Test–retest reliability of electromyographicvariables of masseter and temporal muscles inpatients with cerebral palsy

Lilian Chrystiane Giannasi a,b,*, Miriam Yumi Matsui a, Fabiano Politti b,Sandra Regina F Batista a, Bruna F. Caldas a, Jose Benedito O. Amorim a,Luis Vicente Franco de Oliveira b, Claudia Santos Oliveira b,Monica Fernandes Gomes a

aBioscience, Sao Paulo State University ‘‘Julio de Mesquita Filho’’ – UNESP, SJC – Sao Jose dos Campos, BrazilbMaster’s Program in Rehabilitation Sciences, Nove de Julho University – UNINOVE, Sao Paulo, Brazil

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 8

a r t i c l e i n f o

Article history:

Accepted 18 August 2014

Keywords:

Cerebral palsy

Electromyography

Masticatory muscles

Reliability

a b s t r a c t

Introduction: The aim of this study was to evaluate the reliability of surface electromyogra-

phy of the masticatory muscles in patients with cerebral palsy.

Methods: Surface electromyography was performed over the masseter and temporal mus-

cles in 15 patients with cerebral palsy with the mandible at rest and during maximum

clenching effort in two sessions. The data were analyzed using the root mean square

amplitude, mean frequency, median frequency, zero crossings and approximate entropy.

Results: In the within-day evaluations, intraclass correlation coefficients were higher (0.80–

0.98) for the all electromyography variables and muscles during maximum clenching effort.

In the resting position, the coefficients revealed good to excellent reliability (0.61–0.95) for

root mean square, mean frequency, median frequency and zero crossings and fair to good

reliability (0.53–0.74) for approximate entropy. In the between-day evaluations, the coeffi-

cients revealed good to excellent reliability (0.60–0.86) for mean frequency, median frequen-

cy, zero crossings and approximate entropy. In the resting position, the coefficients revealed

poor to fair reliability (0.23–0.57) for all electromyography variables studied. The root mean

square had the highest standard errors during maximum clenching effort (2.37–5.91) and at

rest (1.47–6.86).

Conclusion: Mean frequency, median frequency and approximate entropy are the most

reliable variables of surface electromyography signals of the masseter and temporal mus-

cles during maximum clenching effort in individuals with cerebral palsy. These measures

can be used to evaluate the function and behaviour of the masticatory muscles in this

population following oral rehabilitation and surgical oral procedures as well as for the study

the physiology of these muscles.

# 2014 Elsevier Ltd. All rights reserved.

* Corresponding author at: R Esperanca, 265, Sao Jose dos Campos, Sao Paulo 12243-700, Brazil. Tel.: +55 12 3951 0800; fax: +55 12 3951 0800.E-mail address: odontogiannasi@uol.com.br (L.C. Giannasi).

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: http://www.elsevier.com/locate/aob

http://dx.doi.org/10.1016/j.archoralbio.2014.08.0110003–9969/# 2014 Elsevier Ltd. All rights reserved.

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 8 1353

1. Introduction

The masticatory muscles participate in complex physiological

actions, such as chewing, swallowing, speech and other

movements that depend on a precise balance between force,

velocity and amplitude. Each muscle has a specific form and

function for effective interaction with other muscles of the

stomatognathic system.1 However, these functions are com-

promised in individuals with cerebral palsy (CP).

