Martin Billinger

FORCe: Fully Online and Automated Artifact Removal for Brain-Computer Interfacing

A fully automated and online artifact removal method for the electroencephalogram (EEG) is developed for use in brain-computer interfacing (BCI). The method (FORCe) is based upon a novel combination of wavelet decomposition, independent component analysis, and thresholding. FORCe is able to operate on a small channel set during online EEG acquisition and does not require additional signals (e.g., electrooculogram signals). Evaluation of FORCe is performed offline on EEG recorded from 13 BCI particpants with cerebral palsy (CP) and online with three healthy participants. The method outperforms the state-of the-art automated artifact removal methods Lagged Auto-Mutual Information Clustering (LAMIC) and Fully Automated Statistical Thresholding for EEG artifact Rejection (FASTER), and is able to remove a wide range of artifact types including blink, electromyogram (EMG), and electrooculogram (EOG) artifacts.

Is it significant? Guidelines for reporting BCI performance

Recent growth in brain-computer interface (BCI) research has increased pressure to report improved performance. However, different research groups report performance in different ways. Hence, it is essential that evaluation procedures are valid and reported in sufficient detail. In this chapter we give an overview of available performance measures such as classification accuracy, cohen’s kappa, information transfer rate (ITR), and written symbol rate. We show how to distinguish results from chance level using confidence intervals for accuracy or kappa. Furthermore, we point out common pitfalls when moving from offline to online analysis and provide a guide on how to conduct statistical tests on (BCI) results.

On the Automated Removal of Artifacts Related to Head Movement From the EEG

Contamination of the electroencephalogram (EEG) by artifacts related to head movement is a major cause of reduced signal quality. This is a problem in both neuroscience and other uses of the EEG. To attempt to reduce the influence, on the EEG, of artifacts related to head movement, an accelerometer is placed on the head and independent component analysis is applied to attempt to separate artifacts which are statistically related to head movements. To evaluate the method, EEG and accelerometer measurements are made from 14 individuals with Cerebral palsy attempting to control a sensorimotor rhythm based brain-computer interface. Results show that the approach significantly reduces the influence of head movement related artifacts in the EEG.

Single-trial connectivity estimation for classification of motor imagery data

OBJECTIVE Many brain-computer interfaces (BCIs) use band power (BP) changes in the electroencephalogram to distinguish between different motor imagery (MI) patterns. Most current approaches do not take connectivity of separated brain areas into account. Our objective is to introduce single-trial connectivity features and apply these features to BCI data. APPROACH We introduce a procedure for extracting single-trial connectivity estimates from vector autoregressive (VAR) models of independent components in a BCI setting. MAIN RESULTS In a simulated BCI, we demonstrate that the directed transfer function (DTF) with full-frequency normalization and the direct DTF give classification results similar to BP, while other measures such as the partial directed coherence perform significantly worse. SIGNIFICANCE We show that single-trial MI classification is possible with connectivity measures extracted from VAR models, and that a BCI could potentially utilize such measures.

A comparison of univariate, vector, bilinear autoregressive, and band power features for brain-computer interfaces.

Selecting suitable feature types is crucial to obtain good overall brain–computer interface performance. Popular feature types include logarithmic band power (logBP), autoregressive (AR) parameters, time-domain parameters, and wavelet-based methods. In this study, we focused on different variants of AR models and compare performance with logBP features. In particular, we analyzed univariate, vector, and bilinear AR models. We used four-class motor imagery data from nine healthy users over two sessions. We used the first session to optimize parameters such as model order and frequency bands. We then evaluated optimized feature extraction methods on the unseen second session. We found that band power yields significantly higher classification accuracies than AR methods. However, we did not update the bias of the classifiers for the second session in our analysis procedure. When updating the bias at the beginning of a new session, we found no significant differences between all methods anymore. Furthermore, our results indicate that subject-specific optimization is not better than globally optimized parameters. The comparison within the AR methods showed that the vector model is significantly better than both univariate and bilinear variants. Finally, adding the prediction error variance to the feature space significantly improved classification results.

Exploration of the neural correlates of cerebral palsy for sensorimotor BCI control.

