Noise tolerance as a measure of channel discrimination for multi-channel neural interfaces

Spatial selectivity can be obtained when recording the activity of a peripheral nerve or corticospinal pathways of the spinal cord by circumferential placement of the metal contacts around the axon bundle. Selectivity indices that have been proposed, however, do not measure how the channel discriminability deteriorates in the presence of noise. The pattern of amplitude distribution across the recording sites during a neural firing (a vector) can be considered as a symbol received at the end of an information channel. Thus, the performance of the neural interface/recording method can be quantified for the noisy case and compared with others using the classic formulae for the information channels. Monte Carlo simulations in this study show that the decay of information transfer rate with noise can differentiate between neural interfaces that have identical spatial selectivity indices based on the Euclidian distance measure. Noise tolerance can be the method of choice to assess the performance of multi-channel neural interfaces in terms of channel discrimination.