Motor unit conduction velocity distribution estimation: assessment of two short-term processing methods

Muscle fibre conduction velocity (MFCV) can be used as an index of the structural and/or functional modifications that can occur during fatigue or pathological processes. Current evaluation of MFCV from surface electromyography (SEMG) classically produces an average value. However, a single mean value is not sufficient when modifications affect only a small part of the conduction velocity distribution. In such a case, an estimation of the whole motor unit conduction velocity distribution (MUCV) would be advantageous. The aim of this study was the evaluation of the quality of two short-term methods based on cross-correlation (CC) and peak-to-peak (PP) estimation. A comprehensive simulation program was used to generate signals with known MUCV distributions. The Dmax statistic of Kolmogorov-Smirnov was used as an error criterion to quantify the estimation error and to optimise the MUCV distribution computation algorithms. The minimum error was observed for an analysing window of 10ms for PP and 15ms for CC. Dmax was significantly lower for PP (0.195+/-0.054) than for CC (0.343+/-0.073). Various simulations showed the strong effect of the variance of the true distribution on the features of the estimated ones. Clinical data measured on the abductor pollicis brevis were studied. MUCV was estimated on a healthy subject (3.63+/-0.87ms(-1)), a patient suffering from a myopathy (2.73+/-0.51ms(-1)) and one suffering from a neuropathy (4.38+/-0.23ms(-1)). The results demonstrate the overall superiority of a peak-to-peak approach.