Improved Neural Unit Separation Using Enhanced Correlation Techniques.
Robert Wesley Bossemeyer · Deep Blue (University of Michigan) · 1981
The purpose of this dissertation is to develop a signal processing technique for measuring nerve fiber conduction latencies in a nerve bundle. The objective is to improve the latency resolution of the cross correlation method by processing nerve bundle correlograms in the frequency domain. Nerve bundle spectra are obtained by the Fourier transform of correlograms computed from data sampled at two locations on the nerve bundle. The nerve bundle cross spectrum is divided by the average of the distal and proximal auto spectra then inverse Fourier transformed to produce nerve fiber latency estimates. This procedure is called the processed correlogram multiple latency estimate (PCMLE). A mathematical model of the nerve bundle was analyzed to show that nerve fiber action potential shape and conduction latency can be separated in the frequency domain. Nerve bundle correlograms of computer simulated nerve fiber signals were processed to test the PCMLE. It was demonstrated that the resolution of closely spaced nerve fiber conduction latencies was 3.4 to 6.8 times better for the PCMLE than for the nerve bundle cross correlogram. The PCMLE also has less variance than the cross correlation estimate. Computer programs for high speed data acquisition and processing of nerve bundle signals were written and tested using an electronic nerve bundle simulation developed for this research. The PCMLE was verified with simulated nerve fiber signals of known latency. The PCMLE was computed for nerve bundle signals from cat peripheral nerve. Nerve fiber latencies were correctly identified in the PCMLE that could not be resolved in the nerve bundle cross correlogram. Three examples of nerve bundle signals from the cat medial articular nerve show conclusively that the PCMLE can identify latencies not shown in the cross correlogram. These latencies were verified by identifying a number of single fiber action potentials in the distal and proximal data records and compiling a histogram of conduction latencies. The PCMLE is a better technique than the nerve bundle cross correlogram for measuring closely spaced nerve fiber conduction latencies in a nerve bundle.