CP is a neuromotor disability stemming from damage to a

specific region of the brain during the prenatal, perinatal or

postnatal period and is associated with sensory and motor

dysfunctions of the orofacial region, such as dysphagia,

dysarthria, drooling and difficulty chewing.2 This irreversible

condition ranges from mild to severe. Spasticity is the most

frequent manifestation of CP and impairment can extend to

oromotor function in the form of difficulty chewing, swallow-

ing and speaking. Moreover, CP is often accompanied by

involuntary movements that compromise the bite and jaw-

opening reflexes, resulting in weakness of the masticatory

muscles and difficulty coordinating these muscles.2,3 Due the

lack of motor control, individuals with CP are more likely to

exhibit temporomandibular disorder (TMD) and parafunc-

tional oral habits, such as finger/thumb sucking and bruxism

(grinding/clenching one’s teeth).4–6 Previous studies report

that the prevalence rate of signs and symptoms of TMD ranges

from 13.3% to 67.6% in individuals with CP.5,6

Surface electromyography (EMG) has been extensively and

safely used to evaluate chewing function7–9 and assess the

efficacy of different therapies, such as global postural re-

education, fixed implant support rehabilitation and the use of

botulinum toxin-A.10–13 However, intrinsic and extrinsic factors

can influence the interpretation of EMG signals. Intrinsic

neurophysiological and anatomic factors include the pH level

in muscle fibres, blood flow, number of active motor units

during muscle contraction, shape of the intracellular action

potential, distribution of motor unit discharge rates, motor unit

synchronization and muscle fibre geometry. 14 Extrinsic factors

include the size and shape of the electrodes and placement on

the skin overlying the muscle (inter-electrode distance, location

and the orientation of detection surfaces relative to the muscle

fibres).14 Thus, in clinical studies assessing changes within the

same subject over time (inter-session evaluations), small

changes in electrode placement in relation to previous positions

may lead to measurement errors, since different electrode

locations over the same muscle provide signals with signifi-

cantly different features.15,16

Different EMG variables are used to quantify muscle

activity. The time and frequency domains have been

employed with linear methods for the evaluation of EMG

signals of the masticatory muscles. However, EMG signals are

highly complex and the mechanisms underlying the genera-

tion of such signals seem to be nonlinear or even chaotic in

nature.17–19 Indeed, a number of papers have found that

nonlinear methods of EMG analysis are more sensitive to

changes in myoelectrical signals in comparison to the linear

methods.20,21 In clinical practice, nonlinear time-series anal-

yses reveal the inherent complexity of normal variability,

indicating features of motor control that are important for

physiotherapists to measure.22 Information entropy has been

proposed as a measure of irregularity (nonlinear behaviour) in

biological signals.23–25 The measurement of entropy is

reported to be a reliable method for characterizing neuromus-

cular alterations26 and approximate entropy (ApEn) provides a

general understanding of the complexity of EMG data.26–29

In statistics, ApEn is used to establish the uncertainty or

variability of a system. ApEn calculated from EMG signals is

dependent on the variability of both amplitude and frequency

and better represents this aspect than the amplitude or

frequency alone.26 Thus, ApEn can be used to quantify the

irregularity or complexity of EMG signals of the masseter and

temporal muscles in patients with CP. The combination of

linear and nonlinear measures for the characterization and

classification of EMG signals of the masticatory muscles in

such individuals may be important to the quantification and

understanding of the neurophysiological conditions of these

muscles.

The selection of a measure for research or clinical use is

motivated by several factors, including reliability. A number of

studies have substantiated the reliability and reproducibility

of EMG in the evaluation of the masticatory muscles in healthy

individuals.30–32 However, as EMG signals are influenced by

anatomic and neurophysiological factors,14 it is not possible to

state whether the reproducibility found in healthy individuals

holds true for patients with CP due to the oromotor

abnormalities in these patients.2–6 Thus, reliability studies

are particularly important for the masticatory muscles in this

population due to the potential variation in measurements

(intrinsic factors) that can to affect the interpretation of the

findings. To the best of our knowledge, no studies have

addressed the reliability of surface EMG of the mascticatory

muscles in adults with CP or the reproducibility of nonlinear

EMG variables of the masticatory muscles. It is therefore

important to determine the variation in EMG measurements

(within-day and between-day reliability) before proposing the

use of EMG as a tool for evaluating the efficacy of therapies

administered to improve the function of the masticatory

muscles in this population.

The aim of the present study was to evaluate within-day

and between-day reliability of EMG variables of the masseter

and temporal muscles in patients with CP.

2. Methods

2.1. Subjects

Twenty-three adult patients with CP were evaluated at the

Oral Special Care Clinic of the Institute of Science and

Technology – Campus Sao Jose dos Campos/UNESP (Brazil).

Only 15 individuals with spastic diparetic CP (8 males and 7

females) met the eligibility criteria. The sample was classified

using the Gross Motor Functional Classification Scale (GMFCS):

two individuals were classified on level I; two were classified

on level II; four were classified on level III; and seven were

classified on level IV. The inclusion criteria were spastic

diparetic CP, partially preserved cognitive function (ability to

respond to verbal commands, such as ‘‘open your mouth’’,

‘‘close your mouth’’ and ‘‘clench your teeth’’) and a statement

Fig. 1 – EMG signal of the right masseter (RM), right

temporal (RT), left masseter (LM) and left temporal (LT)

muscled during maximum clenching effort. The grey band

shows three-second period selected after visual inspection

(Patient with cerebral palsy – age: 23 years; body mass:

51.4 kg). Three-second period selected through a visual

inspection of the raw data.

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 81354

of informed consent signed by the participant or legal guardian

agreeing to voluntary participation in the study. The exclusion

criteria were having undergone orthodontic or orthopaedic

treatment of the jaws or therapies to reduce spasticity (e.g.,

botulin toxin) in the 6 months prior to the study.

This study received approval from the local ethics

committee (process number: 25000.058696/2010-74) and the

Brazilian National Human Research Ethics Committee (CONEP

number: 007/2011). All participants/guardians were properly

informed regarding the objectives and procedures and signed

a statement of informed consent prior to testing.

2.2. Electromyography

The EMG signals were captured using an eight-channel

module (EMG System do Brasil Ltda1) consisting of a

conditioner with a band pass filter with cutoff frequencies

at 20–500 Hz, an amplifier gain of 1000 and a common mode

rejection ratio >120 dB. All data were acquired and processed

using a 16-bit analogue-to-digital converter (EMG System do

Brasil Ltda1), with a sampling frequency 2 kHz per channel.

Active bipolar electrodes with a pre-amplification gain of 20

times were used.

