Investigation of a Multichannel Surface Electromyogram Analysis

Method Considering Superimposed Waveforms in a Elbow Flexion

Movement

Jun Akazawa

and Ryuhei Okuno

School of Health Science and Medical Care, Meiji University of Integrative Medicine, Nantan-shi, Kyoto, Japan

Faculty of Science and Engineering, Setsunan University, Neyagawa-shi, Osaka, Japan

Keywords: Surface EMG, Superimposed Signals.

Abstract: The purpose of this study was to develop a method of decomposing the surface motor unit acton potential

(SMUAP) of a biceps brachii short head muscle when the distance from the surface electrodes to the motor

units (MUs) changes during voluntary isovelocity elbow flexion. In the preparatory study, a subject’s elbow

flexion movement had changed the shape of the SMUAP, which was probably made by a single MU larger

than the previous study. Thus, we had to develop a SMUAP decomposition method that focused on tracking

the SMUAP waveform changes and superimposed signals. The developed SMUAP decomposition

algorithm was based on a sequentially modified template matching method, considering the superimposed

signals. This was applied to the measured SMUAPs. The MU firing rates calculated with our algorithm

were almost the same as those of previous physiological studies; our algorithm was capable of decomposing

SMUAPs when the waveform of the SMUAP was generated from a single MU and responded with each

change in firing time.

1 INTRODUCTION

In physiology and medicine, methods to investigate

the behaviors of motor units (MUs) are desired.

Studies have shown that high-density surface

electrodes are suitable for analyzing the

characteristics of MUs during isometric contraction

(Merletti and Parker, 2004).

Needle electrodes have been used to analyze the

motor unit acton potential (MUAP) behavior from

the tibialis muscle during ankle joint flexion [Kato,

Murakami, and Yasuda, 1985). However, needle

electrodes restricted the angle to a small range.

To solve this problem, we used multi-channel

surface electrodes to investigate the behavior of the

MU in the biceps brachii short head muscle. Our

results showed that the firing rates (FRs) of activated

MUs were almost the same when the degree of

elbow flexion varied from 0 to 120 degree;

additionally, surface MUAPs (SMUAPs) were

identified by visual observation (Okuno, Maekawa,

Akazawa, Yhoshida, and K. Akazawa, 2005). The

measured SMUAP waveforms changed gradually;

thus, it was difficult to perform SMUAP

decomposition quantitatively.

In this study, we developed an algorithm to

decompose the SMUAPs quantitatively during

voluntary isovelocity elbow flexion. This algorithm

was based on the similar shape of SMUAP

waveforms of a single MU extracted for a short

period during isovelocity movements (Akazawa and

Okuno, 2013).

Notably, in some subjects whose fat tissue was

thin, the shape of the SMUAP was most likely

generated by a single MU; in this case, the

waveform shapes changed, making it difficult to

decompose the SMUAPs. Thus, further adjustments

to the algorithm were required to address this issue.

Akazawa, J. and Okuno, R.

Investigation of a Multichannel Surface Electromyogram Analysis Method Considering Superimposed Waveforms in a Elbow Flexion Movement.

DOI: 10.5220/0006634901950200

In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 4: BIOSIGNALS, pages 195-200

ISBN: 978-989-758-279-0

195

2 METHODS

2.1 Experimental Setup

Figure 1: Schematic overview of the experimental setup.

A schematic diagram of the experimental setup is

shown in Fig. 1. The subject was instructed to flex

the elbow joint smoothly to approximately a 110°

angle at constant angular velocity (5°/s) against 10%

of the maximum voluntary contraction (MVC). The

experiments were performed with one healthy

subject, who gave informed consent. The

investigation was approved by the local Ethics

Committee of Meiji University of Integrative

Medicine.

An eight-channel surface electrode was used.

The surface electromyogram (SEMG) signals were

amplified with a gain of 80 dB. The band pass filter

was set at 43 Hz–2.8k Hz.

2.2 Algorithm

Figure 2: Algorithm to monitor motor unit (MU) activity

continuously over a sufficiently large range of motion.

The algorithm shown in Fig. 2 was developed

based on our previous method [Akazawa J., Okuno

R., 2013], with several modifications. The algorithm

steps are described below.

When identifying the action potential of a single

motor unit, we use the SMUAP Profile to extract the

characteristics of each SMUAP.

Fig. 3 shows the method for making the SMUAP

profile [Akazawa, Sato, Minato, and Yoshida, 2005].

