Coherence and Phase Locking Disruption in Electromyograms of

Patients with Amyotrophic Lateral Sclerosis

Mafalda Camara

, Mamede de Carvalho

, Tiago Ara

ujo

1, 2

, Hugo Gamboa

1, 2

and Carla Quint

1, 4

Faculdade de Ci

encias e Tecnologia, Universidade Nova de Lisboa, Lisbon, Portugal

PLUX - Wireless Biosignals, Lisbon, Portugal

Instituto de Medicina Molecular, Faculty of Medicine, University of Lisbon, Lisbon, Portugal

Institute of Biophysics and Biomedical Engineering, Faculty of Sciences, University of Lisbon, Lisbon, Portugal

Keywords:

Contra and Ipsilateral, Coherence, Beta-band, Phase Locking Factor (PLF), Amyotrophic Lateral Sclerosis

(ALS), Electromyography (EMG).

Abstract:

In motor neuron disease, the aim of therapy is to prevent or slow neuronal degeneration and early diagnosis

is thus essential. Hypothesising that beta-band (15−30 Hz) is a measure of pathways integrity as shown

in literature, coherence and phase locking factor (PLF) could be used as an electrophysiological indicator

of upper and lower neuron integrity in patients with amyotrophic lateral sclerosis (ALS). In this work are

applied such tools in different variable situations. Coherence and PLF analysis was computed for EMG signals

registered from 2 groups: control subjects and ALS patients. The data was recorded during instants of steady

contraction for both contra and ipsilateral acquisitions. Ipsilateral coherence and PLF was computed for one

member of each group and results present signiﬁcant differences between both groups. Contrarily, contralateral

acquisitions were performed on 6 members of each group and both coherence and PLF results present no

signiﬁcant differences. So, while control subjects present no neuronal or muscular disorders and therefore

higher synchrony and coherence for beta-band EMG signals, patients with ALS do not present synchronism

or coherence in any frequency, specially for beta-band. All results allowed to conclude that contralateral

coherence is not a good measure of corticospinal pathways integrity. However, ipsilateral acquisitions show

promising results and it is possible to afﬁrm that ipsilateral measurements may reﬂect neuronal degeneration.

For future work is suggested a deeper analysis of PLF, that appear to have potential as a quantitative test of

upper and lower neuron integrity related to ALS.

1 INTRODUCTION

Amyotrophic lateral sclerosis, one of the major neu-

rodegenerative diseases, is a progressive incurable

motor neuron disorder, fatal in all cases. Associ-

ated therapy involves slowing down or even prevent-

ing neuronal degeneration. Since the effectiveness of

this medication depends on an early diagnosis, many

efforts have been made to ﬁnd accurate indicators of

this disease. As a general rule, patients with ALS are

diagnosed when there is already extensive motor neu-

ron degeneration present, since the diagnosis is ham-

pered by the impossibility to access the corticospinal

tract.

Some motor cortex cells are capable of synchro-

nizing their discharge with local oscillations in a

15−30 Hz range of frequency, known as beta-band.

This synchronism happens during rest or steady con-

traction but is not common during movement. Despite

all the research on this subject, there is no agreement

on which function might induce such periodic activity

(Baker et al., 2006). Some oscillatory cortical activity

involves pyramidal neurones and is reﬂected in the de-

scending drive to the muscles, being distributed to ag-

onist and antagonist muscles, which can be observed

in the EMG of these muscle pairs through coherence

in beta-band (Cordivari et al., 2002). Corticomuscular

coherence measures contributions from both ascend-

ing and descending pathways, which was proved by

phase analysis, suggest that muscles led the cortical

recordings (de Carvalho, 2012). The timing of syn-

chronism is dependent on the intrinsic properties of

the inhibitory interneurons and their conduction de-

lay (Jackson et al., 2004). In both upper and down

motor neuron disease, the focus is to prevent, if pos-

sible, or slow neuronal degeneration. Based on the

163

Camara M., de Carvalho M., Araújo T., Gamboa H. and Quintão C..

Coherence and Phase Locking Disruption in Electromyograms of Patients with Amyotrophic Lateral Sclerosis.