2.3. Procedure

The subjects visited the laboratory on two different occasions,

with a 1-week interval between visits. During the sessions, the

participants were instructed to remain seated in a chair, feet

apart, shoulders relaxed and hands resting on thighs. The

subjects were seated in a well-illuminated, silent recording

room in a comfortable upright position with eyes open and

without head support. A short training period was conducted

prior to beginning the tests to prepare the subjects for the

activities. Explanations concerning the procedures and elec-

trode placement were given and the subjects were trained to

bite as hard as possible (maximum clenching effort [MCE]).

Pre-gelled, self-adhesive, bipolar, silver–silver chloride

electrodes (MediTrace1) were positioned over the right masse-

ter (RM), left masseter (LM), right temporal (RT) and left

temporal (LT) muscles, with an inter-electrode distance of

20 mm. The sites for the electrodes were cleaned with a cotton

ball soaked in 70% alcohol to diminish impedance. Surface

electrodes were bilaterally placed according to anatomical

references and the procedures were guided by the direction of

muscle fibres at two points: anterior temporal muscle – 2 to 3 cm

superoposterior distant to the lateral corner of the eyes in the

region of greatest evidence of muscle mass, no hair, parallel to

the muscle fibres, but with the sensing surface perpendicularly

oriented; the superficial portion of masseter – 1 to 2 cm above

the gonial angle of the mandible in the region of greatest

evidence of muscle mass, with muscle fibres parallel to the

surface. A rectangular metallic electrode measuring 3 cm by

2 cm coated with Lectron II conductive gel (Pharmaceutical

Innovations1) to increase the conduction capacity and avoid

interference from external noise was attached to the left wrist of

the volunteers for reference. The evaluations were carried out at

rest and during MCE. In session 1 (test), three readings were

performed in the resting position with a 2-min interval between

readings. After an additional 3 min, three readings were

performed during MCE, with a 5-min interval between readings.

The signals were recorded for 10 s each under each condition.

The same procedures (resting position and MCE) were repeated

after a 1-week interval in session 2 (retest).

2.4. Data processing

The EMG signals were processed using specific routines

carried out in the Matlab program, version 7.1 (The Math-

Works Inc., Natick, MA, USA). For MCE, a three-second period

was selected through a visual inspection of the raw data

(Fig. 1). A moving window was used to select the EMG signals of

the RT, RM, LT and LM muscles based on the greatest

amplitude and regularity of the four muscles simultaneously.

For the resting position, the entire 10-s period of the EMG

signal was used in the analysis.

The EMG signals were analyzed using traditional linear

analyses (amplitude and frequency domain). The complexity of

the EMG signal was analyzed using approximate entropy (ApEn)

(nonlinear analysis). The amplitude of the raw EMG signal was

defined as the root mean square (RMSraw) calculated using a 200-

msmoving window. The mean amplitudeduring the 3-s and10-s

trials recorded during MCE and the resting position, respectively,

was used for analysis. The amplitude of the EMG signal during

MCE (RMSraw � MCE) was normalized by the mean amplitude of

the three EMG signals recorded during 10 s in the resting position

(RMSraw � rest) as follows: RMSMCE = RMSraw � MCE/RMSraw � rest.

This normalization procedure is an alternative for patients with

neurological disorders.33 The amplitude of the signal in the

resting position was expressed as the percentage of the mean

RMSraw � MCE recorded in the three readings, as follows:

RMSrest = (RMSraw � rest/RMSraw � MCE) � 100.

In the analysis of the frequency domain, the power spectral

density of the EMG signal was calculated using Welch’s

averaged periodogram with a Hamming window length of

2048 points. Overlap was 50% of the window length. Mean

frequency (MNF) and median frequency (MDF) of the power

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 8 1355

spectrum were calculated. The number of zero crossings (ZCs)

was also analyzed. The ZC rate of the signals is defined as half

the number of ZCs per second.34

ApEn was calculated to quantify the irregularity or com-

plexity of the EMG signals. This analysis returns a value

between 0 and 2, with higher values reflecting greater

irregularity within the time series.34–36 Given N points and

tolerance r, ApEn (m,r,N) is approximately equal to the negative

mean natural logarithm of the conditional probability that two

sequences similar for m points within the tolerance remain

similar at the next point. The embedding dimension (m) and

tolerance distance (r) were set to m = 2 and r = 0.20 of the

standard deviation (�SD) of the data sequence, as suggested by

Pincus (1991).

For a better understanding of the use of this index in the

analysis of the EMG signal, a number of studies have compared

entropy values between subjects with and without illness/

dysfunction37–39 and higher entropy values have been found

among in nondisabled subjects. These findings suggest that

the absence of physiological complexity is related to patholo-

gy. In other words, values near 0 indicate the presence of

illness or dysfunction and values near 2 indicate the absence

of illness or dysfunction.

