Feature detection in motor cortical spikes by principal component analysis.

Principal component analysis was performed on recorded neural spike trains in rats' motor cortices when rats were involved in real-time control tasks using brain-machine interfaces. The rat with implanted microelectrode array was placed in a conditioning chamber, but freely moving, to decide which one of the two paddles should be activated to shift the cue light to the center. It is found that the principal component feature vectors revealed the importance of individual neurons and windows of time in the decision making process. In addition, one of the first principal components has much higher discriminative capability than others, although it represents only a small percentage of the total variance in the data. Using one to six principal components with a Bayes classifier achieved classification accuracy comparable to that obtained by a more sophisticated high performance support vector classifier.