Surface EMG of jaw elevator muscles: effect of electrode
location and inter-electrode distance
T. CASTROFLORIO*, D. FARINA
†
,A.BOTTIN
†
,M.G.PIANCINO*,P.BRACCO*&
R. MERLETTI
†
*Specialization School of Orthodontics, Department of Biomedical Sciences and Human Oncology, Universita`di
Torino, Torino, Italy and
†
Centro di Bioingegneria, Dip. di Elettronica, Politecnico di Torino, Torino, Italy
SUMMARY This study addresses methodological
issues on surface electromyographic (EMG) signal
recording from jaw elevator muscles. The aims were
(i) to investigate the sensitivity to electrode dis-
placements of amplitude and spectral surface EMG
variables, (ii) to analyse if this sensitivity is affected
by the inter-electrode distance of the bipolar record-
ing, and (iii) to investigate the effect of inter-
electrode distance on the estimated amplitude and
spectral EMG variables. The superficial masseter and
anterior temporalis muscles of 13 subjects were
investigated by means of a linear electrode array.
The percentage difference in EMG variable estimates
from signals detected at different locations over the
muscle was larger than 100% of the estimated value.
Increasing the inter-electrode distance resulted in a
significant reduction of the estimation variability
because of electrode displacement. A criterion for
electrode placement selection is suggested, with
which the sensitivity of EMG variables to small
electrode displacements was of the order of 2% for
spectral and 6% for amplitude variables. Finally,
spectral and, in particular, amplitude EMG variables
were very sensitive to inter-electrode distance,
which thus should be fixed when subjects or mus-
cles are compared in the same or different experi-
mental conditions.
KEYWORDS: surface electromyography, linear elec-
trode arrays, electrode location, jaw elevator mus-
cles
Accepted for publication 1 July 2004
Introduction
Surface electromyography (EMG) provides non-inva-
sively information on muscle properties. Modelling
studies indicate that electrode location over the muscle
has a large influence on the characteristics of the
recorded EMG signals (1–4). For example, surface EMG
features are sensitive to small electrode displacements if
the detection point is close to the innervation zones or
tendons (5–9). Thus, comparison of results obtained in
different days with replacement of the electrodes may
be critical. These issues have been investigated in
modelling studies(5, 6) and experimental works (7–9),
especially for muscles with simple architecture.
Many surface EMG studies on the masticatory mus-
cles were performed in the past [see, for example Refs
(10–14)]. However, for these muscles, there have been
few works addressing methodological issues of the
surface EMG detection (15, 16). In most studies, the
electrodes were placed on the muscle belly without
indications of specific points along the muscle length
and the inter-electrode distance was usually different
among different studies.
For jaw elevator muscles, there is a lack in the
literature on quantitative indications on the sensitivity
of variables extracted from the surface EMG on the type
of recording (location of the electrodes, inter-electrode
distance). These issues are important for the repeata-
bility of the results and the possibility of comparing data
from different studies.
In this work we apply the multi-channel surface
EMG technique for the investigation of the myoelec-
ª 2005 Blackwell Publishing Ltd
411
Journal of Oral Rehabilitation 2005 32; 411–417
tric activity of jaw elevator muscles. Placing many
electrodes over the muscles, instead of the classic
single bipolar recordings, allows to assess the myo-
electric activity at different muscle locations. The
study is mainly methodological with direct implica-
tions on the application of the surface EMG approach
for muscle characterization in both basic research and
clinical studies. The work is performed on the
superficial masseter and the anterior temporalis mus-
cles. The aims of the work were 1) to investigate the
sensitivity of amplitude and spectral surface
EMG variables to small electrode displacements, 2)
to analyse if this sensitivity is affected by the inter-
electrode distance of the bipolar recording, and 3) to
investigate the effect of inter-electrode distance
on the estimated amplitude and spectral EMG
variables.
Materials and methods
Subjects
A sample of 13 healthy subjects [nine males, four
females, age, mean SD, 27Æ3 2Æ 7 years, overjet
3 1 mm, overbite 3 1, SpP^GoGn 22 5 – mes-
omorphic craniofacial typology according to Bracco
et al. (17, 18)], was selected from a group of 38 students
and voluntary assistants of the Orthodontics School of
the University of Torino, Torino, Italy. Inclusion criteria
were 1) no signs or symptoms of temporomandibular
disorders, 2) no cross-bite, 3) no orthodontic treatment,
4) no prosthetic rehabilitation, and 5) no missing teeth
(with the exception of the third molars). All subjects
participated to the experiment after giving written
informed consent. The study was approved by the local
Ethics Committee.
