nins-14-593360 Janua y 15, 2021 Time: 14:6 # 1
ORIGINAL RESEARCH
published: 15 Janua y 2021
doi: 10.3389/ nins.2020.593360
Edi ed by:
Julia S ephen,
Mind Resea ch Ne wo k (MRN),
Uni ed S a es
Re iewed by:
Sidney G osp ê e,
Uni e si y o F anche-Com é, F ance
Dipanjan Roy,
Na ional B ain Resea ch Cen e
(NBRC), India
*Co espondence:
Ainhoa Insaus i-Delgado
ainhoa.insaus i-delgado@uni-
uebingen.de
Special y sec ion:
This a icle was submi ed o
B ain Imaging Me hods,
a sec ion o he jou nal
F on ie s in Neu oscience
Recei ed: 10 Augus 2020
Accep ed: 30 No embe 2020
Published: 15 Janua y 2021
Ci a ion:
Insaus i-Delgado A,
López-La az E, Omedes J and
Ramos-Mu guialday A (2021) In ensi y
and Dose o Neu omuscula Elec ical
S imula ion In luence Senso imo o
Co ical Exci abili y.
F on . Neu osci. 14:593360.
doi: 10.3389/ nins.2020.593360
In ensi y and Dose o Neu omuscula
Elec ical S imula ion In luence
Senso imo o Co ical Exci abili y
Ainhoa Insaus i-Delgado1,2,3*, Edua do López-La az1,4, Jason Omedes5,6 and
Ande Ramos-Mu guialday1,7
1Ins i u e o Medical Psychology and Beha io al Neu obiology, Uni e si y o Tübingen, Tübingen, Ge many, 2In e na ional
Max Planck Resea ch School (IMPRS) o Cogni i e and Sys ems Neu oscience, Tübingen, Ge many, 3IKERBASQUE,
Basque Founda ion o Science, Bilbao, Spain, 4Bi b ain, Za agoza, Spain, 5Ins i u o de In es igación en Ingenie ía
de A agón (I3A), Za agoza, Spain, 6Depa amen o de In o má ica e Ingenie ía de Sis emas (DIIS), Uni e si y o Za agoza,
Za agoza, Spain, 7Neu o echnology Labo a o y, TECNALIA, Basque Resea ch and Technology Alliance (BRTA),
Donos ia-San Sebas ián, Spain
Neu omuscula elec ical s imula ion (NMES) o he ne ous sys em has been ex ensi ely
used in neu o ehabili a ion due o i s capaci y o engage he muscle ibe s, imp o ing
muscle one, and he neu al pa hways, sending a e en olleys owa d he b ain.
Al hough di e en neu oimaging ools sugges ed he capabili y o NMES o egula e
he exci abili y o senso imo o co ex and co icospinal ci cui s, how he in ensi y
and dose o NMES can neu omodula e he b ain oscilla o y ac i i y measu ed wi h
elec oencephalog aphy (EEG) is s ill unknown o da e. We quan i ied he e ec
o NMES pa ame e s on b ain oscilla o y ac i i y o 12 heal hy pa icipan s who
unde wen s imula ion o w is ex enso s du ing es . Th ee di e en NMES in ensi ies
we e included, wo below and one abo e he indi idual mo o h eshold, ixing he
s imula ion equency o 35 Hz and he pulse wid h o 300 µs. Fi s ly, we e icien ly
emo ed s imula ion a i ac s om he EEG eco dings. Secondly, we analyzed he
e ec o ampli ude and dose on he senso imo o oscilla o y ac i i y. On he one
hand, we obse ed a signi ican NMES in ensi y-dependen modula ion o b ain ac i i y,
demons a ing he di ec e ec o a e en ecep o ec ui men . On he o he hand,
we desc ibed a signi ican NMES in ensi y-dependen dose-e ec on senso imo o
ac i i y modula ion o e ime, wi h below-mo o - h eshold in ensi ies causing co ical
inhibi ion and abo e-mo o - h eshold in ensi ies causing co ical acili a ion. Ou esul s
highligh he ele ance o in ensi y and dose o NMES, and show ha hese pa ame e s
can in luence he ec ui men o he senso imo o pa hways om he muscle o he
b ain, which should be ca e ully conside ed o he design o no el neu omodula ion
in e en ions based on NMES.
Keywo ds: neu omuscula elec ical s imula ion (NMES), elec oencephalog aphy (EEG), a e en co ical
ac i a ion, senso imo o oscilla o y hy hm, a i ac emo al
F on ie s in Neu oscience | www. on ie sin.o g 1Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 2
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
INTRODUCTION
Neu omuscula elec ical s imula ion (NMES) is an
elec ophysiological echnique ha consis s o applying elec ical
cu en s on he skin o depola ize mo o and senso y ne es
benea h he s imula ing elec odes (Be gquis e al., 2011). NMES
has been used as a neu oscien i ic ool o s udy senso imo o
neu al mechanisms om he muscles o pe iphe al senso y
ecep o s (mechano ecep o s, nocicep o s, e c.) o he spine
and b ain (Collins, 2007;Ca son and Buick, 2020). NMES has
also been u ilized as a neu o ehabili a i e ool o educe muscle
a ophy, and o imp o e muscle one and mo o unc ion in
pa ien s wi h pa alysis a e s oke (Knu son e al., 2015;Yang
e al., 2019) o spinal co d inju y (SCI) (Pa il e al., 2015).