DOI: 10.5220/0004749901630170

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2014), pages 163-170

ISBN: 978-989-758-011-6

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

hypothesize that beta-band coherence could be used

as an indicator of neuronal integrity, the aim of this

work is to obtain coherence values averaged for each

population and evaluate if signiﬁcant differences were

present between them. Therefore, ﬁnd a way of an

early diagnose for ALS, resorting to contra or ipsi-

lalateral acquisitions.

1.1 Amyotrophic Lateral Sclerosis

This pathology is one of the major neurodegener-

ative diseases, a progressive disorder that involves

widespread degeneration of the motor system neu-

rons. Regions affected undergo destruction of layer

V pyramidal neurons from the motor cortex to the an-

terior horn of the spinal cord. To diagnose such a

disease it is necessary to ﬁnd upper and lower mo-

tor neuron degeneration in multiple regions: bulbar,

cervical, thoracic and lumbar. This disorder is char-

acterized by the neurological regions affected, but

there are common features observed in all patients

such as a rapidly progressive weakness, muscle at-

rophy, muscle cramps, fasciculations, muscle spas-

ticity, difﬁculties in breathing (dysponea), difﬁculties

in swallowing (dysphagia) and difﬁculties in speak-

ing (dysarthria). Patients tend to lose their abilities

to control voluntary movements and symptoms tend

to greatly reduce their quality of life (Kiernan et al.,

2011).

1.2 Electromyography and Beta-band

Voluntary movement is associated with the presence

of rhythmic activity in motor cortex. Beta-band, com-

prising a range of frequencies from 15 to 30 Hz, ap-

pears to vary its magnitude prior and during volun-

tary movements and is associated with the attempt to

perform certain tasks (Halliday et al., 1998). In this

same frequency band, cortical activity and motor unit

ﬁring are correlated during sustained voluntary con-

tractions. Intermuscular coherence (EMG - EMG) be-

tween different muscle groups appears to detect char-

acteristics of the same rhythmic processes, suggesting

a common drive from corticospinal pathways. The

precise function and genesis of these same cortical os-

cillations still remains elusive (Marsden et al., 2000).

Motor units from muscle pairs can be modulated by a

descending 15 - 30 Hz drive. So, this frequency band

oscillations are mediated via corticospinal pathways,

originating from sensorimotor cortex and are coher-

ent with muscles (Marsden et al., 2000; Nishimura

et al., 2009). These oscillations may arise in mo-

tor systems in order to promote synchronous neuronal

ﬁring between neurons populations that are spatially

distributed but functionally related, providing means

of linking different neuronal populations (Marsden

et al., 2000). Coherent activity may represent a com-

mon element in coding activity in simultaneous active

motor centres. Cortical areas involved in the same

motor task may be coherent with each other. Active

muscles show coherence around beta-band, reﬂecting

the activity of neuronal structures involved in driving

the spinal motoneurons (Nishimura et al., 2009). Co-

herence has demonstrated that some of these oscilla-

tions are probably transmitted via pyramidal tract to

activate muscles and may induce the same rhythm on

them (Nishimura et al., 2009).

1.3 Coherence

Coherence analysis of motor unit ﬁring behaviour

can provide information about the organization of

networks responsible for driving spinal motoneurons

during task performance. It also assesses common

presynaptic inputs that synchronize motor units pop-

ulations. In the human body, different activities may

be characterized by functional activities in distinct

circuits, due to muscles discharges at a certain fre-

quency. Some of these oscillating frequencies let to

spinal motoneurones (Grosse et al., 2002). Coherence

is a measure of the linear correlation between the fre-

quencies presented in two signals, being usually com-

puted using discrete Fourier transforms. As a function

of frequency (λ), for the rectiﬁed x and y signals, auto

spectra f

x,x

(λ), f

y,y

(λ) and cross spectra f

x,y

(λ) are

calculated to assess measures of correlation (Farmer

et al., 2007; Grosse et al., 2004). Cross-correlation is

assessed by coherence function - |R

x,y

(λ)|

- deﬁned

as the squared magnitude of cross spectrum, normal-

ized by the product of the two auto spectra as show in

equation 1.

x,y

(λ)|

| f

x,y

(λ)|

| f

x,x

(λ) f

y,y

(λ)|

(1)