Experimental protocol
The subject sat on a comfortable chair without head
support, with the trunk in an erect posture and natural
head position. The eyes were fixed on a target in front
of him/her, 1 m away. The muscles investigated were
the masseter, superficial bundle, and anterior tempo-
ralis of the right and left side. For each muscle, the
subject was asked to produce six maximum voluntary
clenching in the intercuspal position sustained for 10 s.
A rest period of 2 min was given between each
clenching.
Surface EMG signal recording
Surface EMG signals were detected by a linear array
of 16 electrodes (19, 20) (silver point electrodes,
1 mm diameter, 2Æ5 mm inter-electrode distance be-
tween centres). The superficial bundle of the masseter
was investigated with the array placed on the man-
dibular angle – cantus straight line, with the last
electrode corresponding to the mandibular angle
(Fig. 1). The gonion–cantus distance was
100Æ76 8Æ86 mm. The electrode locations over the
muscle will be specified in the following as the
distance from the mandibular angle, normalized with
respect to the mandibular angle – cantus distance
(and reported in percentage).
Two reference lines were considered for the non-
invasive assessment of the anterior temporalis muscle.
The first was the straight line passing between the
mandibular angle and condylar head, rotated forward
with an inclination of 20. The second was the line
tangent to the auricle ear and passing through the
cantus. The array was placed along the first line with
the last electrode on the crossing between the two
reference lines (Fig. 1). The electrode location over the
anterior temporalis will be indicated in the following as
the distance from the crossing of the two selected lines
(in absolute and not percentage values).
The array was fixed on the skin with adhesive tape.
The portion of each muscle covered by the surface
electrodes was 37Æ5 mm (16 electrodes with 2Æ5mm
inter-electrode distance). The skin was cleaned and
slightly abraded with abrasive paste (Every, MEDITEC,
Parma, Italy) before electrode placement.
The signals were detected in single differential mode
to minimise power line interference, amplified
(16-channel surface EMG amplifier, EMG 16, LISiN;
Prima Biomedical & Sport, Treviso, Italy), filtered (3 dB
bandwidth, 10–500 Hz), sampled at 2048 Hz, and
converted to a numerical format using a 12-bit A/D
converter. The single differential detection was per-
formed between adjacent electrodes resulting in an
inter-electrode distance of 2Æ5 mm.
EMG signal analysis
By summation of consecutive single differential signals
detected with the array, it was possible to derive bipolar
recordings as acquired with inter-electrode distances
multiple of 2Æ 5 mm [for details refer to Farina et al.
T. CASTROFLORIO et al.412
ª 2005 Blackwell Publishing Ltd, Journal of Oral Rehabilitation 32; 411–417
(21)]. Thus, signals detected with inter-electrode dis-
tance in the range 2Æ5–20 mm (2Æ5 mm increments)
were obtained and analysed (eight inter-electrode
distances in total).
The average rectified value (ARV) and mean power
spectral frequency (MNF) were calculated from the
recorded signals for the eight inter-electrode distances
and for all electrode locations (22). These variables
were computed from non-overlapping, adjacent signal
epochs of 1 s, resulting in 10 estimates for the 10-s long
maximal contractions performed. The regression line of
these estimates was computed and its initial value (at
time t ¼ 0) was used for further analysis.
The sensitivity to electrode displacements was
assessed by two indexes of variability of EMG estimates
with electrode location (21). The total variability D
tot
was defined, for each EMG variable, as the difference
between the highest and the lowest variable value
estimated over all electrode locations and normalized
with respect to the mean value of the variable over the
same locations. D
tot
was computed for the eight inter-
electrode distances. This index indicates the largest
difference (as percentage of the mean) which can occur
between estimates of an EMG variable when the
electrodes are placed without specific criteria on the
muscle region covered by the array.
The minimum variability D
min
was defined, for each
EMG variable, as the minimal percentage difference
among all the percentage differences in EMG variable
estimates from adjacent electrode locations. Thus, it
indicates the difference which may occur in EMG
variable estimates when the recording system is placed
in the location with minimal sensitivity to small
electrode displacements with a possible error of
2Æ5 mm (the original inter-electrode distance). The
location resulting in D
min
for 2Æ5 mm inter-electrode
distance was considered as the optimal one.
Statistical analysis
The data were analysed using three-way repeated
measurements analysis of variance (
ANOVA
), followed
by post hoc Student–Newman–Keuls (SNK) pair-wise
comparison, when required. Statistical significance was
set to P <0Æ05. Data are reported as mean and standard
deviation (s.d.) or mean and standard error of the mean
(s.e.), as specified in the text.