The wo king p inciple o ehabili a i e NMES is based on
he ollowing: (1) he di ec e ec on muscle one and (2) he
ac i a ion o ecep o s and senso y axons ha send a e en
olleys o he senso imo o co ex, a e being p ocessed by
spinal ne wo ks and subco ical s uc u es.
Nume ous s udies using unc ional magne ic esonance
imaging ( MRI) and nea -in a ed spec oscopy (NIRS) ha e
in es iga ed he ac i i y p oduced by NMES in di e en b ain
s uc u es (Kampe e al., 2000;Blickens o e e al., 2009;I ime-
Nielsen e al., 2012;Schü holz e al., 2012;Weg zyk e al.,
2017). These wo ks demons a ed ha he b ain ac i i y is
p opo ionally inc eased wi h he applied s imula ion in ensi y
(Backes e al., 2000;Smi h e al., 2003;Schü holz e al.,
2012). Recen expe imen s ha e also shown ha pe iphe al
s imula ion can modula e co icospinal exci abili y, measu ed
by mo o e oked po en ials (MEP) (Chipchase e al., 2011;
Veldman e al., 2014). As he s imula ion in ensi y inc eases,
he e is a p og essi e ec ui men o mo e a e en ecep o s
(i.e., cu aneous mechano ecep o s, muscle spindles, and Golgi
endon o gans) ha modula e spinal and co ical ci cui s o a
di e en ex en (Ma iule i e al., 2008;Be gquis e al., 2011;
Golaszewski e al., 2012). I has been p o en ha he a e ence
p o ided by muscle spindles and Golgi endon o gans due
o muscle con ac ion a els h ough he spinal co d o he
soma osenso y co ex and can di ec ly p ojec o he mo o
co ex (Ca son and Buick, 2020). The e o e, he p esence o
absence o muscle con ac ion elici ed by NMES has a di ec
impac on soma osenso y co ex and, indi ec ly, on mo o
co ex exci abili y (Sasaki e al., 2017). The neu al exci abili y
depends on he modula ion o ne ous s uc u es, such as
spinal ne wo ks, in ol ed in he a e en ansmission om he
s imula ed muscle o he b ain.
Al hough he e ec o in ensi y on co ical ac i a ion and
co icospinal exci abili y has been in es iga ed, he dose o
ene gy o he s imula ion has no a ac ed much a en ion
and migh play a pi o al ole in modula ing he ac i i y o
senso y egions. The e is e idence showing ha he numbe
o pe iphe al s imula ion pulses (i.e., dose) o e ime and
he in e -pulse in e al in luence ongoing neu al oscilla ions
and ha e a neu omodula o y e ec on he soma osenso y
co ex a ec ing indi ec ly co icospinal connec i i y (Ni sche
and Paulus, 2000;Schawo onkow e al., 2018;Z enne e al.,
2018). The ne ous sys em main ains i s exci abili y wi hin
an equilib ium ange h ough adjus men s de i ed om he
his o y o neu onal ac i i y, p e en ing excessi e inhibi ion
o acili a ion (Pozo and Goda, 2010). P olonged pe iods o
s imula ion-induced exci abili y/inhibi ion ha e been shown o
ac i a e homeos a ic plas ici y mechanisms ha d i e he sys em
owa d a mo e inhibi ed/exci ed s a e, educing he e ec s o he
s imula ion (Gamboa e al., 2010;And ews e al., 2013). Fu he ,
a p og essi e pe cep ual adap a ion (o educ ion o senso y
esponsi eness) has been e idenced a e p olonged in e als
o pe iphe al ib o ac ile and elec ocu aneous s imula ion
(Leung e al., 2005;Buma e al., 2007;G aczyk e al.,
2018), indica ing a neu al compensa ion a e a pe u ba ion
o he oscilla o y neu al sys em. The e o e, i could be
concei able ha p olonged pe iods o NMES esul in changes
o neu al exci abili y, esul ing in a ebalance o he co ical
esponse along ime.
