Exploration of the mental attention mechanisms in motor imagery-based EEG decoding.

BACKGROUND: Brain-Computer Interface (BCI) systems enable direct communication between the brain and external devices, with motor imagery (MI)-based BCIs as a key paradigm. Although decoding neural signals has advanced via machine learning and deep learning, the influence of human factors, especially mental attention on performance remains underexplored. NEW METHOD: This study quantitatively investigates how mental attention modulates MI decoding. Specifically, it examines the enhancement of Common Spatial Pattern (CSP) features under high attention and evaluates attention-based data selection as a decoding criterion. RESULTS: Experimental results demonstrate that applying mental attention as a trial selection strategy (Strategy 2) markedly improves MI decoding performance, yielding an 11.6% increase relative to the baseline accuracy of 61.3% observed without attention. These findings highlight that integrating real-time mental attention monitoring into BCI systems can enhance decoding robustness and stability, paving the way for personalized and context-aware brain-computer interactions in neurorehabilitation, cognitive training, and intelligent assistive technologies. COMPARISON WITH EXISTING METHODS: Prior studies focused largely on algorithmic innovations. In contrast, this work adopts a user-centric perspective, showing that attention-informed trial selection significantly improves performance even within standard CSP-based pipelines. CONCLUSIONS: Incorporating mental attention into decoding frameworks enhances MI-BCI performance. This approach may improve the robustness and user-adaptability of online BCI systems, contributing to more effective and user-friendly neurotechnology.