1.4 Phase Locking Factor

During oscillatory activity, neurons ﬁre syn-

chronously. Therefore, common target cells will

receive neural activity synchronously and so, os-

cillations play an important role for the timing of

neural activity (Klimesch et al., 2008). On the

assumption that coherence between two signals exists

for beta-band, it would be interesting to investigate if

both signals are synchronized within this frequency

range. I.e. if the phase difference between the signals

is kept constant. To evaluate such behaviour, one

can determine the phase locking factor between both

BIOSIGNALS2014-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

164

signals, using Hilbert transforms. For signals x and y,

(t) and φ

(t) represent signals phase dependency

on time, respectively, for t = 1, ..., T . PLF between

both signals is deﬁned by (Almeida et al., 2011):

≡ |

∑

t=1

i[φ

(t)−φ

(t)]

| = |he

i[φ

(t)−φ

(t)]

i| (2)

PLF assumes values from 0 to 1. 0 stands for

signals entirely asynchronous, with phases randomly

distributed; 1 stands for signals perfectly synchro-

nized and with a constant phase lag. Values between 0

and 1 represent partial synchrony. To perform the PLF

computation it is necessary to isolate the frequencies

of interest, applying a band-pass ﬁlter with a narrow

band centred on each value of frequency in study.

2 ACQUISITIONS

2.1 Subjects

Focussing on previous published results (Fisher et al.,

2012), ipsilateral measurements were performed on

1 member from the group of patients and 1 mem-

ber from the group of control. Contralateral measure-

ments required the existence of two different groups

of subjects: group of 6 patients presenting ALS and

a control group of 6 subjects. All participants from

the control group do not present any known neuronal

or muscular disease, whereas patients with ALS had

been diagnosed within less than one year. For patients

with ALS that were in a more advanced stage of the

disease, presenting more difﬁculties that limited their

own movement control, it was impossible to collect

an analysable EMG signal.

2.2 Recordings

For each ipsilateral acquisition, two signals were si-

multaneously acquired from each subject using EMG

sensors attached to a bioPlux device, as observed in

ﬁgure 1. For both right and left hand, signals were

collected using two sensors attached to ﬁrst dorsal

interosseus muscle; for both right and left forearm,

signals were collected using two sensors attached to

extensor digitorum communis muscle. Ground was

placed in ulna bone inferior extremity, where no mus-

cle activity is present.

For each contralateral measurement, two signals

were simultaneously acquired from each subject, us-

ing two EMG sensors attached to a bioPlux device.

Each sensor (one for each hand) has two connected

Figure 1: Ipsilateral acquisitions experimental setup for left

member: Bioplux research device, placement of two EMG

sensors and ground.

Figure 2: Contralateral acquisitions experimental setup:

Bioplux research device, placement of two EMG sensors

and ground.

electrodes placed in ﬁrst dorsal interosseus muscle.

Ground was placed as in ipsilateral recordings. Sur-

face electrodes placements are shown in ﬁgure 2.

The used device collects real time biosignals at a

frequency rate of 5 kHz and EMG sensors have sec-

ond order band pass ﬁlter with cutoff frequencies of

25 and 450 Hz . Data is transmitted via bluetooth to a

computer, where the signals can be saved and visual-

ized.

2.3 Acquisition Protocol

Subjects were asked to seat and place both hands on

a desk, 10 cm away from each other in a parallel po-

sition and with hand palms facing each other in 90

degrees of ﬂexion with the elbow. Subjects had to el-

evate both index ﬁngers vertically with a maximum

articular amplitude in a direction opposite to the other

ﬁngers position, hold that position for 3 seconds while

maintaining a certain force/pressure and then return

to the initial position, where it remains for 3 seconds

while relaxing as much as possible. This movement

was repeated for 5 minutes or less according to max-

imum time tolerated by the patients. The coordinated

movement was guided by a programmed sound and

CoherenceandPhaseLockingDisruptioninElectromyogramsofPatientswithAmyotrophicLateralSclerosis

165

both ﬁngers had to be as much coordinated as possi-

ble one to another. The protocol was used for both

contra and ipsilateral acquisitions.

