Firmas neuronales y decodificación multicognitiva de señales EEG inducidas por estímulos compartidos: un estudio...

Multi-task decoding from electroencephalogram (EEG) signals is valuable for brain-computer interface (BCI) applications in naturalistic settings. Most existing studies focus on decoding distinctly different tasks, leaving the diversity of cognitive responses elicited by a single stimulus underexplored. We introduced a novel experimental paradigm where a common visual stimulus elicits five distinct cognitive processes: single reach, interception reach, sequence reach, attention reach, and inhibition reach. EEG signatures were analyzed using temporal and spectral methods. A regularized linear discriminant analysis (RLDA) classifier was employed for decoding, utilizing both temporal and event-related spectral perturbation (ERSP) features. Significant neural activation differences (p < 0.05) were observed across tasks and brain regions. The RLDA classifier achieved high decoding accuracy: 91.72% ± 6.10% for classifying the five cognitive states using ERSP features. Furthermore, for the sequence reach task, temporal features enabled classification of normal versus catch trials with 77.96% ± 7.03% accuracy. All these results demonstrate the potential for EEG-based BCI applications to distinguish diverse cognitive states elicited by identical stimuli, offering new insights for improving the naturalness and intelligence of BCI systems. Future work will focus on enhancing decoding performance and extending this research to online applications.