چكيده به لاتين
Decoding brain signals is a cornerstone of neuroscience, particularly in unraveling the complexities of movement-related signals. These investigations have been pivotal in propelling the development of brain-computer interface (BCI) systems. Recently, neuroscience has embraced a holistic paradigm, portraying the nervous system as a complex network governed by linear dynamic system (LDS) principles. This viewpoint posits that neural network behavior is shaped by hidden variables, giving rise to dynamic patterns intrinsic to neural activity. Consequently, this framework guides data processing and output generation within this intricate system. This study capitalizes on the advantages of the dynamic system approach, aiming to illuminate the nervous system's processing model, condense data into meaningful representations, and explore dynamics within the state-space. Specifically, our objective is to decode movement patterns using neuronal activity data sourced from the primary motor cortex (M1) during center-out reaching task. By employing the neuronal dynamic filter (NDF) to model the dynamic rules governing neuronal population activity, we assess its effectiveness in predicting hand position and velocity. Furthermore, we explore the performance of the NDF-based method with two other dimensionality reduction techniques, Principal Component Analysis (PCA) and Canonical Correlation Analysis (CCA). Our findings reveal that the NDF method, utilizing eight dimensions, significantly outperforms alternative approaches in decoding hand position, achieving average correlation coefficients of 0.917 and 0.831 horizontally and vertically, respectively. In decoding velocity, although the Kalman filter with input from canonical components slightly outperforms the dynamical filter, yielding average correlation coefficients of 0.872 and 0.762 in the horizontal and vertical directions, the latter provides a more meaningful understanding of neural processing mechanisms, as demonstrated in state-space plots. In conclusion, our results offer compelling evidence of dynamic rules shaping neuronal population activity, presenting promising avenues for decoding endeavors. This research underscores the importance of integrating dynamic system principles in comprehending and harnessing brain signals for applications like BCI systems.
