Improved signal and reduced noise in neural recordings from close-spaced electrode arrays using independent component analysis as a preprocessor.

Noise can greatly complicate the isolation of individual cell action potential waveforms for the sake of electrophysiological analysis. For an experiment involving recording in the thalamus/subthalamus areas of a rat brain, a hybrid hardware/software method was utilized to improve the signal-to-noise quality of the recorded signal on each recording channel. The procedure uses closely spaced recording electrode arrays and independent component analysis (ICA) to fortify the signal energy of a single spike by combining input from several channels, and concurrently to reduce the noise on each channel by isolating common mode components such as artifacts, slow waves, and correlated distant spike activation. In the next step, a wavelet denoising-based signal-to-noise assessment is used to quantify the improvement in data quality for each data record. The data presented here demonstrate that this method, which can be applied off-line as a preprocessor to other time domain or transform domain spike sorting methods, is consistently effective at improving data quality and facilitating subsequent detection and classification of neurons.