The SMUAP profile consists of both plus and minus

amplitude components, because the minimum

potential (

) was sometimes larger than

maximum potential (

) from the preliminary

experiment. The eight-channel SEMG is shown in

Fig. 3(a). Fig. 3(b) shows the parameter for detecting

the SMUAP. The maximum potential at t =

denoted by

, and the minimum potential at t = t

is denoted by

. The threshold parameters

were and . In Fig. 3(c), the values of

were plotted against each channel to create a

SMUAP profile.

is the maximum of

. In

Fig. 3(d), the values of

were plotted against

each channel to create a SMUAP profile.

the minimum of .

Figure 3: Method for creating an SMUAP profile from

measured surface electromyogram (SEMG) signals.

Step 1:

In Step 1, the SEMG is measured for 30.0 s. The

SEMG is used to retain the MUAPs that belong to

the target MUAP and remove both the noise and

low-amplitude MUAPs that do not reach the given

thresholds.

Step 2:

We calculate the coincidence between SMUAP

profiles to decompose the SMUAPs. The

performance index (PI) characterizing the fitness

between two SMUAPs is given by Eq. (1):

BIOSIGNALS 2018 - 11th International Conference on Bio-inspired Systems and Signal Processing

196

… (1)

If the PI is larger than the threshold value, then both

and are defined as generated by

a single MU. Both previous (Akazawa J., Okuno R.,

2013) and present decomposition methods are

shown at Fig. 4. The parameters are described as

follows.

: Potential distribution of the SMUAP.

: Number of the MU.

: firing time of the MU.

Figure 4: Surface motor unit action potential (SMUAP)

decomposition method: previous method versus present

method.

When the SMUAPs of a single MU were

decomposed in a previous study (Akazawa J., Okuno

R., 2013), did not change. In the present

study, we used a sequential method to account for

changes in . As such, our sequentially

modified template matching method was expected to

be more robust than our previous method (Akazawa

J., Okuno R., 2013)when the amplitude of the

SMUAP increases or decreases upon firing.

In Step 2, synthetic waveforms are created using

templates based on the single motor unit detected at

Step 2.

The difference between the waveforms of this

synthesized waveform and the measured

superimposed waveform are calculated and then

decomposed. As shown at Fig. 5, a superimposed

waveform is created by adding SMUAP of Motor

Unit 01 (MU01) and MU 02 which are single motor

units. Identification processing is performed by

calculating the difference between the waveforms of

the superimposed waveform and the measured

superimposed waveform.

Figure 5: Creation of superimposed waveform.

Step 3:

\In Step 3, coupling is used to connect the MUAP

trains (MUAPTs). As shown in Fig. 6, since the

muscle length changes, the time period of 30.0 s was

too long to decompose all of the SMUAPs at once.

Thus, short-period (3.0 s) signals were used. The

coupling period to connect the MUAPTs was set to

1.5 s.

To decompose the SEMG for 30.0 s, it was

necessary to merge each 3.0-s MUAPT. Fig. 6

shows a schematic diagram of this process. Each bar

represents one firing. Firings of the same MU were

aligned horizontally. Notably, at some points in Fig.

6, the firing time of MUAPT 1 was the same as in

MUAPT 2.

Investigation of a Multichannel Surface Electromyogram Analysis Method Considering Superimposed Waveforms in a Elbow Flexion

Movement

197

Figure 6: Schematic diagram for the coupling.

Step 4:

In Step 4, the firing patterns of all identified MUs

are plotted as MUAPTs to resolve the activity

patterns of the MU.

3 RESULTS

In this study, we instructed a normal subject to flex

the elbow joint from 0 to 110° against a constant

load torque of 10% MVC with an isovelocity of

5.0°/s. The SEMG signal was detected with an

electrode array from the biceps brachii short head

muscle. When the elbow joint angle reached 40° at 9

s, the active MU was observed.

As a subject flexed the elbow smoothly, the

SMUAPs that had a large amplitude at CH1 at 10 s

changed smoothly from a low channel number to a

high channel number. As the elbow angle increased,

moved from a small channel number to a

large channel number.

Figure 7: SEMG signals obtained from the biceps brachii

short head muscle during voluntary isovelocity elbow

flexion.

Figure 8: Extracting the SMUAP signals. The active MU

was observed from roughly 9 s.

Fig. 8 shows the extracted SMUAPs. Our algorithm

revealed that these SMUAPs were generated by a

single MU. On the left side of the figure, the

amplitude becomes larger over time; whereas on the

right side of the figure, the SMUAP retains the same

shape. Using our previously developed algorithm

(Akazawa J., Okuno R., 2013), we calculated the

rate of fitness between the two SMUAPs. The rate of

fitness decreased as a function of elbow flexion

angle because the SMUAP of a single MU changed

rapidly. Thus, it was necessary to improve the

previous decomposition method (Akazawa J., Okuno

R., 2013).

Figure 9: Relationship between the range and performance

index PI [%].

