Precision Decomposition Technique
for EMG Signals

Figure 1: Three channels of EMG signal recorded with the quadrifilar electrode. The signals were bandpassed from 1 kHz to 10 kHz. Note that the shapes of the motor unit action potentials are different in the individual channels. This is an important feature that enables the computer algorithms to decompose the signals.
Please click here to see a large version of video in a new window.

Figure 2: Shapes of the motor unit action potentials extracted from the signals on the left. Note that 17 different motor unit action potentials were identified, but as may be seen in Figure 3, not all could be traced continuously.

Figure 3a: The BAR PLOT of the old system, Precision Decomposition I. These are the motor unit action potential firings in the EMG signal shown in Figure 1.  The signal was first decomposed automatically and then it was edited by an experienced operator employing rules.   A bar is placed at the time a motor unit firing is identified. The colored bars correspond to correctly identified firings and the black ones indicate errors. The latter are false positives. False negatives appear as gaps in the train. The signal was first decomposed in an automatic mode and then edited with operator assistance. Only the first 8 Motor units are correctly identified; the others contain numerous errors and may not even be individual motor units. The solid line represents the force produced by the muscle.

Figure 3b: The BAR PLOT of the new system, Precision Decomposition II. These motor unit firings were obtained from the same EMG signal as Figure 3a.  The EMG signal was decomposed automatically; there was no operator assisted editing. Note that 14 motor units were identified and there are relatively fewer errors.

Figure 4a: The DOT PLOT of Precision Decomposition I. These are the same data as in figure 2a. Here the time interval between firings is plotted vertically. This presentation is helpful for visualizing the regularity of the firings. Note that there are two categories of trains: those that have relatively regular firing patterns and those that appear as very irregular. The later contain numerous low-level and high-level firing intervals. This pattern occurs when firings are incorrectly allocated to the wrong motor unit.

Figure 4b: The DOT PLOT of the Precision Decomposition II. Note that it contains far fewer low-level and high-level intervals. This is indicative of fewer errors.

Figure 5: The FIRING RATE PLOT of Precision Decomposition I (left) and Precision Decomposition II (right). The same 8 motor units are presented in both plots. Note that they are nearly identical. This is remarkable considering that Precision Decomposition II ran completely in the automatic mode.