Al hough mos o he s udies so a ha e used MRI and
NIRS o ack slow b ain co ela es, o TMS-induced MEPs
o asses co icospinal exci abili y, elec oencephalog aphy (EEG)
is a powe ul neu oimaging echnique wi h high empo al
esolu ion ha has been used o s udy senso imo o p ocesses
(including senso y e oked po en ials and hy hms) (Bi baume
e al., 1990;Buzsaki, 2006;Shibasaki and Halle , 2006;Leodo i
e al., 2019). The senso imo o oscilla ions ha mainly comp ise
he olandic alpha [(7–13) Hz] and be a [(14–30) Hz] hy hms
ha e been ho oughly used o s udy co ical in ol emen du ing
senso imo o asks (Ramos-Mu guialday and Bi baume , 2015;
López-La az e al., 2018), being quan i ied as he e en -
ela ed (de)synch oniza ion (ERD/ERS) (P u schelle and Lopes
da Sil a, 1999). Fu he mo e, i has also been used as a
ea u e o neu omodula ion o senso imo o neu al ne wo k
ia p op iocep ion and hap ics (Ray e al., 2020;Sebas ián-
Romagosa e al., 2020). To da e, only a ew s udies ha e epo ed
how oscilla o y ac i i y measu ed wi h EEG is modula ed by
NMES (Vidau e e al., 2016, 2019;Tu-Chan e al., 2017;Co be
e al., 2018). Unde s anding his p ocess is o g ea impo ance
gi en he ele ance o EEG and NMES o neu o ehabili a ion,
especially o EEG-based neu al in e aces. I is impo an o no e
ha , al hough he neu al esponse can be di ec ly in luenced by
e e y s imula ion pa ame e (e.g., equency, pulse wid h, and
pulse wa e o m), we ocused on in es iga ing he in luence o
in ensi y and dose o s imula ion, since hese a e he pa ame e s
ha a e usually pe sonalized o each pa ien .
In his wo k, we acqui ed EEG ac i i y om 12 heal hy
pa icipan s du ing NMES o he w is ex enso muscles a
h ee di e en in ensi ies ( wo below and one abo e he mo o
h eshold) o in es iga e he neu omodula o y e ec o pe iphe al
s imula ion on he ongoing co ical oscilla o y ac i i y, while
he subjec s es ed in a com o able si ing posi ion. Fi s ly,
we wan ed o con i m ha as he in ensi y inc eases, he e is
a p opo ional neu al exci a ion, p obably esul ing om he
ec ui men o mo e a e en ibe s, which p o ide a g ea e
p ojec ion o a e en olleys o he senso y co ex. Secondly, we
wan ed o assess i he dose o s imula ion has an e ec on he
magni ude o neu al exci a ion o e ime, expec ing ha a e
he s imula ion has been p o ided du ing a p olonged ime, he
co ical esponsi eness is lowe .
F on ie s in Neu oscience | www. on ie sin.o g 2Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 3
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
MATERIALS AND METHODS
Pa icipan s
Twel e igh -handed heal hy pa icipan s ( ou emales,
age = 27.5 ±3.0 yea s, heigh = 176.3 ±8.6 cm,
weigh = 69.8 ±9.8 kg) we e ec ui ed o pa icipa e in he s udy.
All o hem signed an in o med consen o m. The expe imen al
p ocedu e was app o ed by he E hics Commi ee o he Facul y
o Medicine o he Uni e si y o Tübingen (Ge many).
Pa icipan s we e asked o s ay com o ably sea ed on a chai
wi h hei igh a m es ing on a side able and he hand hanging
wi h he palm acing downwa ds. Neu omuscula elec ical
s imula ion (NMES) elec odes we e placed on he igh -
hand ex enso s (mo e de ailed desc ip ion in “Neu omuscula
Elec ical S imula ion” sec ion), as desc ibed in Figu e 1A. EEG
and muscle ac i i y o wo senso s we e eco ded du ing he
expe imen . The elec ical a i ac eco ded om he muscle
senso s was used o align s imula ion onse du ing he EEG
signal p ocessing.
Expe imen al Design and P ocedu e
The main pu pose o he expe imen was o in es iga e NMES
neu omodula o y e ec s (ins an aneous and cumula i e) on
b ain oscilla o y ac i i y. Wi h his aim, we compa ed he
a e en co ical ac i i y gene a ed by h ee di e en NMES
in ensi ies. Pa icipan s we e passi ely s imula ed, meaning ha
hey we e es ing, and no oli ional mo o command was
gene a ed du ing s imula ion. Each pa icipan unde wen one
session consis ing o nine blocks, each comp ising 18 ials.
One o he h ee NMES in ensi ies was andomly assigned o
each block (de e mina ion o he cu en in ensi ies explained
in “Neu omuscula Elec ical S imula ion” sec ion), esul ing in
h ee blocks pe in ensi y. A eady cue was p esen ed 2.6–
3 s be o e he NMES in e al, which had a andom du a ion
be ween 3.4 and 3.8 s. F om he o se o he NMES o he
nex eady cue, a 3 s in e - ial pe iod was in oduced (see
Figu e 1B). Audi o y cues announced he beginning o each
in e al. The ime be ween blocks was used as b eaks, las ing
a ound 150 s (i.e., 2.5 min). The en i e session including se up
did no exceed 90 min.
