Comparison of Spike Detectors based on Simultaneous Intracellular and Extracellular Recordings

The ability to accurately detect action potentials in cortical in vivo recordings is a necessary first step in any multichannel wireless brain-machine interface (BMI) data acquisition system. This work presents a comparison of simple spike detection algorithms appropriate for implementation in an autonomous low power wireless BMI chip. The detectors were applied to pre-recorded cortical extracellular potentials. A simultaneously recorded intracellular transmembrane voltage gave the precise spike times of a local neuron; these times were then used as the "gold standard" against which to compare the output of the spike detectors under test. In contrast to earlier work on simulated data that showed the superiority of a simple absolute value-based spike detector, this work demonstrated that the non-linear energy operator provides an effective balance between correct detections and false alarms, and that the relative difference between detection algorithms diminishes as SNR increases