| The application of multi-channel recording electrodes is expanding rapidly, as neuroscientists seek methods to understand how multiple neurons carry out brain functions. In this dissertation, methods are described to improve the recording properties of micro-machined silicon-substrate electrodes. The methods are based on multi-channel signal processing techniques and empirical determination of optimal recording surface area. Also described is the analysis of neurons recorded in the dorsal cochlear nucleus, a region of the auditory system, revealing a possible inhibitory interaction between two specific cell types.; Background noise was found to obscure action potentials of neurons recorded in the auditory system, particularly in the presence of a synchronizing stimulus. Using silicon electrodes with closely-spaced recording elements, the signal-to-noise ratio was substantially improved by application of a linear array signal processing technique. The denoising process resulted in more easily detected spike events, with signal-to-noise ratio improvements as large as 12 dB. The spikes were subsequently classified into individual neural sources by multi-channel template or principal component methods. Refinements to the array processing method were made by expansion to the spatio-temporal domain and by an adaptive processing procedure.; Design optimization experiments involved the use of a custom electrode having recording elements of various surface areas along its length. The electrode was moved slowly past a neuron, allowing an empirical determination of the electrode size yielding the best signal-to-noise ratio of detected spikes. The optimal electrode surface area was found to be significantly larger than that currently in use. Iridium activation of the electrodes did not improve the signal-to-noise ratio.; The optimized recording methods were applied to neural ensembles of the guinea pig dorsal cochlear nucleus. To verify previous theories that an inhibitory interaction between cartwheel and fusiform cell types exists, we recorded simultaneously from these units using a 16-channel electrode. Cartwheel cells were identified by their tendency to fire spikes in bursts, and fusiform cells by their high spontaneous rate and response to noise and tones. A small number of putative cartwheel-fusiform pairs exhibited a significant feature in cross-correlograms derived from spontaneous activity, consistent with a monosynaptic inhibitory connection. |