Da a Acquisi ion
The elec oencephalog aphic ac i i y was eco ded wi h a
comme cial 32-channel ac iCAP sys em (B ain P oduc s GmbH,
Ge many) and a monopola B ainAmp ampli ie (B ain P oduc s
GmbH, Ge many). The eco ding elec odes we e placed a FP1,
FP2, F7, F3, Fz, F4, F8, FC3, FC1, FCz, FC2, FC4, C5, C3, C1, Cz,
C2, C4, C6, CP5, CP3, CP1, CPz, CP2, CP4, CP6, P7, P3, P4, P8,
O1, and O2, ollowing he in e na ional 10/20 sys em. G ound
and e e ence elec odes we e placed a AFz and Pz, espec i ely.
Muscle ac i i y o he igh o ea m o he pa icipan s
was eco ded by an MR-compa ible B ainAmp ampli ie
(B ain P oduc s GmbH, Ge many) using wo Ag/AgCl bipola
senso s (Myo onics-No omed, Tukwila, Wa, Uni ed S a es).
The senso s we e placed la e ally o he s imula ion pads (see
Figu e 1A), using he igh colla bone as g ound. Bo h EEG
and muscle ac i i y we e synch onously acqui ed a a sampling
a e o 1,000 Hz.
Neu omuscula Elec ical S imula ion
A p og ammable neu omuscula s imula o Bones im (Tecnalia,
Se bia) was used o deli e he s imula ion. The ca hode
(3 ×3.5 cm, sel -adhesi e elec ode) was placed o e he muscles
in ol ed in w is ex ension (ex enso digi o um and ex enso
ca pi ulna is), a one hi d o he dis ance be ween he la e al
epicondyle o he hume us and he Lis e ’s ube cle a he w is .
The anode (5 ×5 cm, sel -adhesi e elec ode) was placed
5 cm dis al o he ca hode. To ensu e he co ec loca ion
o he elec odes, indi idually de e mined o each pa icipan ,
s imula ion abo e he mo o h eshold was applied un il a
comple e w is ex ension was induced.
The equency o he NMES was se o 35 Hz, and he pulse
wid h o 300 µs (Lynch and Popo ic, 2008). The indi idual
in ensi ies o each subjec we e ob ained by a scan o cu en s,
s a ing a 1 mA and inc easing in s eps o 1 mA. The pa icipan s
we e asked o epo he ini ia ion o he ollowing sensa ions:
(i) ingling o he o ea m (i.e., senso y h eshold—STh), (ii)
wi ching o he inge s (i.e., mo o h eshold—MTh), and (iii)
comple e ex ension o he w is (i.e., unc ional h eshold—FTh).
Acco ding o hese h esholds, he h ee NMES in ensi ies ( wo
below and one abo e he mo o h eshold) we e calcula ed,
ollowing he equa ions de ined in he s udy o Smi h e al.
(2003). Low in ensi y was de ined as one hi d be ween he
STh and MTh; medium in ensi y as wo hi ds be ween he
STh and MTh; and high in ensi y as he FTh (see Eq. 1–3 and
Figu e 1C). These in ensi ies we e kep cons an h oughou he
expe imen , and he comple e ex ension o he w is induced
by unc ional s imula ion was isually e i ied. None o he
pa icipan s epo ed pain o any ha m ul e ec due o he
s imula ion.
Low in ensi y =MTh −STh×0.33 +STh (1)
Medium in ensi y =MTh −STh×0.66 +STh (2)
High in ensi y =FTh (3)
Da a P ep ocessing and Analysis
A i ac Remo al P ocedu es
One impo an limi a ion o he quan i ica ion o EEG ac i i y
du ing con inuous s imula ion is he con amina ion o he
signals due o he elec ical cu en s deli e ed o he body.
The EEG is easily pollu ed by hese cu en s, and a i ac
emo al me hodologies a e essen ial o p ope ly es ima e co ical
ac i a ion. Wi h his aim, di e en echniques o con amina ion
emo al in in asi e and non-in asi e b ain ac i i y eco dings
ha e been p oposed, such as in e pola ion, blanking, o linea
eg ession e e ence (LRR) (Wal e e al., 2012;I u a e e al.,
2018;Young e al., 2018). Blanking o he da a is he mos
es ic i e me hod as con amina ed da a a e ejec ed and signals
ha could be o in e es a e neglec ed o u he analysis.
F on ie s in Neu oscience | www. on ie sin.o g 3Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 4
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
Bones im
In ensi y
Senso y
h eshold Mo o
h eshold Func ional
h eshold
33% 66%0
Low in ensi y
Medium in ensi y
High in ensi y
S a ReadyReady
NMES
3 s2.6 ~ 3 s 0.8 s 3.4 ~ 3.8 s
AB
C
FIGURE 1 | Expe imen al design and p ocedu e. (A) Rep esen a ion o he loca ion o s imula ion elec odes and senso s o measu e muscle ac i i y placed on igh
w is ex enso s. (B) Timeline o he h ee phases included in each ial: p epa a ion, NMES, and in e - ial pe iod. (C) De e mina ion o NMES in ensi ies: low in ensi y
as one hi d be ween he senso y h eshold and mo o h eshold, medium in ensi y as wo hi ds be ween he senso y h eshold and mo o h eshold, and high
in ensi y as unc ional mo o h eshold.