2.5. Data analysis

The Shapiro–Wilk test demonstrated that the data were

normally distributed. Data on the EMG readings obtained

Table 1 – Within-day (test and retest) and between-day (test–r

Test R

Mean (SD) ICC (2,1) SEM Mean (SD)

RMSMCE (mV)

RT 31.89 (19.13) 0.95 4.27 31.85 (17.55)

RM 24.43 (24.12) 0.94 5.91 21.85 (15.70)

LT 18.53 (16.77) 0.98 2.37 22.86 (12.03)

LM 28.85 (27.32) 0.98 3.86 36.41(21.95)

MDF (Hz)

RT 140.01 (27.37) 0.97 4.74 141.99 (21.59)

RM 132.51 (27.11) 0.93 7.17 140.08 (24.99)

LT 144.34 (31.08) 0.97 5.38 143.95 (29.25)

LM 125.20 (16.46) 0.88 5.70 130.62 (24.47)

MNF (Hz)

RT 152.4 (25.08) 0.97 4.34 153.75 (20.52)

RM 144.53 (26.16) 0.97 4.53 152.75 (24.62)

LT 157.04 (28.01) 0.97 4.85 156.46 (26.84)

LM 136.39 (17.4) 0.91 5.22 141.68 (25.20)

ZC (times/s)

RT 1308.71 (185.50) 0.94 45.43 1305.08 (175.21)

RM 1266.22 (197.68) 0.94 48.42 1332.37 (211.27)

LT 1351.46 (201.27) 0.93 53.25 1342.00 (225.84)

LM 1192.51 (140.31) 0.91 42.09 1229.51 (214.58)

ApEn

RT 1.03 (0.08) 0.89 0.02 1.02 (0.09)

RM 1.01 (0.12) 0.93 0.03 1.06 (0.12)

LT 1.06 (0.08) 0.81 0.03 1.05 (0.12)

LM 0.96 (0.09) 0.92 0.02 0.97 (0.13)

ICC, intraclass correlation coefficient; SEM, standard error of mean; R

median frequency; MNF, mean frequency; ZC, zero crossing; ApEn, app

temporal; LM, left masseter.

in two sessions (test–retest) were expressed as mean and

standard deviation (SD). The reliability of each EMG

measure was quantified using intraclass correlation coeffi-

cients (ICCs) and the standard error of the mean (SEM). ICCs

were calculated using the ICC2,1 model40 and SEM was

estimated by subtracting the ICC value from 1, taking the

square root of this value and multiplying by the SD

(SEM ¼ SDffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1 � ICCp

) 41. For the purposes of the present

study, ICCs were interpreted using the following criteria:

0.00–0.39 = poor; 0.40 to 0.59 = fair; 0.60–0.74 = good; and

0.75–1.00 = excellent.42. The SEM was used to express

reliability in absolute values, with higher values indicating

a high level of error and implying non-reproducibility

of the tested values. All data were analyzed using

the Statistical Package for Social Sciences (SPSS) version

20.0.

3. Results and discussion

Fifteen individuals with spastic diparetic CP (8 males and 7

females; mean age: 26.9 � 5.8 years; body mass: 54.4 � 10.9 kg;

height: 160.0 � 9.0 cm; body mass index: 22 � 3.63 kg/m2) were

included in the study. The data were analyzed to determine

the reliability of the EMG variables in this population. Tables 1

and 2 respectively display the mean ICCs and SEM for each

EMG reading of each muscle analyzed with the mandible at

rest and during MCE.

etest) reliability of EMG signal readings during MCE.

etest Test–retest

ICC (2,1) SEM Mean (SD) ICC (2,1) SEM

0.89 5.82 31.87 (15.74) 0.65 9.31

0.96 3.14 22.98(15.84) 0.56 10.50

0.92 3.40 21.11 (13.20) 0.72 6.98

0.97 3.80 32.63 (19.36) 0.62 11.93

0.92 6.10 141.00 (21.81) 0.72 11.11

0.92 7.06 136.30 (25.08) 0.86 9.38

0.96 5.85 144.15 (28.30) 0.84 11.32

0.91 7.34 127.91 (18.84) 0.71 10.15

0.93 5.42 153.08 (19.74) 0.68 11.16

0.94 6.03 148.64 (24.34) 0.85 9.42

0.96 5.36 156.75 (25.66) 0.84 10.26

0.93 6.66 139.03 (19.63) 0.74 10.01

0.95 39.17 1306.90 (145.45) 0.61 90.83

0.94 51.75 1299.30 (193.43) 0.81 84.31

0.94 55.31 1346.73 (197.40) 0.79 90.46

0.93 56.77 1211.01 (160.05) 0.70 87.66

0.83 0.03 1.03 (0.06) 0.60 0.04

0.93 0.03 1.03 (0.11) 0.73 0.05

0.80 0.05 1.05 (0.09) 0.65 0.05

0.89 0.04 0.97 (0.10) 0.65 0.05

MSMCE, root mean square (MCE: maximum clenching effort); MDF,

roximate entropy; RT, right temporal; RM, right masseter; LT, left

Table 2 – Within-day (test and retest) and between-day (test–retest) reliability of EMG signal readings in resting position.

