Selection of filtering frequencies for neuronal spike signals

In order to determine the proper filtering frequency ranges for accurately recording and analyzing the extracellular neuron unit action potential (i.e., spike), changes in spike waveform distortion, signal-to-noise ratio (SNR), and accuracy of spike classification under different lower and higher cut-off frequencies were investigated. The spike signals were recorded from rat hippocampal regions by microelectrode arrays. The results show that with lower cut-off frequency ≤100 Hz and higher cut-off frequency ≥5 kHz, the spike waveform distortions caused by filtering are small. The optimized higher cut-off frequency ranges for spike detection and spike classification are both at 3-5 kHz. However, the optimized lower cut-off frequency for spike detection is 500-600 Hz, whereas for spike classification the frequency is about 200 Hz. The reason is that the accuracy of spike classifications depends on both the SNR and the waveform distortion.