Howe e , i he emo al is implemen ed using ha dwa e, he
a i ac has less in luence on he eco e y pe iod o he ampli ie ,
p e en ing i om being sa u a ed and allowing he use o o he
me hods o compensa e o he missing da a (Ken and G ill,
2012). Ano he app oach is o linea ly in e pola e he co up ed
da a, connec ing he las poin be o e he a i ac and he i s
poin a e he a i ac . Howe e , in e pola ion induces a bias
in he es ima ion o powe spec um o he signals (Wal e
e al., 2012). LRR e- e e ences he signals h ough weigh s ha
a e assigned o each channel. The weigh s a e calcula ed in a
aining block acco ding o he noise o each channel gene a ed
by he elec ical s imula ion. This me hod e ec i ely educes
a i ac s, bu i ixes he weigh s and assumes no changes in
channel noise du ing he in e en ion. So a , he easibili y
o his me hod has only been p o en in in asi e eco dings
(Young e al., 2018), in which impedances a e less likely o
change wi hin sessions and a e mo e simila among channels
(Ball e al., 2009). No mally, impedances de e io a e and noise-
in luence inc eases h oughou an EEG session, complica ing
he implemen a ion o LRR in non-in asi e eco dings o b ain
ac i i y. The e o e, we implemen ed an al e na i e wo-s ep
a i ac emo al me hod and demons a ed i s easibili y. The aw
EEG signals we e p e-p ocessed using cus om-de eloped sc ip s
in MATLAB (Ma hWo ks, Na ick, MA, Uni ed S a es).
Channel emo al based on powe -line noise
Du ing an EEG session, pa icula ly du ing se up and du ing
pe iods be ween expe imen al blocks, special ca e is equi ed
o main ain EEG signal clean (i.e., aw da a inspec ion and
impedance check). Howe e , ou empi ical expe ience
shows ha , some imes, ce ain EEG elec odes p esen
highe con amina ion due o he s imula ion han o he s
(see Figu e 2A). These elec odes p esen b oadband a i ac s
ha impede u he analyses e en a e applying he median
il e p ep ocessing desc ibed below. We hypo hesized ha his
e ec migh be due o deg aded impedances, which occasionally
de e io a e e en when du ing he se up we e se below 5 kOhm.
Despi e we did no s o e he impedances o each elec ode o
check his and disca d he elec odes wi h high impedance, we
idea ed an au oma ized me hod o de ec and disca d hem
o line. This way, we could au oma ically elimina e con amina ed
channels wi hou he human bias ha would cons i u e a
manual ejec ion.
Ha ing high impedance be ween he eco ding elec ode and
he skin esembles an open ci cui , whe e he elec ode beha es
like an an enna and cap u es ou side elec ic equencies (like he
powe -line noise). Ou me hod exploi s his e ec and iden i ies
he EEG elec odes wi h unusually high powe -line noise (50 Hz
in Eu ope). Since all he elec odes should be app oxima ely
equally exposed o elec omagne ic signals a 50 Hz, we assume
ha e y high powe a his equency is an indi ec indica o
o high skin-elec ode impedance. The p ocedu e was applied
block-wise, meaning ha i was used o de ec and emo e
con amina ed channels wi hin each indi idual EEG block. The
EEG ac i i y was high-pass il e ed a 0.1 Hz wi h a 4 h-o de
Bu e wo h. The powe spec al dis ibu ion a 48–52 Hz was
es ima ed using Welch’s me hod, a e aging he pe iodog am
o 1 s Hamming windows wi h 50% o e lapping. The powe
mean and s anda d de ia ion (SD) o all he EEG channels we e
calcula ed om all ials in each block. Channels whose powe
was highe han 4 SD abo e he mean we e disca ded om ha
speci ic block. The emaining channels we e used o e-compu e
he mean and SD. The p ocedu e was i e a i ely epea ed un il no
channels exceeded he ejec ion h eshold (see Figu e 3 dashed
box, Supplemen a y Figu es 1,2, and Supplemen a y Table 1).
Median il e ing o emo al o elec ical s imula ion
con amina ion
I is well known ha applying elec o-magne ic cu en s o
s imula e he neu al sys em can in oduce undesi ed noise o
he eco dings. The NMES con igu a ion used in his s udy
in oduces la ge peaks o sho la ency (∼5 ms) o he eco ded
EEG signal. The e o e, median il e ing was used o minimize
he NMES-induced a i ac s (Insaus i-Delgado e al., 2017). This
il e is sui ed o elimina e high-ampli ude peaks om a ime
se ies (Gallaghe and Wise, 1981) and can emo e he sho -
la ency high-ampli ude a i ac s caused by he NMES. A sliding
window o 10 ms was applied o he EEG signal in s eps o one
F on ie s in Neu oscience | www. on ie sin.o g 4Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 5
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
0123-1-2-3 44- Time (s)
C3
C1
Cz
CP3
CP1 120 µV
Con amina ion o channels wi h bad impedances
Time (s)
2.10 2.12 2.14 2.16 2.18 2.20
C1
NMES a i ac emo al by median il e
Raw signal
Median il e ed
15 µV
AB
FIGURE 2 | Cha ac e iza ion o con amina ion induced by NMES. (A) EEG o a ep esen a i e ial showing non-con amina ed channels (C3, C1, CP3, and CP1) and
one con amina ed channel (Cz) du ing high-in ensi y s imula ion. Channels wi h bad impedances a e mo e p one o be con amina ed by elec ical s imula ion.
