THE DETECTION AND CLASSIFICATION OF SOMATOSENSORY EVOKED POTENTIALS BASED ON NEO AND ICA FOR MULTICHANNEL INTRACORTICAL RECORDINGS

Abstract

In this study, we apply a multichannel integrated system to record and analyze intracortical neural signals from the primary somatosensory cortex of rats. Four neural signals are evoked by without stimulation, by stimulation using a toothbrush, pen shaft, and needle. These signals are processed according to the presented procedures. First, spectral subtraction is used to reduce noise and then the nonlinear energy operator is adopted to detect spikes. This process is the signal preprocessing. Independent component analysis is performed with dynamic dimension increase to extract the features and form a feature vector. Then, k-means is employed to group the feature vector into different clusters. The 100% of signals intercepted without stimulation that we observe are separated into Cluster 1; the 67% evoked signals of stimulation by using a toothbrush are divided into Cluster 2; and the 73% evoked signals of stimulation by using a needle are separated into Cluster 4. Some of the features of evoked signals stimulated using a pen shaft are similar to those stimulated by using a needle or a toothbrush. The monitoring subsystem records synchronously the timing of the external stimuli, the waveform of the evoked potentials, the audio of the neural signals, and the action of an experimental rat using a video recording device. The information is applied to assist us in proving of experimental results. Finally, the presented methods are utilized to extract the features from various evoked potentials and distinguish the stimulants from different sensory signals.