Cerebral palsy (CP) includes a broad range of disorders, which can result in impairment of posture and movement control. Brain-computer interfaces (BCIs) have been proposed as assistive devices for individuals with CP. Better understanding of the neural processing underlying motor control in affected individuals could lead to more targeted BCI rehabilitation and treatment options. We have explored well-known neural correlates of movement, including event-related desynchronization (ERD), phase synchrony, and a recently-introduced measure of phase dynamics, in participants with CP and healthy control participants. Although present, significantly less ERD and phase locking were found in the group with CP. Additionally, inter-group differences in phase dynamics were also significant. Taken together these findings suggest that users with CP exhibit lower levels of motor cortex activation during motor imagery, as reflected in lower levels of ongoing mu suppression and less functional connectivity. These differences indicate that development of BCIs for individuals with CP may pose additional challenges beyond those faced in providing BCIs to healthy individuals.

SCoT: a Python toolbox for EEG source connectivity

Analysis of brain connectivity has become an important research tool in neuroscience. Connectivity can be estimated between cortical sources reconstructed from the electroencephalogram (EEG). Such analysis often relies on trial averaging to obtain reliable results. However, some applications such as brain-computer interfaces (BCIs) require single-trial estimation methods. In this paper, we present SCoT—a source connectivity toolbox for Python. This toolbox implements routines for blind source decomposition and connectivity estimation with the MVARICA approach. Additionally, a novel extension called CSPVARICA is available for labeled data. SCoT estimates connectivity from various spectral measures relying on vector autoregressive (VAR) models. Optionally, these VAR models can be regularized to facilitate ill posed applications such as single-trial fitting. We demonstrate basic usage of SCoT on motor imagery (MI) data. Furthermore, we show simulation results of utilizing SCoT for feature extraction in a BCI application. These results indicate that CSPVARICA and correct regularization can significantly improve MI classification. While SCoT was mainly designed for application in BCIs, it contains useful tools for other areas of neuroscience. SCoT is a software package that (1) brings combined source decomposition and connectivtiy estimation to the open Python platform, and (2) offers tools for single-trial connectivity estimation. The source code is released under the MIT license and is available online at github.com/SCoT-dev/SCoT.

Towards a framework based on single trial connectivity for enhancing knowledge discovery in BCI

We developed a framework for systematic evaluation of BCI systems. This framework is intended to compare features extracted from a variety of spectral measures related to functional connectivity, effective connectivity, or instantaneous power. Different measures are treated in a consistent manner, allowing fair comparison within a repeated measures design. We applied the framework to BCI data from 14 subjects recorded on two days each, and demonstrated the frameworks feasibility by confirming results from the literature. Furthermore, we could show that electrode selection becomes more focal in the second BCI session, but classification accuracy stays unchanged.

Brain-Computer interfacing for users with cerebral palsy, challenges and opportunities

It has been proposed that hybrid Brain-computer interfaces (hBCIs) could benefit individuals with Cerebral palsy (CP). To this end we review the results of two BCI studies undertaken with a total of 20 individuals with CP to determine if individuals in this user group can achieve BCI control. Large performance differences are found between individuals. These are investigated to determine their possible causes. Differences in subject characteristics are observed to significantly relate to BCI performance accuracy. Additionally, significant relationships are also found between some subject characteristics and EEG components that are important for BCI control. Therefore, it is suggested that knowledge of individual users may guide development towards overcoming the challenges involved in providing BCIs that work well for individuals with CP.

Online visualization of brain connectivity.

BACKGROUND While visualization of brain activity has well established practical applications such as real-time functional mapping or neurofeedback, visual representation of brain connectivity is not widely used. In addition, technically challenging single-trial connectivity estimation may have hindered practical usage of connectivity in online applications. NEW METHOD In this work, we developed algorithms that are capable of estimating and visualizing (effective) connectivity between independent cortical sources during online EEG recordings. RESULTS The core routines of our procedure, such as CSPVARICA source extraction and regularized connectivity estimation, are available in our open source Python-based toolbox SCoT. We demonstrate for the first time that online connectivity visualization is feasible. We show this in a feasibility study with twelve participants performing two different tasks, namely motor execution and resting with eyes open or closed. Connectivity patterns were significantly different between two motor tasks in four participants, whereas significant differences between resting task patterns were found in seven participants. COMPARISON WITH EXISTING METHODS Existing connectivity studies have focused on offline methods. In contrast, there are only a small number of examples in the literature that explored online connectivity estimation. For example, a system based on wearable EEG has been demonstrated to work for one subject, and the Glass Brain project has received considerable attention in popular sciences last year. However, none of these attempts validate their methods on multiple subjects. CONCLUSIONS Our results show that causal connectivity patterns can be observed online during EEG measurements, which is a first step towards real-time connectivity analysis.