(B) Zoom in 100-ms segmen o a ep esen a i e EEG ial in a non-con amina ed channel ha p esen s NMES a i ac s ( ed line). The e ec o s imula ion a i ac s is
minimized by median il e (g een line).
Time- equency
analysis
HP il e 0.1 Hz
Powe o EEG
channels [48-52] Hz
Compu e mean and SD
powe o e all
channels
Powe EEG
Ch > Mean
powe + 4SD ?
EEG channels wi h
good impedances
Disca d EEG
channel
YES
NO
Quan i ica ion o b ain oscilla o y ac i i y
Es ima ion o powe
in alpha (7-13) Hz
and be a (
14
-30) Hz
Non-s imula ion (-3, -1) s
Low NMES (0.5, 2.5) s
Med. NMES (0.5, 2.5) s
High NMES (0.5, 2.5) s
Raw da a
Channel emo al
Median il e
BP il e
0.1-45 Hz
T ial ex ac ion
(-4, 4) s
Common a e age
e- e e encing
Subsampling o
100 Hz
FIGURE 3 | Flowcha wi h he s eps o he quan i ica ion o b ain oscilla o y ac i i y. The whole se o da a is i s ly p ep ocessed by a wo-s ep p ocedu e based on
channel emo al and median il e ing. In his le el, channels wi h bad impedances and a i ac s due o elec ical s imula ion a e emo ed. Then, he emaining clean
da a a e il e ed and di ided in o ials. Finally, powe is es ima ed in alpha (7–13) Hz and be a (14–30) Hz bands o non-s imula ion [(−3, −1) s] and NMES [(0.5, 2.5)
s] in e als, being he baseline (−2.5, −1.5) s.
F on ie s in Neu oscience | www. on ie sin.o g 5Janua y 2021 | Volume 14 | A icle 593360

nins-14-593360 Janua y 15, 2021 Time: 14:6 # 6
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
sample, p o iding as ou pu he median alue o each window.
We selec ed a 10 ms window as i ully co e s he elec ical
a i ac . This il e p oduces a equency-dependen a enua ion
ha ollows an exponen ial unc ion om 0 o 100 Hz (i.e., he
equency wi h a pe iod ha comple ely i s wi hin he 10 ms
window), leading o low a enua ion a low equencies and a
comple e a enua ion a 100 Hz (Supplemen a y Figu e 3). Wi h
his window size, he a enua ion o he signal a 10 Hz, 20 Hz,
and 30 Hz is 1.28%, 4.89%, and 10.90% espec i ely. The ela i ely
low a enua ion a low equencies makes his me hod sui able
o analyzing alpha and be a senso imo o oscilla ions. Figu e 2B
displays a zoomed segmen o 100 ms o ac i i y, showing he
e ec o he median il e on he s imula ion a i ac s.
Quan i ica ion o B ain Oscilla o y
Ac i i y
A common a e age e e ence (CAR) was applied o he EEG
signals. The e- e e enced signals we e band-pass il e ed a 0.1–
45 Hz using a 1s -o de Bu e wo h il e . Each block was
immed down o (18×) 8-s ials, om -4 o + 4 s, being 0 he
beginning o he s imula ion. The ials we e down sampled a
100 Hz, and hose belonging o he same le el o NMES in ensi y
we e pooled oge he .
The quan i ica ion o co ical ac i i y was pe o med by
e alua ing he spec um di e ences o he senso imo o hy hms,
by means o he alpha and be a e en - ela ed (de)synch oniza ion
(ERD/ERS), i.e., dec ease o inc ease in powe gene a ed by
an e en compa ed o a baseline (P u schelle and Lopes da
Sil a, 1999). La ge ERD alues (i.e., mo e nega i e powe alues)
ep esen s onge co ical ac i a ion compa ed o baseline
ime in e al, as i ep esen s disinhibi ion/exci a ion o neu al
popula ion ac i i y (Ri e e al., 2009). Fo he quan i ica ion
o b ain ac i i y, we used he FieldT ip oolbox1 o MATLAB.
Time- equency maps we e calcula ed using Mo le wa ele s in
he equency ange om 1 o 45 Hz, wi h a esolu ion o 0.5 Hz.
The powe change was compu ed as he pe cen age o inc ease o
dec ease in powe (i.e., ERS o ERD) wi h espec o he baseline
[(−2.5, −1.5) s], whe e Pj ep esen s he signal powe a he j h
sample, as desc ibed in Eq. 4.