3 SIGNAL PROCESSING

The acquired signals were processed using Python

language. Signals were ﬁltered by a third order But-

terworth band pass ﬁlter of 30−2000 Hz. In order to

extract information about coherence and PLF, inter-

vals of contraction common to both signals had to be

isolated from intervals of relaxation, since coherence

is better estimated during periods of steady contrac-

tion (Fisher et al., 2012). Signals presenting higher

amount of noise will conceal real information, inhibit-

ing to distinguished contractions from relaxation in-

tervals. This and the differences among individual

signals, does not allow to predeﬁne an onset value

common to all signals. So, instead of using a method

based on the EMG signal envelope, a method based

on statistical model was used to deﬁne contractions

onset. Initially, more than one value was assigned to

both on and offset for each contraction. In order to ob-

tain the correct on and offset for each contraction, the

excessive ones were removed when: (1) the number

of samples between the on and offset is too short or

(2) when the number of samples between the off and

following onset is too short. To guarantee that these

signals have common intervals of contractions, it is

chosen, for each contraction, the highest value from

both onsets and the lowest from both offsets.

3.1 Coherence Processing

A ﬁrst analysis was performed using a long interval

of contraction record. The signal was divided on two

equal epochs, and the coherence between them was

computed. Regarding the EMG signals, they were

full-wave rectiﬁed before any fast Fourier transform

(FFT) analysis. Sampling frequency is placed as 5

kHz, the nonequispaced parameter of the FFT (NFFT)

as 2048, and the value that dictates the dependency

between FFT windows as 1024. Deﬁning all this pa-

rameters enables to compute coherence.

Coherence is reported in two different ways. First,

to provide a visual representation of coherence de-

pendency on frequency (in the imaginary domain),

coherence mean values among intervals of contrac-

tion for a given muscle pair was performed across all

patients within a group of subjects. This allows to

obtain precise coherence values for each patient ac-

quisition based on averaging multiple independent in-

stants of data. Second, to provide an estimation of

coherence dependency on frequency across the popu-

lation of each group, mean coherence was calculated

for the same values of frequency for a given muscle

pair among all subjects within the same group. To as-

sure that the wanted band of frequency was present in

both signals of each patient, frequency spectra were

computed for all instants of contraction and then av-

eraged to present an individual spectrum for each pa-

tient.

3.2 Phase Locking Factor Processing

Since the beta-band frequencies seem to be the most

relevant for this study, PLF was calculated for speciﬁc

values of frequency f within this same band [15, 30]

Hz with a resolution of 1 Hz. This procedure was per-

formed among both control and patient groups. Each

signal was band pass ﬁltered [ f − 2, f + 2] Hz, be-

ing f the analysed frequency. Instants of contraction

where again isolated and for each, PLF between each

pair of contra and ipsilateral measurements was cal-

culated. To present a ﬁnal value for each member of

each group, PLF was averaged between all contrac-

tions within the same acquisition. This procedure, is

performed as many times as the number of the anal-

ysed frequencies. PLF was averaged among all mem-

bers within the same group to present a PLF value,

dependent on frequency choice, for a population.

4 VALIDATION

4.1 Synthetic Electromyography

For validation of the algorithms, synthetic EMG sig-

nals were used to compute coherence. To prove that

coherence exists between two signals linearly depen-

dent on each other for particular values of frequency,

sets of signals were constructed and deﬁned by the

following equation:

signal = (sin(t × 2π f ) + k)× n(t) × mod(t) (3)

where t is a sequence of integer numbers, incre-

mented by one unit, with a desirable length; t is re-

ferred in seconds by dividing the desirable length for

5000, to take into account the sampling frequency.

n(t) represents Gaussian noise (µ = 0 and σ = 1), k

is the signal envelope, f is the signal’s frequency and

the portion mod(t) represents the rest of the division

of t by 6 bigger then 3 - guaranteeing instants of sim-

ulated contraction and relaxation of 3000 ms.

Multiple signals were created, choosing t : 8, 80

or 391s and f : 10, 20 or 40 Hz. Number of sam-

ples were chosen to provide signals with 1, 13 and 65

BIOSIGNALS2014-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

166

instants of contractions to decide the more appropri-

ate time sampling for acquisitions. Frequency values

were chosen taken into account that beta-band is be-

ing studied.

