Brain-Computer Interface (BCI) translates brain activity into commands for control and communication. However, the high inter-subject variability limits its efficiency. Most BCI studies use local measurements without considering the interconnected nature of brain functioning. As a result, mastering non-invasive BCI systems remains a learned skill that yields suboptimal performance in ~30% of users, referred to as the “BCI inefficiency” phenomenon [1]. In this study, we are therefore considering how communication between brain areas impacts BCI performance by studying neuronal avalanches that can be described as a non-linear dynamic biomarker where neurons fire together in a cascade-like pattern following an inverse power law [2]. To test our hypothesis, we are creating a BCI pipeline with neuronal avalanches as features and comparing this with state-of-the-art methods.