ERD/ERSj(%)=Pj−Baseline
Baseline ×100 (4)
F om he ime- equency maps, we calcula ed he senso imo o
a e aged changes in powe in alpha (7–13) Hz and be a (14–
30) Hz bands o non-s imula ion [(−3, −1) s] and NMES [(0.5,
2.5) s] in e als, using as baseline he (−2.5, −1.5) s in e al
(see Figu e 3). The NMES pe iod was de ined as s a ing a
0.5 s o a oid po en ial bias and in luence o he s imula ion
onse like e en - ela ed b ain po en ials (e.g., soma osenso y
e oked po en ial) a = 0. Fo he opog aphical inspec ion
and he desc ip i e analysis o he en i e b ain ac i i y, all
EEG channels we e analyzed indi idually. Fo he quan i a i e
analysis, we calcula ed he mean change in powe o channels
C1, C3, CP1, and CP3 (i.e., a ea o e he senso imo o co ex
1h p:// ield ip oolbox.o g/
ep esen ing he igh o ea m, con ala e al hemisphe e o he
s imula ed limb), since we conside ed ha a e aged alues o e
hese elec odes could be e quan i y he o e all changes in he
senso imo o a eas.
S a is ical Analysis
The s a is ical es s we e pe o med in IBM SPSS 25.0 S a is ics
so wa e (SPSS Inc., Chicago, IL, Uni ed S a es) and MATLAB.
We used he Shapi o–Wilk es o de e mine he no mali y
o he da a. Acco dingly, a mul i a ia e analysis o a iance
(MANOVA) o epea ed measu es was pe o med o ind
di e ences in he dependen a iables, alpha and be a ERD/ERS,
wi h NMES in ensi y ( ou le els: no s imula ion, low-, medium-,
and high-in ensi y s imula ion) as wi hin-subjec ac o . In o de
o de e mine he o igin o he signi ican e ec , pos hoc es s wi h
Bon e oni co ec ion we e pe o med.
In o de o analyze whe he NMES can induce a dose-
e ec , we s udied he ERD/ERS changes o e ime. Fo ha , we
compu ed he alpha and be a ERD/ERS o each single ial (i.e.,
in Eq. 4, Pjwas he alpha/be a powe o each ial du ing he
NMES pe iod, and he baseline was calcula ed om he g and
a e age o all he ials o each in ensi y). A linea eg ession
was es ima ed o he ERD/ERS alues o e ials o he wo
equency bands (i.e., alpha and be a) and he h ee NMES
in ensi ies (i.e., low, medium, and high). Co ela ion be ween
ERD/ERS and sequence o ials we e calcula ed using Pea son’s
co ela ion coe icien o s udy s imula ion e ec s o e ime.
RESULTS
E ec o A i ac Remo al
The p e-p ocessing o he da a elimina ed sa is ac o ily he
elec ical noise con amina ion coming om he pe iphe al
elec ical s imula ion. I educed he e ec o he a i ac s o an
ex en ha allowed us o pe o m EEG spec al analysis o he
b ain oscilla o y ac i i y.
Channels wi h good impedances a e also in luenced by he
elec ical s imula ion a i ac ha is in oduced in o he signal
as la ge peaks. Figu e 2B illus a es in a 100 ms segmen o
a ep esen a i e ial how he median il e deals wi h hese
undesi ed a i ac s. To p o e he e icacy o he me hod, we
ocused on he wo s -case scena io, as he con amina ion is
la ge o highe s imula ion in ensi ies. This e ec can be
obse ed in Figu e 4, which depic s he EEG ime- equency
ac i i y a he di e en NMES in ensi ies including a i ac s and
a e he median and spa ial il e s a e applied. The NMES
gene a es an inc ease o powe , o ERS, a ound 35 Hz (i.e., he
s imula ion equency), which inc eases as he NMES in ensi y is
inc emen ed (Figu e 4, le column). This powe inc ease in high-
be a/low-gamma band due o s imula ion a i ac was elimina ed
o all in ensi ies a e median il e ing. Applying he median il e
did no change he powe in alpha and be a equencies be o e
he s imula ion onse ( = 0 s), bu elimina ed he ERS du ing
he s imula ion pe iod, minimizing he a i ac s and e ealing he
alpha and be a modula ion. The common a e age e- e e ence
(CAR) a e median il e ing enhanced he powe dec ease o
F on ie s in Neu oscience | www. on ie sin.o g 6Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 7
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
Wi hou median il e Wi h median il e
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
50400102030-50 -40 -30 -20 -10
ERD/ERS (%)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
Medium
in ensi y
Low
in ensi y
High
in ensi y
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
5
15
25
35
45
0123-1-2-3 Time (s)
F equency (Hz)
Wi h median il e and CAR
FIGURE 4 | Compa ison o co ical ac i a ion a e median and spa ial il e ing. Time- equency maps a e aged o e all pa icipan s, ep esen ing ERD/ERS, o he
a e age o channels (C3, C1, CP3, and CP1) loca ed o e he con ala e al senso imo o co ex o he s imula ed limb. A e aged ime- equency maps wi hou
median il e (le column), wi h median il e (cen e column), and wi h median and CAR il e ( igh column) a e emo al o con amina ed channels. Rows show he
di e en NMES in ensi ies: high (uppe ow), medium (middle ow), and low in ensi y (lowe ow). The pe cen age o ERD/ERS is compu ed acco ding o he baseline
(−2.5, −1.5) s. Time 0 s is aligned wi h he onse o he s imula ion.