Fig. 9 shows the relationship between the range R

and the SMUAP profile used to extract the SMUAP

characteristics and the PI [%] calculated from the

difference of the two SMUAPs. When the range

changed from 10 to 40, the ratio increased from 9 to

10 s. On the other hand, when the range was set at 0,

PI held a nearly constant value.

BIOSIGNALS 2018 - 11th International Conference on Bio-inspired Systems and Signal Processing

198

Figure 10: Our developed algorithm decomposed the MU

action potentials.

A typical SEMG record of biceps brachii muscle

(BIC) is shown in Fig. 10. To determine whether

SMUAPs were from the same MU, individual

SMUAP shapes (channels 1–8) shapes were

compared by the algorithm. We detected nine

SMUAPs from two MUs.

Figure 11: SMUAP wave shape of the MU01 observed at

four points.

Fig. 11 shows that the SMUAP waveform of MU01

changed during elbow flexion movement. The

SMUAP of the MU01 was observed at (a) 10, (b)

15, (c) 20, and (d) 25 s. These results confirmed that

our developed method continuously tracked the

SMUAP waveform changes during elbow flexion

movement.

Figure 12: Decomposition process of the superimposed

waveform.

Fig. 12 shows the result of the decomposition

process for the superimposed waveform using the

proposed method. The target was the signal of (iii)

in Fig. 8. The blue line waveform is a measured

SEMG, and the red line waveform is created by

synthesizing the template of MU 01 and MU 02. The

degree of coincidence of the waveforms was 73.93

Fig. 13 shows the MUAPT in which the solid

lines correspond to the elbow joint angle (the

average velocity was approximately 5°/s); identified

firings of the nine MUs are shown. Each bar

represents one firing and firings of the same MUs

are aligned horizontally. The average frequency of

Motor Unit 01 (MU01) was 22.14 Hz and the

standard deviation was 18.46 Hz. The average

frequency of MU02 was 10.87 Hz, and the standard

deviation was 6.20 Hz, and the average frequency of

MU03 was 21.82 Hz, and the standard deviation was

7.18 Hz. MU01 and MU02 continued to fire for a

relatively long period of time. The results of

calculated MU’s FRs agree with the generally

accepted behavior of MU FRs.

Figure 13: MUAP Train.

4 CONCLUSIONS

In physiology and medicine, methods to investigate

the behavior of MUs are desired. We have

developed an algorithm to decompose SMUAPs

quantitatively during voluntary isovelocity elbow

flexion. However, in some subjects whose fat tissue

was thin, the shape of the SMUAP was most likely

Investigation of a Multichannel Surface Electromyogram Analysis Method Considering Superimposed Waveforms in a Elbow Flexion

Movement

199

generated by a single MU and, as such, did not

retain the same shape. For this reason, we modified

our original SMUAP decomposition method to focus

on tracking changes in the SMUAP waveform. The

newly developed algorithm used a sequentially

modified template matching method, based on

superimposed SMUAPs, to calculate the FRs of

MUs. Our results showed that the FRs were nearly

the same as those cited in previous physiological

studies. Thus, the algorithm proposed is expected to

be useful for decomposing SMUAPs when the shape

of SMUAPs are generated from a single MU and

change with each firing.

ACKNOWLEDGEMENTS

This work was supported by JSPS KAKENHI Grant

Numbers 26350678.

REFERENCES

Merletti, R. and Parker P., Eds., 2004, Electromyography

Physiology, Engineering, and Noninvasive

Applications. Piscataway, NJ: IEEE Press.

Kato M., Murakami S., Yasuda K., 1985. Behavior of

single motor units of human tibialis anterior muscle

during voluntary shortening contraction under constant

load torque. Exp Neurol., vol. 90(1), pp. 238-53.

Okuno R., Maekawa K., Akazawa J., Yhoshida M., and

Akazawa K., 2005. Motor Unit Activity in Biceps

Brachii Muscle during Voluntary Isovelocity Elbow

Flexion. IEICE transactions on information and

systems, E88-D(6), pp. 1265-1272.

Akazawa J., Okuno R., 2013. A Method for Quantitative

SEMG Decomposition and MUAP Classification

during Voluntary Isovelocity Elbow Flexion, IEEE

EMBS, 35th Annual International Conference of the,

pp.6776-6779, Osaka, Japan

Akazawa J., Sato T., Minato K., and Yoshida M., 2005.

Method of estimating location and territory of motor

units in human first dorsal interosseous muscle with

multi-channel surface electromyograms,” JSMBE, vol.

43(4), pp. 595-604., Dec., 2005. (in Japanese)

BIOSIGNALS 2018 - 11th International Conference on Bio-inspired Systems and Signal Processing

200