural activity changes during sentence listening comprehension according to different levels of lexical knowledge (productive, only comprehensive, uncomprehensive), we measured participants' electrical activity in the brain via electroencephalography (EEG) and conducted a time-frequency-based EEG power analysis. Additionally, we employed a decoding model to verify the predictability of vocabulary knowledge levels based on neural activity. The decoding results showed that EEG activity could discriminate between listening to sentences containing phrases that include productive knowledge and ones without. The positive impact of productive vocabulary knowledge on sentence comprehension, driven by distinctive neural processing during sentence comprehension, was unequivocally evident. Our study emphasizes the importance of productive vocabulary knowledge acquisition to enhance the process of second language listening comprehension.

Detecting how time is subjectively perceived based on event-related potentials (ERPs): a machine learning approach

Background and objective: Time perception is essential for the precise performance of many of our activities and the coordination between different modalities. But it is distorted in many diseases and disorders. Event-related potentials (ERP) have long been used to understand better how the human brain perceives time, but machine learning methods have rarely been used to detect a person's time perception from his/her ERPs. Methods: In this study, EEG signals of the individuals were recorded while performing an auditory oddball time discrimination task. After features were extracted from ERPs, data balancing, and feature selection, machine learning models were used to distinguish between the oddball durations of 400 ms and 600 ms from standard durations of 500 ms. ERP results showed that the P3 evoked by the 600 ms oddball stimuli appeared about 200 ms later than that of the 400 ms oddball tones. Classification performance results indicated that support vector machine (SVM) outperformed K-nearest neighbors (KNN), Random Forest, and Logistic regression models. Results: The accuracy of SVM was 91.24, 92.96, and 89.9 for the three used labeling modes, respectively. Another important finding was that most features selected for classification were in the P3 component range, supporting the observed significant effect of duration on the P3. Although all N1, P2, N2, and P3 components contributed to detecting the desired durations. Conclusion: Therefore, results of this study suggest the P3 component as a potential candidate to detect sub-second periods in future researches on brain-computer interface (BCI) applications.

Machines, mathematics, and modules: the potential to provide real-time metrics for pain under anesthesia

The brain-based assessments under anesthesia have provided the ability to evaluate pain/nociception during surgery and the potential to prevent long-term evolution of chronic pain. Prior studies have shown that the functional near-infrared spectroscopy (fNIRS)-measured changes in cortical regions such as the primary somatosensory and the polar frontal cortices show consistent response to evoked and ongoing pain in awake, sedated, and anesthetized patients. We take this basic approach and integrate it into a potential framework that could provide real-time measures of pain/nociception during the peri-surgical period. This application could have significant implications for providing analgesia during surgery, a practice that currently lacks quantitative evidence to guide patient tailored pain management. Through a simple readout of "pain" or "no pain," the proposed system could diminish or eliminate levels of intraoperative, early post-operative, and potentially, the transition to chronic post-surgical pain. The system, when validated, could also be applied to measures of analgesic efficacy in the clinic.

External globus pallidus input to the dorsal striatum regulates habitual seeking behavior in male mice

The external globus pallidus (GPe) coordinates action-selection through GABAergic projections throughout the basal ganglia. GPe arkypallidal (arky) neurons project exclusively to the dorsal striatum, which regulates goal-directed and habitual seeking. However, the role of GPe arky neurons in reward-seeking remains unknown. Here, we identified that a majority of arky neurons target the dorsolateral striatum (DLS). Using fiber photometry, we found that arky activities were higher during random interval (RI; habit) compared to random ratio (RR; goal) operant conditioning. Support vector machine analysis demonstrated that arky neuron activities have sufficient information to distinguish between RR and RI behavior. Genetic ablation of this arkyGPe→DLS circuit facilitated a shift from goal-directed to habitual behavior. Conversely, chemogenetic activation globally reduced seeking behaviors, which was blocked by systemic D1R agonism. Our findings reveal a role of this arkyGPe→DLS circuit in constraining habitual seeking in male mice, which is relevant to addictive behaviors and other compulsive disorders.