Test Retest Test–retest

Mean (SD) ICC (2,1) SEM Mean (SD) ICC (2,1) SEM Mean (SD) ICC (2,1) SEM

RMSRest (%MCE)

RT 17.21 (18.98) 0.81 8.27 8.44 (6.59) 0.95 1.47 12.83 (10.44) 0.35 8.41

RM 11.18 (10.77) 0.69 5.99 6.15 (4.97) 0.73 2.58 8.66 (6.41) 0.34 5.19

LT 13.59 (11.00) 0.61 6.86 7.42 (5.23) 0.94 1.28 10.50 (7.32) 0.43 5.52

LM 9.36 (7.74) 0.73 4.02 5.57 (4.82) 0.75 2.41 7.47 (5.38) 0.46 3.95

MDF (Hz)

RT 98.17 (32.04) 0.81 13.96 99.64 (27.86) 0.80 12.45 98.91 (24.68) 0.52 17.09

RM 126.09 (31.22) 0.63 18.99 118.38 (33.22) 0.80 14.85 122.23 (28.54) 0.56 18.93

LT 109.89 (35.34) 0.81 15.40 96.67 (28.8) 0.90 9.10 103.28 (27.47) 0.57 18.01

LM 108.50 (21.89) 0.65 12.95 115.89 (26.92) 0.77 12.91 112.20 (19.04) 0.39 14.87

MNF (Hz)

RT 121.32 (29.20) 0.62 18.00 120.66 (32.13) 0.84 12.85 120.99 (22.96) 0.36 18.36

RM 156.38 (35.56) 0.76 17.42 141.08 (25.59) 0.77 12.27 148.73 (25.81) 0.27 22.05

LT 137.99 (34.12) 0.70 18.68 117.81 (31.81) 0.93 8.41 127.90 (26.91) 0.45 19.95

LM 133.39 (24.81) 0.69 13.81 138.10 (23.36) 0.73 12.13 135.75 (18.18) 0.35 14.65

ZC (times/s)

RT 2663.06 (707.81) 0.69 394.09 2586.35 (456.45) 0.80 204.13 2624.71 (424.22) 0.30 354.92

RM 3454.15 (700.57) 0.71 377.26 3260.06 (397.32) 0.74 202.59 3357.11 (447.29) 0.44 334.72

LT 3006.77 (693.14) 0.66 404.16 2677.84 (565.17) 0.93 149.52 2842.31 (449.45) 0.29 378.71

LM 2892.35 (502.87) 0.69 279.98 2998.60 (442.51) 0.80 197.89 2945.47 (379,96) 0.44 284.33

ApEn

RT 1.16 (0.34) 0.68 0.19 1.22 (0.23) 0.53 0.15 1.19 (0.21) 0.29 0.17

RM 1.17 (0.26) 0.68 0.14 1.29 (0.27) 0.74 0.13 1.23 (0.18) 0.23 0.15

LT 1.09 (0.27) 0.72 0.14 1.23 (0.23) 0.67 0.13 1.16 (0.17) 0.25 0.14

LM 1.17 (0.25) 0.61 0.15 1.35 (0.14) 0.50 0.09 1.26 (0.17) 0.35 0.13

ICC, intraclass correlation coefficient; SEM, standard error of mean; RMSRest, root mean square en rest position; MDF, median frequency; MNF,

mean frequency; ZC, zero crossing; ApEn, approximate entropy; RT, right temporal; RM, right masseter; LT, left temporal; LM, left masseter.

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 81356

3.1. Intra-session reliability

During MCE, the ICCs for the three trials recorded in the two

sessions (test and retest) revealed excellent intra-session

reliability (range: 0.80–0.98) for all EMG variables in all muscles

(Table 1). The RMSMCE was the variable with the highest SEM

(range: 2.37–5.91). In the resting position (Table 2), the ICCs

revealed good to excellent reliability (range: 0.61–0.95) for

RMSrest, mean and median frequencies (MNF, MDF) and zero

crossings (ZC). However, the approximate entropy (ApEn)

analysis revealed fair to good reliability (range: 53–74). The

SEM was highest for RMSrest (range: 1.47–6.86).

3.2. Inter-session reliability

Inter-session reliability was calculated between the test and

retest sessions with three readings under each condition.

During MCE, the ICCs revealed good to excellent reliability

(range: 0.60–0.86) mean and median frequencies (MNF, MDF)

and zero crossings (ZC) and approximate entropy (ApEn). ICCs

values for RMSMCE were fair to good (range: 0.56–0.72) and the

SEM was the highest for this variable (range: 6.98–11.93). In the

resting position, the ICCs revealed poor to fair reliability

(range: 0.23–0.57) for all EMG variables studied. The aim of the

present study was to determine the reliability of linear and

nonlinear measures of EMG activity in the masticatory

muscles of individuals with CP during intra-session and

inter-session analyses. The findings reveal that several

variables are reliable and precise in the reading of EMG

signals during MCE.

