Improving the Efficacy of ERP-Based BCIs Using Different Modalities of Covert Visuospatial Attention and a Genetic Algorithm-Based Classifier

Mauro Marchetti, Francesco Onorati, Matteo Matteucci, Luca Mainardi, Francesco Piccione, Stefano Silvoni, Konstantinos PriftisPLoS 2013 · DOI

Improving ERP-Based BCI Efficacy with Voluntary Covert Attention and Genetic Algorithms

Abstract

We investigated whether the covert orienting of visuospatial attention can be effectively used in a brain-computer interface (BCI) guided by event-related potentials (ERPs). Three visual interfaces were tested: one that activated voluntary orienting of visuospatial attention and two that elicited automatic orienting. We compared two classification procedures: an online method using Independent Component Analysis (ICA) and Support Vector Machines (SVM), and an offline method using a Genetic Algorithm (GA) for personalized feature extraction. The offline classification revealed a significant performance advantage for the "voluntary" interface, supported by neurophysiological data. Furthermore, the GA classifier outperformed the ICA-based approach, suggesting that combining voluntary attention with personalized feature extraction can lead to more efficient BCI control.

Introduction

Brain-Computer Interfaces (BCIs) offer a vital communication channel for individuals with severe motor impairments, such as those in the late stages of Amyotrophic Lateral Sclerosis (ALS) or Locked-In Syndrome (LIS). Many visual BCIs rely on "overt" attention, requiring physical eye movements toward a target. However, as these conditions progress, eye movement control is often lost. To address this, researchers are developing BCIs based on "covert" attention—the ability to focus on peripheral stimuli while maintaining gaze on a central point.

This study explores two key factors to improve covert BCI performance: the cognitive strategy used to orient attention (voluntary vs. automatic) and the signal processing method used to decode brain waves (standard fixed features vs. personalized genetic algorithms).

Experimental Design and Methods

The study involved 12 healthy participants who interacted with three distinct visual interfaces designed to trigger different types of covert attention:

  • "Arrows" (Automatic): Peripheral arrows changed color to trigger an automatic shift in attention.
  • "Auto" (Automatic): Peripheral icons briefly disappeared, eliciting an automatic response.
  • "Vol" (Voluntary): A central letter appeared at the fixation point, instructing the user to voluntarily attend to a specific peripheral icon.
Comparison of the three visual interfaces: Arrows (left), Auto (middle), and Vol (right).

Each trial followed a specific temporal sequence to ensure gaze remained fixed while attention shifted covertly. Eye movements were monitored via electrooculography (EOG) to confirm that participants were not looking directly at the targets.

The temporal sequence of a trial, illustrating the stimulus presentation and the maintenance of central gaze.

Classification Strategies

Two different pipelines were used to classify the EEG signals:

  1. ICA Classifier (Online): This method used Independent Component Analysis to remove artifacts, followed by an SVM using a standard, fixed set of signal features.
  2. GA Classifier (Offline): This approach employed a Genetic Algorithm to "evolve" a personalized set of features for each participant. By treating feature selection as an evolutionary process, the system identified the specific timing and wave shapes unique to an individual's brain response.

Performance was measured using the F-measure, which provides a more accurate assessment than simple accuracy in datasets where non-target stimuli significantly outnumber targets.

Key Results and Neurophysiological Insights

The study yielded several critical findings:

  • Superiority of Voluntary Attention: In offline analysis, the "Vol" interface achieved a significantly higher F-measure (0.67) compared to the automatic interfaces (0.55). This suggests that mental effort directed by a central cue creates a clearer neural signal than automatic triggers.
  • The GA Advantage: The personalized Genetic Algorithm classifier significantly outperformed the standard ICA/SVM approach. For the voluntary interface, the GA's F-measure of 0.67 far exceeded the ICA's 0.51.
  • Noise Suppression: Neurophysiological data showed that voluntary attention acts as a powerful filter. The Late Negative Component (LNC) amplitude for non-target stimuli was significantly lower in the "Vol" condition (0.87 µV) compared to the "Auto" condition (-2.47 µV). This indicates that voluntary focus helps the brain actively inhibit distracting, non-target information.

Conclusion and Future Directions

The combination of voluntary covert attention and personalized feature extraction via Genetic Algorithms represents a significant step forward for non-invasive BCIs. While the GA results were obtained offline, they demonstrate the potential for "ad personam" systems that adapt to the user's unique neurophysiology.

Future work will focus on implementing the GA classifier for real-time, online use and validating these strategies with patients in locked-in states, for whom covert attention BCIs are a critical necessity.

Summary updated

Abstract

We investigated whether the covert orienting of visuospatial attention can be effectively used in a brain-computer interface guided by event-related potentials. Three visual interfaces were tested: one interface that activated voluntary orienting of visuospatial attention and two interfaces that elicited automatic orienting of visuospatial attention. We used two epoch classification procedures. The online epoch classification was performed via Independent Component Analysis, and then it was followed by fixed features extraction and support vector machines classification. The offline epoch classification was performed by means of a genetic algorithm that permitted us to retrieve the relevant features of the signal, and then to categorise the features with a logistic classifier. The offline classification, but not the online one, allowed us to differentiate between the performances of the interface that required voluntary orienting of visuospatial attention and those that required automatic orienting of visuospatial attention. The offline classification revealed an advantage of the participants in using the "voluntary" interface. This advantage was further supported, for the first time, by neurophysiological data. Moreover, epoch analysis was performed better with the "genetic algorithm classifier" than with the "independent component analysis classifier". We suggest that the combined use of voluntary orienting of visuospatial attention and of a classifier that permits feature extraction ad personam (i.e., genetic algorithm classifier) can lead to a more efficient control of visual BCIs.
BibTeX
@inproceedings{marchetti2013improving,
  author = {Mauro Marchetti and Francesco Onorati and Matteo Matteucci and Luca Mainardi and Francesco Piccione and Stefano Silvoni and Konstantinos Priftis},
  title = {{Improving the Efficacy of ERP-Based BCIs Using Different Modalities of Covert Visuospatial Attention and a Genetic Algorithm-Based Classifier}},
  booktitle = {PLoS},
  year = {2013},
  doi = {10.1371/journal.pone.0053946},
}
Figure