Leveraging Grating-Based Flickers: A Leap Toward Practical, Visually Comfortable, and High-Performance Dry EEG Code-VEP BCI

Frederic Dehais,Kalou Cabrera Castillos,Simon Ladouce,Pierre Clisson

DOI: https://doi.org/10.1101/2024.07.17.603960

2024-07-22

Abstract:Purpose: Reactive Brain-Computer Interfaces (rBCIs) typically rely on repetitive visual stimuli, which can strain the eyes and cause attentional distraction. To address these challenges, we propose a novel approach rooted in visual neuroscience to design visual Stimuli for Augmented Response (StAR). The StAR stimuli consist of small randomly-oriented Gabor or Ricker patches that optimize foveal neural response while reducing peripheral distraction. Methods: In a factorial design study, 24 participants equipped with an 8-dry electrodes EEG system focused on series of target flickers presented under three formats: traditional 'Plain' flickers, Gabor-based, or Ricker-based flickers. These flickers were part of a five-classes Code Visually Evoked Potentials (c-VEP) paradigm featuring low frequency, short, and aperiodic visual flashes. Results: Subjective ratings revealed that Gabor and Ricker gratings were visually comfortable and nearly invisible in peripheral vision compared to plain flickers. Moreover, Gabor and Ricker-based textures achieved higher accuracy (93.6% and 96.3%, respectively) with only 88 seconds of calibration data, compared to plain flickers (65.6%). A follow-up online implementation of this experiment was conducted to validate our findings within the frame of naturalistic operations. During this trial, remarkable accuracies of 97.5% in a cued task and 94.3% in an asynchronous digicode task were achieved, with a mean decoding time as low as 1.68 seconds. Conclusion: This work demonstrates the potential to expand BCI applications beyond the lab by integrating visually unobtrusive systems with gel-free, low density EEG technology, thereby making BCIs more accessible and efficient. The datasets, algorithms, and BCI implementations are shared through open-access repositories.

Neuroscience