Regarding the intra-session analysis, all EMG variables

exhibited good to excellent reliability at rest and during

MCE. However, lesser reliability was found regarding the

RMS, especially when the muscles were at rest. This lower

degree of reliability may be related to the manner by which

the amplitude of the EMG signal was normalized. In

individuals without cognitive impairment, the indication

is to normalize the amplitude of the signal by MCE.43

However, considerable variability was found in the EMG

readings during MCE in the present study, indicating that

the subjects had difficulty in maintaining MCE in a constant

fashion (Fig. 1) and the decision was made to normalize the

EMG signal based on the rest position, as suggested by

Soderberg and Knutson (2000).33 Therefore, the amplitude

values of the EMG signals should be interpreted with

caution. With regard to the signal captured with the

muscles at rest, the normalization process was not

responsible for the low degree of inter-session reliability

(test–retest), since low reliability was found for all EMG

variables (MDF, MNF, ZC and ApEn).

Approximate entropy (ApEn) exhibited good to excellent

intra-session and inter-session reliability during MCE. More-

over, the standard error of the mean (SEM) was low in both

analyses. In the intra-session analysis with the muscles at

rest, good reliability was found in the ApEn in the first session

(test), but weak reliability was found for the right temporalis

(RT) and left masseter (LM) muscles in the second session

(retest). Nonetheless, the SEM was low. In the inter-session

analysis, poor reliability was found for ApEn with the muscles

at rest.

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 8 1357

The variables extracted from the analysis of the frequency

of the EMG signal exhibited better inter-session reliability

(good to excellent) during MCE. The ICCs and SEM were similar

for mean and median frequencies (MNF, MDF). Zero crossings

(ZC) exhibited lower ICCs than those found for MDF and MNF,

but the values remained within the same range of reliability

and the SEM was low. The ZC rate has properties close to MDF

and MNF,44 but is highly dependent on the signal-to-noise

ratio in EMG signals and is also very sensitive to deviations in

the amplitude of the signal. Thus, the use of this variable in the

analysis of EMG signals in individuals with CP could be

unnecessary when MDF and MNF are employed.

The better reliability of the frequency domain in relation-

ship to amplitude domain during MCE may be related to the

weakness of the masticatory muscles and difficulty coordi-

nating these muscles in this group of patients.2,3 These

problems are reflected in the fluctuation of the amplitude of

the EMG signal, as demonstrated in Fig. 1. The amplitude of the

signal is related to the recruitment and the discharge rates of

active motor units.14 The results of computer simulation

studies45,46 suggest that motor unit synchronization can alter

the amplitude of the EMG signal due to the summation of

motor unit potentials at the surface of the muscle. As motor

unit synchronization reflects connections within the central

nervous system, the fluctuation in the amplitude do EMG

signal in the present sample may stem from abnormalities in

the synchronization, recruitment and discharge rates of the

active motor units caused by CP. Thus, the variation in the

amplitude of the EMG signal stemming from these abnormali-

ties and the method employed for the normalization of the

amplitude of the signal may explain the lower degree of

reliability regarding the RMS.

In relation to the mean and median frequencies (MNF,

MDF), it has been suggested that shifts in these indices reflect

the recruitment of progressively larger and faster motor

units.47,48 Thus, the 3-s range used to analyze the EMG signal

in MCE may not have been sufficient for possible alterations in

the order of motor unit recruitment to be directly reflected in

the MDF and MNF, which may explain the greater reliability of

these variables.

Studies involving healthy individuals report that the

masseter and temporal muscles exhibit a high degree of

reliability (determined by the intraclass correlation coeffi-

cient),32,49 as found in the present study for individuals with

CP. However, the studies cited did not determine reproduc-

ibility with the muscles at rest, which hinders comparisons

with the present findings under this condition.

4. Conclusions

The present findings demonstrate that mean frequency,

median frequency and approximate entropy are the most

reliable variables of surface electromyography signals of the

masseter and temporal muscles during maximum clenching

effort in individuals with cerebral palsy. These measures can

be used to evaluate the function and behaviour of the

masticatory muscles in this population following oral reha-

bilitation and surgical oral procedures as well as for the study

the physiology of these muscles. However, the ICCs of the EMG

data recorded in the resting position demonstrate poor

reliability in the inter-session analysis of the following EMG

variables: RMS, MDF, MNF, ZC and ApEn.

Funding

None.

Competing interests

The authors declare that there are no conflicts of interest.

Ethical approval

Not required.

Acknowledgement

This project was supported by the Brazilian Coordination

for the Improvement of Higher Education Personnel/Postdoc-

toral National Program (CAPES/PNPD) – grant number 02495/

09-0.

r e f e r e n c e s

1. Vitti M, Basmajian JV. Integrated actions of masticatorymuscles: simultaneous EMG from eight intramuscularelectrodes. Anat Rec 1977;187:173–89.

2. Briesemeister M, Schmidt KC, Ries LG. Changes inmasticatory muscle activity in children with cerebral palsy. JElectromyogr Kinesiol 2013;23:260–6.

