Principle component feature detector for motor cortical control

Principle component analysis (PCA) was performed on recorded neuronal action potentials from neural ensembles in rat´s motor cortex when the rat was involved in a closed-loop real-time brain machine interface (BCI). The implanted rat was placed in a conditioning chamber, but freely moving, to decide which one of the two paddles should be activated to shift the light to the center. It is found that the principle component feature vectors revealed the importance of individual neurons and their temporal dynamics in relation to the intention of activating either left or right paddle. In addition, the first principle component feature has much higher discriminative capability than others although it represents only a few percentage of the total variance. Using the first principle component with the Bayes classifier achieved 90% classification accuracy, which is comparable with the accuracy obtained by a more sophisticated high performance support vector classifiers.