Results
Maximal variability of surface EMG variables in the muscle
portion investigated
Figure 2a shows D
tot
of ARV and MNF as a function of
the inter-electrode distance for the two muscles inves-
tigated. ARV showed higher sensitivity to location than
Innervation zone
4 A.U.
1 A.U.
10 ms
8 A.U.
1
2
Auricle ear-cantus line
Gonion-cantus line
Gonion-condylar head line
Gonion-condylar head line
rotated forward (20 degree)
Surface EMG array
location
Masseter-superficial bundle
Temporalis anterior
1
2
Array upper part
Array upper part
Array upper part
2·5 mm interelectrode distance
10
mm interelectrode distance
15
mm interelectrode distance
(a) (b)
Fig. 1. (a) Schematic representation
of the anatomical lines used for ele-
ctrode location and examples of sur-
face multi-channel EMG signals
detected from the anterior temporalis
muscle of one subject (right side).
(b) Representative signals detected
from the anterior temporalis muscle.
The signals resulting from three
inter-electrode distances are shown.
Note the different amplitude scale in
the three cases. The location of the
innervation zones of a few potentials
is indicated for the recordings with
2Æ5-mm inter-electrode distance.
A.U. stands for arbitrary units.
SURFACEEMGDETECTIONINJAW ELEVATOR MUSCLES 413
ª 2005 Blackwell Publishing Ltd, Journal of Oral Rehabilitation 32; 411–417
MNF; D
tot
reached values higher than 100% for small
inter-electrode distances and values near 40% for
frequently used inter-electrode distances.
A three-way (factors: muscle, side, inter-electrode
distance)
ANOVA
of MNF D
tot
revealed significant
dependency on the muscle (F ¼ 9Æ21, P <0Æ01) and
inter-electrode distance (F ¼ 117Æ06, P > 0Æ001). The
post hoc SNK test disclosed pair-wise differences
(P <0Æ01) between the temporalis and masseter mus-
cles (Fig. 2a). Moreover, the post hoc SNK test disclosed
differences between all the inter-electrode distances
(P <0Æ05).
The D
tot
of ARV also depended (three-way
ANOVA
with factors muscle, side, and inter-electrode distance)
on the muscle and inter-electrode distance (F ¼ 3Æ73,
P <0Æ05; F ¼ 130Æ16, P > 0Æ001, respectively), with
the temporalis muscle showing significantly larger
values of D
tot
than the masseter muscle (SNK test,
P <0Æ05). ARV D
tot
was different (pair-wise compari-
sons by SNK test) for all the inter-electrode distances
(P <0Æ05), except between 17Æ5 mm and 20 mm.
Minimal variability of EMG variables to electrode
displacements
Figure 2b reports D
min
of ARV and MNF. As for
D
tot
,D
min
is smaller for the frequency than for the
amplitude variable. D
min
depends on inter-electrode
distance but to a less extent than D
tot
.
A three-way (factors: muscle, side, and inter-electrode
distance)
ANOVA
of MNF D
min
was not significant for any
of the factors considered (Fig. 3). On the contrary, ARV
D
min
depended (three-way
ANOVA
on factors the muscle,
side, and inter-electrode distance) on the inter-electrode
distance (F ¼ 8Æ 93, P > 0Æ001), with 2Æ5 mm inter-
electrode distance resulting in larger D
min
than the other
inter-electrode distances (SNK test, P <0Æ001).
Optimal electrode location
With the distances defined in the ‘Materials and
methods’ section, the optimal electrode location was
23Æ1 7Æ7% of the mandibular angle–cantus distance,
for the superficial masseter, and 24Æ4 6Æ6 mm, for
the anterior temporalis.
EMG variables and inter-electrode distance
Average rectified value and MNF were computed for
the eight inter-electrode distances in the location
resulting in D
min
. Figure 3 reports the results. A three-
way (factors: muscle, side, and inter-electrode distance)
ANOVA
of MNF revealed a dependency on the inter-
electrode distance (F ¼ 9Æ02, P > 0Æ001). The post hoc
SNK test disclosed pair-wise differences (P <0Æ05)
between MNF computed with inter-electrode distances
which differed by at least 10 mm.