Trial-Specific Feature Performance on Single-Channel Auditory Mismatch Negativity Detection

Successful detection of uncommon events is vital in the survival of an organism. Specifically, the study of neuro-sensory detection lends itself widely to understanding the human brain. Mismatch Negativity (MMN) is an important Event-Related Potential (ERP) response to an oddball stimulus which is preceded by repeated homogeneous stimulation. MMN is associated with perceptual learning and medical diagnostics among other applications. Currently, MMN detection relies on visual inspection of ERPs by skilled clinicians which makes for a costly, slow and subjective tool. In this paper, we use MMN to quantify the discriminative abilities of healthy or diagnosed subjects. We introduce a novel algorithmic method to extract and select important trial-specific features for discriminating standard from deviant responses. We utilize machine learning and classification approaches to evaluate our novel model using single-subject trial data while minimizing the number of necessary selection features provided by statistical test parameters and Genetic Algorithm (GA). In this work, a large variety of methods with 27 subjects, hundreds of trials and electrode counts compete for the definitive discrimination of MMN events. Our model requires only one EEG channel, a single subject and as low as five deviant tones. The results show statistically significant detection improvement over the traditional methods while maximizing resource economy.

Distant Sensor Prediction of Event-Related Potentials

OBJECTIVE

The ability to measure event-related potentials (ERPs) as practical, portable brain vital signs is limited by the physical locations of electrodes. Standard electrode locations embedded within the hair result in challenges to obtaining quality signals in a rapid manner. Moreover, these sites require electrode gel, which can be inconvenient. As electrical activity in the brain is spatially volume distributed, it should be possible to predict ERPs from distant sensor locations at easily accessible mastoid and forehead scalp regions.

METHODS

An artificial neural network was trained on ERP signals recorded from below hairline electrode locations (Tp9, Tp10, Af7, Af8 referenced to Fp1, Fp2) to predict signals recorded at the ideal Cz location.

RESULTS

The model resulted in mean improvements in intraclass correlation coefficient relative to control for all stimulus types (Standard Tones: +9.74%, Deviant Tones: +3.23%, Congruent Words: +15.25%, Incongruent Words: +25.43%) and decreases in RMS Error (Standard Tones: -26.72%, Deviant Tones: -17.80%, Congruent Words: -28.78%, Incongruent Words: -29.61%) compared to the individual distant channels. Measured vs predicted ERP amplitudes were highly and significantly correlated with control for the N100 (R = 0.5, p_adj < 0.05), P300 (R = 0.75, p_adj < 0.01), and N400 (R = 0.75, p_adj < 0.01) ERPs.

CONCLUSION

ERP waveforms at distant channels can be combined using a neural network autoencoder to model the control channel features with better precision than those at individual distant channels. This is the first demonstration of feasibility of predicting evoked potentials and brain vital signs using signals recorded from more distant, practical locations.

SIGNIFICANCE

This solves a key engineering challenge for applications that require portability, comfort, and speed of measurement as design priorities for measurement of event-related potentials across a range of individuals, settings, and circumstances.

A condition-independent framework for the classification of error-related brain activity

The cognitive processing and detection of errors is important in the adaptation of the behavioral and learning processes. This brain activity is often reflected as distinct patterns of event-related potentials (ERPs) that can be employed in the detection and interpretation of the cerebral responses to erroneous stimuli. However, high-accuracy cross-condition classification is challenging due to the significant variations of the error-related ERP components (ErrPs) between complexity conditions, thus hindering the development of error recognition systems. In this study, we employed support vector machines (SVM) classification methods, based on waveform characteristics of ErrPs from different time windows, to detect correct and incorrect responses in an audio identification task with two conditions of different complexity. Since the performance of the classifiers usually depends on the salience of the features employed, a combination of the sequential forward floating feature selection (SFFS) and sequential forward feature selection (SFS) methods was implemented to detect condition-independent and condition-specific feature subsets. Our framework achieved high accuracy using a small subset of the available features both for cross- and within-condition classification, hence supporting the notion that machine learning techniques can detect hidden patterns of ErrP-based features, irrespective of task complexity while additionally elucidating complexity-related error processing variations. Graphical abstract A schematic of the proposed approach. (a) EEG recordings in an auditory experiment in two conditions of different complexity. (b) Characteristic event related activity feature extraction. (c) Selection of feature vector subsets for easy and hard conditions corresponding to correct (Class1) and incorrect (Class2) responses. (d) Performance for individual and cross-condition classification.