3. Dos Santos MT, Nogueira Ml. Infantile reflexes and theireffects on dental caries and oral hygiene in cerebral palsyindividuals. J Oral Rehabil 2005;32:880–5.

4. Bhowate R, Dubey A. Dentofacial changes and oral healthstatus in mentally challenged children. J Indian Soc PedodPrev Dent 2005;23:71–3.

5. Ortega AO, Guimaraes AS, Ciamponi AL, Marie SK.Frequency of parafunctional oral habits in patients withcerebral palsy. J Oral Rehabil 2008;35:191–5.

6. Miamoto CB, Pereira LJ, Paiva SM, Pordeus IA, Ramos-JorgeML, Marques LS. Prevalence and risk indicators oftemporomandibular disorder signs and symptoms in apediatric population with spastic cerebral palsy. J Clin PediatrDent 2011;35:259–63.

7. Giannasi LC, Batista SRF, Matsui MY, Hardt CT, Gomes CP,Amorim JBO, et al. Effect of a hyperbolide masticationapparatus for the treatment of severe sleep bruxism in achild with cerebral palsy: long-term follow-up. J Bodyw MovTher 2014;18:62–7.

8. Castroflorio T, Falla D, Tartaglia GM, Sforza C, Deregibus A.Myoelectric manifestations of jaw elevator muscle fatigueand recovery in healthy and TMD subjects. J Oral Rehabil2012;39:648–58.

9. Piancino MG, Isola G, Merlo A, Dalessandri D, Debernardi C,Bracco P. Chewing pattern and muscular activation in openbite patients. J Electromyogr Kinesiol 2012;22:273–9.

10. Sochaniwskyj AE, Koheil RM, Bablich K, Milner M, KennyDJ. Oral motor functioning, frequency of swallowing and

a r c h i v e s o f o r a l b i o l o g y 5 9 ( 2 0 1 4 ) 1 3 5 2 – 1 3 5 81358

drooling in normal children and in children with cerebralpalsy. Arch Phys Med Rehabil 1986;67:866–74.

11. Monteiro W, Francisco de Oliveira Dantas da Gama T, dosSantos RM, Collange Grecco LA, Pasini Neto H, Oliveira CS.Effectiveness of global postural reeducation in thetreatment of temporomandibular disorder: case report. JBodyw Mov Ther 2013;17:53–8.

12. De Rossi M, Santos CM, Miglioranca R, Regalo SC. All onfour1 fixed implant support rehabilitation: a masticatoryfunction study. Clin Implant Dent Relat Res 2013. http://dx.doi.org/10.1111/cid.12031.

13. Bar-On L, Aertbelien E, Molenaers G, Van Campenhout A,Vandendoorent B, Nieuwenhuys A, et al. Instrumentedassessment of the effect of botulinum toxin-A in the medialhamstrings in children with cerebral palsy. Gait Posture 2013.pii:S0966-6362(13)00252-X.

14. Farina D, Merletti R, Enoka RM. The extraction of neuralstrategies from the surface EMG. J Appl Physiol 2004;96:1486–95.

15. Roy SH, De Luca CJ, Schneider J. Effects of electrode locationon myoelectric conduction velocity and median frequencyestimates. J Appl Physiol 1986;61:1510–7.

16. Jensen C, Vasseljen O, Westgaard RH. The influence ofelectrode position on bipolar surface electromyogramrecordings of the upper trapezius muscle. Eur J Appl Physiol1993;67:266–73.

17. Lei M, Wang ZZ, Feng ZJ. Detection nonlinearity of actionsurface EMG signal. Phys Lett A 2001;290:297–303.

18. Rodrick D, Karwowski W. Nonlinear dynamical behavior ofsurface electromyographical signals of biceps muscle undertwo simulated static work postures. Nonlinear DynamicsPsychol Life Sci 2006;10:21–35.

19. Swie YW, Sakamoto K, Shimizu Y. Chaotic analysis ofelectromyography signal at low back and lower limbmuscles during forward bending posture. Electromyogr ClinNeurophysiol 2000;545:329–42.

20. Felici F, Rosponi A, Sbriccoli P, Filligoi GC, Fattorini L,Marchetti M. Linear and non-linear analysis of surfaceelectromyograms in weight-lifters. Eur J Appl Physiol2001;84:337–42.

21. Fattorini L, Felici F, Filligoi GC, Traballesi M, Farina D.Influence of high motor unit synchronization levels on non-linear and spectral variables of the surface EMG. J NeurosciMethods 2005;143:133–9.

22. Harbourne RT, Stergiou N. Movement variability and the useof nonlinear tools: principles to guide physical therapistpractice. Phys Ther 2009;89:267–82. http://dx.doi.org/10.2522/ptj.20080130. [review].

23. Costa M, Goldberger AL, Peng CK. Multiscale entropy todistinguish physiologic and synthetic RR time series. ComputCardiol 2000;29:137–40.