A three-way (factors: muscle, side, and inter-elec-
trode distance)
ANOVA
of ARV revealed a dependency
on the muscle (F ¼ 12Æ97, P <0Æ001) and inter-elec-
trode distance (F ¼ 88Æ34, P > 0Æ001). The anterior
Interelectrode distance (mm)
∆tot (%, mean ± SE)∆min (%, mean ± SE)
(a)
(b)
160
2 4 6 8 10 12 14 16 18 20
MNF,temporalis
MNF, sup. masseter
ARV,temporalis
ARV, sup. masseter
22
140
120
100
80
60
40
20
0
0
9
8
7
6
5
4
3
2
222 4 6 8 10 12 14
16
18
ARV
MNF
200
1
Interelectrode distance (mm)
Fig. 2. (a) ARV and MNF D
tot
as a function of inter-electrode
distance for the two muscles investigated. Both sides are pooled
together as the side is not a factor influencing D
tot
(see text for the
statistical analysis). b) ARV and MNF D
min
as a function of the
inter-electrode distance. The results from the two muscles and
both sides are pooled together as neither the muscle nor the side
are factors significantly influencing D
min
(see text for the statistical
analysis). MNF D
min
is not influenced by inter-electrode distance
but it is reported as a function of the inter-electrode distance for
comparison with ARV.
T. CASTROFLORIO et al.414
ª 2005 Blackwell Publishing Ltd, Journal of Oral Rehabilitation 32; 411–417
temporalis resulted in larger EMG amplitudes (Fig. 3)
than the masseter muscle (SNK test, P <0Æ001). ARV
values corresponding to different inter-electrode dis-
tances were significantly different (P <0Æ05).
Discussion
In this study we analysed the sensitivity of EMG
variables in relation to electrode placement and inter-
electrode distance. Classic amplitude and spectral EMG
variables have been considered as they have been used
in many previous surface EMG studies for the assess-
ment of jaw elevator muscles [see, for example Refs
(14, 22–29)]. The fibre length of the superficial mass-
eter and anterior temporalis is approximately 25 mm
(30), thus the end-plate and end-of-fiber components
(2) constitute a major part of these signals. The
sensitivity of EMG variables to electrode location was
high when all the possible locations in the muscle
portion investigated were compared (Fig. 2). Increasing
the inter-electrode distance, the variability with elec-
trode location significantly decreased, thus inter-elec-
trode distances of at least 10–15 mm are suggested if the
electrodes are placed without reference to optimisation
criteria. Even with large inter-electrode distances,
however, differences in EMG variable estimates from
close locations over the muscle may be as large as
20–30%. These values should correspond to the min-
imal differences which can be considered reliable when
comparing subject groups or sets of measurements
without using specific criteria for electrode placements.
The sensitivity to electrode location found in this study
on the jaw elevator muscles is similar to that observed
in a previous study on the upper trapezius muscle
(note, however, that in the latter case a larger skin area
was covered) (21).
The variability of the estimates with electrode loca-
tion probably also depends on the size of the electrodes.
In this study we used small electrodes (1 mm diameter)
in order to place them at a distance of 2Æ5 mm, which
allowed to compare results from a number of inter-
electrode distances. As the increase of electrode size
may increase the detection volume, it is expected that
the variability of the estimates reported in this study are
larger than those obtained with larger electrodes.
However, the dependency of the variability on inter-
electrode distance is not affected by this factor, thus the
conclusions drawn in this study can be considered
general. In addition, variability of amplitude estimates
may be further decreased by normalization procedures
(such as expressing the signal amplitude as a percentage
of the amplitude recorded during a maximal voluntary
contraction). While this approach may reduce the effect
of some of the factors affecting absolute amplitude
values, it may have other drawbacks, such as the poor
reproducibility of the normalization contraction.
The minimal variability to electrode displacement of
2Æ5 mm is the percentage difference between EMG
variables expected when a linear array with 2Æ5mm
inter-electrode distance (as that used in this study) is
applied in different experimental sessions and when the
optimal location defined in this study is used. The optimal
location found in this study for the two muscles inves-
tigated varied considerably within subjects, which was
Interelectrode distance (mm)
MNF (Hz, mean
±
SE)
(a)
0 2 4 6 8 10 12 14 16 18 20 22
148
150
152
154
156
158
160
162
164
Interelectrode distance (mm)
ARV (µV, mean
±
SE)
(b)
Superficial masseter
Temporalis
0 2 4 6 8 10 12 14 16 18 20 22
0
50
100
150
200
250
Fig. 3. (a) MNF values as a function of the inter-electrode
distance. For each inter-electrode distance, the location leading
to D
min
has been selected. MNF values for the two muscles and
both sides are pooled together as these factors do not affect the
results (see text for the statistical analysis). (b) Same as (a) for ARV
values. In this case the two muscles result in statistically different
results (see text) and, thus, have been separated in reporting the
results.
SURFACEEMGDETECTIONINJAW ELEVATOR MUSCLES 415
ª 2005 Blackwell Publishing Ltd, Journal of Oral Rehabilitation 32; 411–417