Novel Signal Processing Technique for Capture and Isolation of Blink-Related Oscillations Using a Low-Density Electrode Array for Bedside Evaluation of Consciousness

OBJECTIVE

Blink-related oscillations derived from electroencephalography (EEG) have recently emerged as an important measure of awareness. Combined with portable EEG hardware with low-density electrode arrays, this neural marker may crucially augment the existing bedside assessments of consciousness in unresponsive patients. Nonetheless, the close relationship between signal characteristics of the neural response of interest and blink-induced oculomotor artifacts poses particular challenges when measuring blink-related oscillations using a point-of-care platform. This study presents a novel denoising approach based on time-frequency (TF) filtering that exploits the differential temporal and spectral features to isolate the neural response from ocular artifact in a low-density array.

METHODS

We investigated the effectiveness of the TF filtering technique using 64-channel EEG data collected in healthy adults, with focal analysis of the Pz and POz channels.

RESULTS

TF filtering showed comparable performance in denoising the signal relative to the established gold-standard independent component analysis approach, with strong similarities in morphological characteristics as measured by intraclass correlations (p < 0.001), extent of artifact rejection based on the ocular contamination index (p < 0.006), as well as time- and frequency-domain signal capture (p < 0.05). Results are robust at the individual and group levels, and are crucially validated using raw data from only four electrodes comprising Pz, POz, Fp2, and T7.

CONCLUSION

These results demonstrate for the first time that TF filtering enables the successful capture and isolation of the blink-related oscillations response using a four-electrode array.

SIGNIFICANCE

This significantly advances the translation of the blink-related oscillations marker to a point-of-care platform for eventual bedside applications.

Accessing knowledge of the 'here and now': a new technique for capturing electromagnetic markers of orientation processing

OBJECTIVE

The ability to orient with respect to the current context (e.g. current time or location) is crucial for daily functioning, and is used to measure overall cognitive health across many frontline clinical assessments. However, these tests are often hampered by their reliance on verbal probes (e.g. 'What city are we in?') in evaluating orientation. Objective, physiology-based measures of orientation processing are needed, but no such measures are currently in existence. We report the initial development of potential brainwave-based markers of orientation processing as characterized using electroencephlography (EEG) and magnetoencephalography (MEG).

APPROACH

An auditory stimulus sequence embedded with words corresponding to orientation-relevant (i.e. related to the 'here and now') and orientation-irrelevant (i.e. unrelated to the current context) conditions was used to elicit orientation processing responses. EEG/MEG data, in concert with clinical assessments, were collected from 29 healthy adults. Analysis at sensor and source levels identified and characterized neural signals related to orientation processing.

MAIN RESULTS

Orientation-irrelevant stimuli elicited increased negative amplitude in EEG-derived event-related potential (ERP) waveforms during the 390-570 ms window (p  <  0.05), with cortical activations across the left frontal, temporal, and parietal regions. These effects are consistent with the well-known N400 response to semantic incongruence. In contrast, ERP responses to orientation-relevant stimuli exhibited increased positive amplitude during the same interval (p  <  0.05), with activations across the bilateral temporal and parietal regions. Importantly, these differential responses were robust at the individual level, with machine-learning classification showing high accuracy (89%), sensitivity (0.88) and specificity (0.90).

SIGNIFICANCE

This is the first demonstration of a neurotechnology platform that elicits, captures, and evaluates electrophysiological markers of orientation processing. We demonstrate neural responses to orientation stimuli that are validated across EEG and MEG modalities and robust at the individual level. The extraction of physiology-based markers through this technique may enable improved objective brain functional evaluation in clinical applications.