For all created signals, coherence was calculated

between two of the same type signals (simulating both

right and left hand). Coherence deﬁned by equation 3

is performed for instants of contraction and then aver-

aged for the entire measurement. Coherence results

do not present any dependency on the choice of f

value, since it assumes one for that same frequency

and nearly 0 for the remaining.

On the contrary, coherence results seem to depend

on the choice of signals length. Two types of sig-

nals were performed, both deﬁned by equation 3, with

common value of 40 Hz for f but with different value

of t. First, signals with 65 instants of contraction, and

second signals with 1 instant of contraction; two sets

of signals, with different portions of n(t), were cre-

ated to test coherence between them. Results of mean

coherence for these pair of signals allow to afﬁrm that

the higher the signals length, the more accurate are

the coherence results. Since noise is randomly dis-

tributed, averaging coherence of greater amount of in-

stants of contraction, tends to reduce the SNR. There-

fore, the use of long acquisitions increases the preci-

sion of coherence.

5 RESULTS AND DISCUSSION

5.1 Coherence Analysis

5.1.1 Ipsilateral

For both simultaneously acquired signals, ipsilateral

coherence was analysed for instants of contraction us-

ing NFFT as 4096. Graphical representation, for a

patient and a control subject, is shown in ﬁgure 3.

Analysing results, it is possible to observe that the sig-

nal acquired from the patient presents coherence near

0 for all frequencies, while coherence from the con-

trol subject presents higher values for most presented

frequencies, despite the fact that beta-band is not dis-

tinguished from the remaining frequencies.

Differences are signiﬁcant since for beta-band, co-

herence mean value for the patient was 0.024 ± 0.013

and for the control subject 0.694 ± 0.040. Patient re-

sults would be expected since coherence in beta-band

is not visible for ALS. On the other hand, results from

the control group are not expected since coherence

values were higher than those observed in literature.

Differences in results may be explained by differences

in acquisition protocol, used algorithm or parameters.

(a)

(b)

Figure 3: (a) Ipsilateral coherence acquired from a patient.

(b) Ipsilateral coherence acquired from a control subject.

Delimited by the grey box, are represented the frequencies

corresponding to beta-band.

5.1.2 Contralateral

Prior to coherence analysis, the presence of beta-band

frequencies was tested, recurring to frequency spectra

analysis. Instants of muscular contraction were iso-

lated and full-wave rectiﬁed. The power spectrum

was calculated and averaged within all contractions;

these frequency spectra were performed by PSD cal-

culus. Graphical representations of these results for

patients group are shown in ﬁgure 4(a), and for con-

trol group in ﬁgure 4(b), both with NFFT as 2048.

In ﬁgure 4, results show a higher presence of beta-

band frequencies, since they appear more enhanced

than the remaining, for the majority of all results from

both hands and groups. When the frequency spectra

reveal the presence of beta frequencies, one can evalu-

ate the values of coherence within these band frequen-

cies. As shown in 5.1, ipsilateral coherence appears

to be a precise indicator of neuronal degeneration de-

velopment. Contralateral coherence is tested to check

whether results are similar to those obtained for ipsi-

lateral acquisitions. Contralateral coherence between

CoherenceandPhaseLockingDisruptioninElectromyogramsofPatientswithAmyotrophicLateralSclerosis

167

(a)

(b)

Figure 4: Representation of the power spectra. Results col-

lected from the left hand are represent by the dotted line

and results collected from the right hand are represent by

the straight line; delimited by the grey box, are represented

the frequencies corresponding to beta band. (a) Results for

the group of control and (b) for the group of patients.

both interosseous muscles, one from each side, are

shown for patients and control subjects groups in ﬁg-

ures 5(a) and 5(b), respectively.

Results from the control group are not entirely

similar to those observed in literature. Coherence val-

ues for the beta-band frequencies do not stand out

from the ones presented in the remaining frequen-

cies. There was no signiﬁcant intermuscular coher-

ence present on any of the two groups for beta-band.

