Waveform analysis of compound nerve action potentials: a computer simulation.

Abstract

A compound nerve action potential (CNAP) recorded from surface of the skin is composed of single nerve fiber action potentials propagating at various conduction velocities (CVs). Based on this assumption, we constructed two CNAPs simulating those at the digit elicited by stimulation of the median nerve at the wrist (CNAP-Wr) and the elbow (CNAP-El). Distribution of the nerve fiber CVs used for the simulation was estimated from actual recordings of 12 healthy subjects. Mean ratio of the CNAP amplitude (CNAP-Wr/CNAP-El) was 2.2, and that of the duration was 0.76. First, to analyze how CV distributions affect CNAP waveforms artificially in our computer program, we altered mean CV or standard deviation (SD) of the estimated distribution, and reconstructed CNAPs. Results indicated that as the mean CV was made slower, the ratio of the CNAP amplitude increased and that of the duration decreased. Similar changes of the ratios were also shown when the SD of the CV distribution was increased without changing the mean CV. Secondly, to make constructed CNAPs more similar to actual recordings, we added artificial noise to the constructed waves and had 11 electromyographers read onset and peak latencies. The average of maximum CVs calculated from the onset latencies of CNAPs-Wr was 62m/sec, and that from the latency difference between CNAPs-Wr and -El was 70m/sec. Both CVs were slower than the genuine maximum (77m/sec) given to the simulation program. The mean CV calculated from latency difference of the major positive peaks of the CNAPs was 62m/sec being close to the mode (62m/sec) or mean (60m/sec) value of the given distribution.