Review of EEG-based pattern classification frameworks for dyslexia

Dyslexia is a disability that causes difficulties in reading and writing despite average intelligence. This hidden disability often goes undetected since dyslexics are normal and healthy in every other way. Electroencephalography (EEG) is one of the upcoming methods being researched for identifying unique brain activation patterns in dyslexics. The aims of this paper are to examine pros and cons of existing EEG-based pattern classification frameworks for dyslexia and recommend optimisations through the findings to assist future research. A critical analysis of the literature is conducted focusing on each framework’s (1) data collection, (2) pre-processing, (3) analysis and (4) classification methods. A wide range of inputs as well as classification approaches has been experimented for the improvement in EEG-based pattern classification frameworks. It was uncovered that incorporating reading- and writing-related tasks to experiments used in data collection may help improve these frameworks instead of using only simple tasks, and those unwanted artefacts caused by body movements in the EEG signals during reading and writing activities could be minimised using artefact subspace reconstruction. Further, support vector machine is identified as a promising classifier to be used in EEG-based pattern classification frameworks for dyslexia.

Multimodal characterization of the semantic N400 response within a rapid evaluation brain vital sign framework

Background

For nearly four decades, the N400 has been an important brainwave marker of semantic processing. It can be recorded non-invasively from the scalp using electrical and/or magnetic sensors, but largely within the restricted domain of research laboratories specialized to run specific N400 experiments. However, there is increasing evidence of significant clinical utility for the N400 in neurological evaluation, particularly at the individual level. To enable clinical applications, we recently reported a rapid evaluation framework known as “brain vital signs” that successfully incorporated the N400 response as one of the core components for cognitive function evaluation. The current study characterized the rapidly evoked N400 response to demonstrate that it shares consistent features with traditional N400 responses acquired in research laboratory settings—thereby enabling its translation into brain vital signs applications.

Methods

Data were collected from 17 healthy individuals using magnetoencephalography (MEG) and electroencephalography (EEG), with analysis of sensor-level effects as well as evaluation of brain sources. Individual-level N400 responses were classified using machine learning to determine the percentage of participants in whom the response was successfully detected.

Results

The N400 response was observed in both M/EEG modalities showing significant differences to incongruent versus congruent condition in the expected time range (p < 0.05). Also as expected, N400-related brain activity was observed in the temporal and inferior frontal cortical regions, with typical left-hemispheric asymmetry. Classification robustly confirmed the N400 effect at the individual level with high accuracy (89%), sensitivity (0.88) and specificity (0.90).

Conclusion

The brain vital sign N400 characteristics were highly consistent with features of the previously reported N400 responses acquired using traditional laboratory-based experiments. These results provide important evidence supporting clinical translation of the rapidly acquired N400 response as a potential tool for assessments of higher cognitive functions.

Developing Brain Vital Signs: Initial Framework for Monitoring Brain Function Changes Over Time

Clinical assessment of brain function relies heavily on indirect behavior-based tests. Unfortunately, behavior-based assessments are subjective and therefore susceptible to several confounding factors. Event-related brain potentials (ERPs), derived from electroencephalography (EEG), are often used to provide objective, physiological measures of brain function. Historically, ERPs have been characterized extensively within research settings, with limited but growing clinical applications. Over the past 20 years, we have developed clinical ERP applications for the evaluation of functional status following serious injury and/or disease. This work has identified an important gap: the need for a clinically accessible framework to evaluate ERP measures. Crucially, this enables baseline measures before brain dysfunction occurs, and might enable the routine collection of brain function metrics in the future much like blood pressure measures today. Here, we propose such a framework for extracting specific ERPs as potential “brain vital signs.” This framework enabled the translation/transformation of complex ERP data into accessible metrics of brain function for wider clinical utilization. To formalize the framework, three essential ERPs were selected as initial indicators: (1) the auditory N100 (Auditory sensation); (2) the auditory oddball P300 (Basic attention); and (3) the auditory speech processing N400 (Cognitive processing). First step validation was conducted on healthy younger and older adults (age range: 22–82 years). Results confirmed specific ERPs at the individual level (86.81–98.96%), verified predictable age-related differences (P300 latency delays in older adults, p < 0.05), and demonstrated successful linear transformation into the proposed brain vital sign (BVS) framework (basic attention latency sub-component of BVS framework reflects delays in older adults, p < 0.05). The findings represent an initial critical step in developing, extracting, and characterizing ERPs as vital signs, critical for subsequent evaluation of dysfunction in conditions like concussion and/or dementia.