Pooled coherence value for the patients group was

0.0069 ± 0.0019 and for the control group 0.0031 ±

0.0017. Coherence was present within signiﬁcance

(in the same order of magnitude found in literature),

but not higher on the beta-band than the remaining

other frequencies. The analysis from the group of pa-

tients shows a behaviour similar with the group of pa-

tients. Contralateral coherence between coactivated

muscles from left and right side is not considered sig-

niﬁcant within the beta-band. Therefore, despite the

possible presence of these frequencies (beta-band) on

(a)

(b)

Figure 5: (a) Contralateral coherence acquired from a pa-

tient. (b) Contralateral coherence acquired from a control

subject. Delimited by the grey box, are represented the fre-

quencies corresponding to beta-band.

both signals, they do not signiﬁcantly depend linearly

from each other.

5.2 PLF Analysis

5.2.1 Ipsilateral

To present a ﬁnal value for each member of each

group, PLF was averaged between all contractions

within the same acquisition, for each value of fre-

quency among beta-band. Therefore, this procedure

is performed as many times as the number of studied

frequencies. PLF was averaged among all members

within the same group to present pooled results, for

ipsilateral measurements. The frequencies of interest

and its respective PLF values are represented in ﬁgure

The results in ﬁgure 6(a) are similar to those ob-

tained in 5.1.2. PLF values do not stand out from the

remaining in beta-band and do not assume signiﬁcant

BIOSIGNALS2014-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

168

(a)

(b)

Figure 6: Schematic representation of mean PLF values as a

function of frequency, for ipsilateral acquisitions, depicted

as a straight line; standard deviation is displayed as a dotted

line. Delimited by the grey box are represented the frequen-

cies corresponding to beta-band. (a) Results from the group

of patients. (b) Results from the control group.

values for any frequency. On the other hand, observ-

ing ﬁgure 6(b), PLF assumes much higher values and

beta-band frequencies appears to stand out from the

remaining. Differences are signiﬁcant since for beta-

band, PLF mean value for the patient was 0.131 ±

0.031 and for the control subject 0.6664 ± 0.0093.

5.2.2 Contralateral

Mean PLF values for the beta band in contralateral

signals, as a function of frequency, are presented in

ﬁgure 7, for the group of patients and group of con-

trol in 7(a) and 7(b), respectively. As observed in both

graphical representations of ﬁgure 7, PLF values do

not signiﬁcantly stand out for the beta-band frequen-

cies, since for beta-band, PLF pooled value for the

patients group was 0.1169 ± 0.0068 and for the con-

trol subject 0.1169 ± 0.0057.

(a)

(b)

Figure 7: Schematic representation of mean PLF values as

a function of frequency, for contralateral acquisitions, de-

picted as a straight line; standard deviation is displayed as

a dotted line. Delimited by the grey box are represented the

frequencies corresponding to beta-band. (a) Results from

the group of patients. (b) Results from the control group.

6 CONCLUSIONS

Synthetic EMG signals with a speciﬁc known value of

frequency were used to test coherence between two

signals of the same type, to ﬁnd whether coherence

among them is present for that frequency. For two

simulated signals of instants of contraction and relax-

ation with the same frequency, mean coherence be-

tween them for instants of contraction, are close to

1 for the speciﬁc frequency and near 0 for the re-

maining. Varying signals frequency value does not

change coherence general trends. Altering these sig-

nals length, changes the number of instants of con-

traction used to present ﬁnal averaged coherence, and

the lower the number of samples, the higher will be

the inﬂuence of noise.

Both coherence and PLF analysis in ipsilateral

CoherenceandPhaseLockingDisruptioninElectromyogramsofPatientswithAmyotrophicLateralSclerosis

169

acquisitions present signiﬁcant differences between

both groups of patients, as already proved by previ-

ous publications (Fisher et al., 2012). Results from

the group of patients present no signiﬁcant differ-

ences from the results obtained for contra lateral ac-

quisitions while the group of subjects, presents higher

values of PLF and coherence for all frequencies, es-

pecially for beta-band. So, it is possible to assume

that while control subjects present no neuronal or

muscular anomalies and therefore higher synchrony

for beta-band, patients with ALS do not present syn-

chrony in any frequency, specially for beta-band. All

results allow to conclude that ipsilateral analysis is

a good measure of corticospinal pathways integrity.

