Figure 1.
Schematic representation of the model.
A, types of synaptic inputs incorporated in the model. B, equivalent circuit of the motoneuron model (soma compartment).
Table 1.
Motor unit statistics for the 8 subjects.
Figure 2.
Representative example of the behavior of the correlation measures for seven motor unit spike trains recorded from subject #4.
A, spike trains of seven motor units filtered with 1 ms, 150 ms and 400 ms duration low-pass windows (shown on top). The spike trains are ordered in decreasing discharge rates (6–13 pps). B, cross-correlation functions between pairs of the same filtered spike trains. Note the dependence of the peak correlation on the average discharge rates of the pairs and the length of the filters.
Figure 3.
Association between the discharge rates of the pairs of spike trains for four correlation indexes calculated from subject #6.
A, scattered plot of the common drive index calculated using an Hann window of 150 ms duration and the geometric mean of the discharge rates of the pairs of spike trains used in the calculation (Rˆ2 = 0.14, P<0.05). B, the same but using a longer window of 400 ms duration (P>0.05). C, scattered plot of the CIS index and the geometric mean of the discharge rates (Rˆ2 = 0.13, P<0.05). D, scattered plot of the SIP index and the geometric mean of the discharge rates (P>0.05).
Table 2.
Correlation coefficients between motoneuron spike trains for the 8 subjects.
Figure 4.
Normalized correlation values and coherence using the pooling of multiple spike trains.
A, results from the simulations using the band-pass Gaussian noise in the frequency band 0–100 Hz. B, results from the experimental recording on Subject n. 7, the one whit the highest number of motor unit spike trains correctly identified. Note a trend of saturation for all indexes when more than 4–5 motoneuronspike trains were used in the calculations.
Figure 5.
Coherence functions using pooled spike trains.
A, magnitude of coherence for the simulation data (band-pass common synaptic Gaussian noise in the range 0–100 Hz) using 1, 3 and 5 pooled spike trains. Single pair combinations (light grey) and average using all available combinations (black line) B, same results for the experimental motor unit spike trains recorded from subject n. 2. The confidence level is shown with a dashed line.
Figure 6.
Peak of correlation magnitude between spike trains filtered with Hann windows of different lengths.
A, simulations results using combinations of 2, 3 and 7 pooled spike trains. B, inset for the filter lengths in the range 20–200 ms. C, experimental results for subject n. 3. D, inset for the filter length in the range 20–140 ms. Values are reported as the mean across all combinations of the specified number of spike trains. Notice the non-monotonic behavior when the filter length is approximately equal to the mean inter-spike intervals (97±17 ms for simulations and 72±19 ms for experimental).
Figure 7.
Comparison between time and frequency domain correlation for different filter lengths.
A, impulse response of the three filters used in the study: rectangular window of 1 ms (equivalent to raw spike train with 1000 Hz sample rate), Hann window of 150 ms and 400 ms durations. B, transfer function for the same filters. C, cross-correlation functions calculated for two spike trains recorded from subject 7 and filtered with the above windows. D, comparison of the coherence functions for the pair of spike trains filtered with the rectangular window (black line) and the other two filters.