24. Costa M, Goldberger AL, Peng CK. Multiscale entropyanalysis: a new measure of complexity loss in heart failure. JElectrocardiol 2003;36(Suppl.):39–40.

25. Goldberger AL, Peng CK, Lipsitz LA. What is physiologiccomplexity and how does it change with aging and disease?Neurobiol Aging 2002;23:23–6.

26. Sung PS, Zurcher U, Kaufman M. Reliability differencebetween spectral and entropic measures of erector spinaemuscle fatigability. J Electromyogr Kinesiol 2010;20:25–30.

27. Chen W, Wang Z, Ren X. Characterization of surface EMGsignals using improved approximate entropy. J Zhejiang UnivSci 2006;B 7:844–8.

28. Mansur PH, Cury LK, Leite JO, Pereira AA, Penha-Silva N,Andrade AO. The approximate entropy of theelectromyographic signals of tremor correlates with theosmotic fragility of human erythrocytes. Biomed Eng Online2010;22:9–29.

29. Smith BA, Teulier C, Sansom J, Stergiou N, Ulrich BD.Approximate entropy values demonstrate impaired

neuromotor control of spontaneous leg activity in infantswith myelomeningocele. Pediatr Phys Ther 2011;23:241–7.

30. Buxbaum J, Mylinski N, Parente FR. Surface EMG reliabilityusing spectral analysis. J Oral Rehabil 1996;23:77I–5I.

31. Castroflorio T, Icardi K, Torsello F, Deregibus A, DebernardiC, Bracco P. Reproducibility of surface EMG in the humanmasseter and anterior temporalis muscle areas. J CranioPract 2005;23:130–7.

32. De Felıcio CM, Sidequersky FV, Tartaglia GM, Sforza C.Electromyographic standardized indices in healthyBrazilian young adults and data reproducibility. J OralRehabil 2009;36:558–77.

33. Soderberg GL, Knutson LM. A guide for use andinterpretation on kinesiologic electromyographic data. PhysTher 2000;80:485–98.

34. Inbar GF, Allin J, Paiss O, Kranz H. Monitoring surface EMGspectral changes by the zero crossing rate. Med Biol EngComput 1986;24:10–8.

35. Pincus SM. Approximate entropy as a measure of systemcomplexity. Proc Natl Acad Sci USA 1991;88:2297–301.

36. Pincus SM, Goldberger AL. Physiological time-seriesanalysis: what does regularity quantify? Am J Physiol1994;266:H1643–56.

37. Vaillancourt DE, Newell KM. The dynamics of resting andpostural tremor in Parkinson’s disease. Clin Neurophysiol2000:1112046–0.

38. Dragomir A, Akay Y, Curran AK, Akay M. Investigating thecomplexity of respiratory patterns during the laryngealchemoreflex. J NeuroEng Rehabil 2008;20:5–17.

39. Sethi A, Davis S, McGuirk T, Patterson TS, Richards LG.Effect of intense functional task training upon temporalstructure of variability of upper extremity post stroke. J HandTher 2013;26:132–7.

40. Weir JP. Quantifying test–retest reliability using theintraclass correlation coefficient and the SEM. J StrengthCond Res 2005;19:231–40.

41. Bruton A, Conway JH, Holgate ST. Reliability: what is it andhow is it measured? Physiotherapy 2000;86:94–9.

42. Cicchetti DV, Sparrow SS. Developing criteria forestablishing inter-rater reliability of specific items:application to assessment of adaptive behavior. Am J MentDefic 1981;86:127–37.

43. Burden A. How should we normalize electromyogramsobtained from healthy participants? What we have learnedfrom over 25 years of research. J Electromyogr Kinesiol2010;20:1023–35.

44. Hagg GM, Suurkula J. Zero crossing rate of electromyogramsduring occupational work and endurance tests as predictorsfor work related myalgia in the shoulder/neck region. Eur JAppl Physiol Occup Physiol 1991;62:436–44.

45. Yao W, Fuglevand RJ, Enoka RM. Motor-unitsynchronization increases emg amplitude and decreasesforce steadiness of simulated contractions. J Neurophysiol2000;83:441–52.

46. Zhou P, Rymer WZ. Factors governing the form of therelation between muscle force and the EMG: a simulationstudy. J Neurophysiol 2004;92:2878–86.

47. Bernardi M, Felici F, Marchetti M, Montellanico F, PiacentiniMF, Solomonow M. Force generation performance andmotor unit recruitment strategy in muscles of contralaterallimbs. J Electromyogr Kinesiol 1999;9:121–30.

48. Bernardi M, Solomonow M, Nguyen G, Smith A, Baratta R.Motor unit recruitment strategies changes with skillacquisition. Eur J Appl Physiol 1996;74:52–9.

49. Gonzalez Y, Iwasaki LR, McCall Jr WD, Ohrbach R, Lozier E,Nickel JC. Reliability of electromyographic activity vs. bite-force from human masticatory muscles. Eur J Oral Sci2011;119:219–24.