Frequency spectrums were performed for contra lat-

eral acquisitions and it was demonstrated that beta-

band frequencies are present in instants of steady con-

traction. But, as proved by coherence and PLF analy-

sis beta-band frequencies in contralateral acquisitions

was not present in neither for patients or control sub-

jects. Neither differences or signiﬁcance values were

found for both control and patients. Therefore, it is to

conclude that contra lateral frequency analysis is not a

good marker for beta-band frequencies studies. Com-

paring both methods used to study frequency domain,

PLF analysis may be seen as a more practical method

since it requires smaller sections of data, compared

to coherence. On the other hand, PLF can only be

analysed for a speciﬁc value of frequency, while co-

herence is analyzed for the range of frequencies pre-

sented on data. To conclude, as further work is sug-

gested to rely on the possibility of using PLF analysis

to observe beta-band and assess about neuronal path-

ways integrity in ipsilateral acquisitions. Therefore, it

seems possible that studying PLF serves as a method

of diagnosing ALS.

REFERENCES

Almeida, M., Vig

ario, R., and Bioucas-Dias, J. (2011).

Phase locked matrix factorization. In Proc. of the EU-

SIPCO Conference.

Baker, S. N., Chiu, M., and Fetz, E. E. (2006). Afferent

encoding of central oscillations in the monkey arm.

Journal of neurophysiology, 95(6):3904–3910.

Cordivari, C., Lees, A. J., Misra, V. P., and Brown, P.

(2002). Emg–emg coherence in writer’s cramp. Move-

ment disorders, 17(5):1011–1016.

de Carvalho, M. (2012). Testing upper motor neuron func-

tion in amyotrophic lateral sclerosis: the most difﬁcult

task of neurophysiology. Brain, 135(9):2581–2582.

Farmer, S. F., Gibbs, J., Halliday, D. M., Harrison, L. M.,

James, L. M., Mayston, M. J., and Stephens, J. A.

(2007). Changes in emg coherence between long and

short thumb abductor muscles during human develop-

ment. The Journal of physiology, 579(2):389–402.

Fisher, K. M., Zaaimi, B., Williams, T. L., Baker, S. N.,

and Baker, M. R. (2012). Beta-band intermuscular

coherence: a novel biomarker of upper motor neu-

ron dysfunction in motor neuron disease. Brain,

135(9):2849–2864.

Grosse, P., Cassidy, M., Brown, P., et al. (2002). Eeg-emg,

meg-emg and emg-emg frequency analysis: physi-

ological principles and clinical applications. Clin-

ical neurophysiology: ofﬁcial journal of the In-

ternational Federation of Clinical Neurophysiology,

113(10):1523.

Grosse, P., Edwards, M., Tijssen, M., Schrag, A., Lees,

A. J., Bhatia, K., and Brown, P. (2004). Patterns of

emg–emg coherence in limb dystonia. Movement dis-

orders, 19(7):758–769.

Halliday, D. M., Conway, B. A., Farmer, S. F., and Rosen-

berg, J. R. (1998). Using electroencephalography

to study functional coupling between cortical activity

and electromyograms during voluntary contractions in

humans. Neuroscience letters, 241(1):5–8.

Jackson, A., Spinks, R., Freeman, T., Wolpert, D., and

Lemon, R. (2004). Rhythm generation in monkey mo-

tor cortex explored using pyramidal tract stimulation.

The Journal of Physiology, 541(3):685–699.

Kiernan, M. C., Vucic, S., Cheah, B. C., Turner, M. R.,

Eisen, A., Hardiman, O., Burrell, J. R., and Zoing,

M. C. (2011). Amyotrophic lateral sclerosis. The

Lancet, 377(9769):942–955.

Klimesch, W., Freunberger, R., Sauseng, P., and Gruber, W.

(2008). A short review of slow phase synchronization

and memory: evidence for control processes in differ-

ent memory systems? Brain research, 1235:31–44.

Marsden, J., Ashby, P., Limousin-Dowsey, P., Rothwell, J.,

and Brown, P. (2000). Coherence between cerebellar

thalamus, cortex and muscle in man cerebellar thala-

mus interactions. Brain, 123(7):1459–1470.

Nishimura, Y., Morichika, Y., and Isa, T. (2009). A sub-

cortical oscillatory network contributes to recovery

of hand dexterity after spinal cord injury. Brain,

132(3):709–721.

BIOSIGNALS2014-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

170