USING COARSE-GRAINED MEASURES TO CHARACTERIZE ELECTROMYOGRAPHIC SIGNALS

Human electromyograms were obtained from normal muscle and from spontaneously fibrillating, denervated muscle, and the time intervals between spikes in electrical activity were measured. The entropy and measures related to the entropy were used to examine these data. We have achieved a statistically significant discrimination between normal and denervated muscle (p<0.0002) using a measure (the second moment of the distribution) that requires less than 200 consecutive interspike intervals in a calculation that requires less than 2 milliseconds. This suggests that real time monitoring may be feasible. The range of values obtained for the second moment for the two conditions (normal versus denervated) did not overlap. Complete separation was achieved. A measure analogous to the two-dimensional product of inertia which is sensitive to dynamical structure in the data was also calculated. Again, complete separation of values obtained in the two cases was observed. Further research is required to determine if these measures can aid in the differential diagnosis of neuromuscular disorders and in the longitudinal assessment of the efficacy of treatment. In this review we also establish the relationships between conventional statistical measures and the limiting values of coarse grained measures obtained as the number of elements in the partition approaches infinity.