he bands o in e es o e e y NMES in ensi y, as i does wi h
non-con amina ed EEG (Wolpaw e al., 2002). Rega dless o he
in ensi y deli e ed, he ini ia ion o he s imula ion gene a ed
ime- and phase-locked ac i i y (i.e., e en - ela ed po en ials—
ERP), p esumably due o he senso y p ocessing o he NMES
(Reynolds e al., 2015). This can be seen as powe inc ease a low
equencies (1–5 Hz).
In luence o S imula ion In ensi y on
Co ical Ac i a ion
To analyze he in luence o s imula ion in ensi y on co ical
ac i a ion, we compa ed he changes in b ain oscilla o y
ac i i y in ou condi ions: non-s imula ion, low-, medium-,
and high-in ensi y s imula ion (see opoplo s o alpha and be a
hy hms in Figu e 5). Topog aphic maps o non-s imula ion
condi ion we e calcula ed using he in e al (−3, −1) s p io o
he s imula ion, while he o he condi ions we e ex ac ed om
he in e al (0.5, 2.5) s a e s imula ion onse . An inc emen
o he s imula ion in ensi y esul ed in an inc easing ERD (i.e.,
la ge dec ease in powe ) in bo h equency bands o e he
senso imo o co ex as expec ed (Backes e al., 2000;Smi h e al.,
2003;Schü holz e al., 2012), while occipi al a eas showed idling
ac i i y. A high-in ensi y s imula ion, he senso imo o co ex
o bo h hemisphe es p esen ed a dec ease o powe , being mo e
p onounced in he con ala e al hemisphe e, as demons a ed
in p e ious wo k s udying b ain oscilla o y signa u es o mo o
asks (Ramos-Mu guialday and Bi baume , 2015).
F on ie s in Neu oscience | www. on ie sin.o g 7Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 8
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
Medium in ensi y y isne nihgiHy isne niwoLNon-s imula ion
Alpha band
(7-13 Hz)
40020-40 -20
ERD/ERS (%)
*
**
0
-10
-20
ERD/ERS (%)
-30
Non-
S im. Low Med. High
0
-10
-20
ERD/ERS (%)
-30
Non-
S im. Low Med. High
*
**
Be a band
(14-30 Hz)
FIGURE 5 | Compa ison o co ical ac i a ion o di e en s imula ion in ensi ies o alpha and be a band EEG ac i i y. Topog aphic maps, a e aged o e all
pa icipan s, showing ERD/ERS o non-s imula ion pe iods (−3, −1) s and NMES pe iods (0.5, 2.5) s belonging o each in ensi y (i.e., low, medium, and high) o
alpha (uppe ow) and be a (lowe ow) equency bands. Ba g aphs show he mean pe cen age o ERD/ERS a e aged om channels (C3, C1, CP3, and CP1) o
each in ensi y and equency band. The s a is ically signi ican di e ences be ween pai s a e exp essed wi h ho izon al lines and s a s. The pe cen age o ERD/ERS is
calcula ed wi h espec o he baseline (−2.5, −1.5) s. The signals a e p ocessed using he wo-s ep p ocedu e (i.e., emo al o con amina ed channels and median
il e ) and a CAR.
Ou MANOVA analysis e lec ed a signi ican e ec
o in ensi y on alpha and be a ERD [F(6, 86) = 4.356,
p= 0.001]. Righ mos panels in Figu e 5 display he
esul s o he pos hoc compa isons. Fo bo h alpha and
be a, he e was a signi ican ly highe ERD (i.e., mo e
nega i e powe alues) induced by high-in ensi y NMES
compa ed o es (p= 0.004 o alpha, p<0.001 o be a),
o low-in ensi y NMES (p= 0.013 o alpha, p= 0.004
o be a), and o medium in ensi y (p= 0.045 o alpha,
p<0.001 o be a).
S imula ion Dose-E ec
To s udy he in luence o s imula ion dose on co ical
ac i i y, we compu ed he ERD/ERS o each single ial and
pe o med a eg ession in ime wi hin session appending
same s imula ion in ensi y blocks in o de o appea ance
(blocks o di e en in ensi ies we e p esen ed andomly).
Figu e 6 shows he a e age ERD/ERS o all pa icipan s in
bo h equency bands du ing he 54 ials (3 blocks ×18
ials, see e ical yellow lines) o each in ensi y and a
linea eg ession o i hem. We also pe o med eg essions
o e ime wi hin each block wi h consecu i e ials (see
Supplemen a y Figu e 4).
The i s hing we obse ed is a clea modula ion based on
s imula ion, educing i s a iabili y wi h s imula ion ampli ude.
Low and medium in ensi ies caused a signi ican educ ion
o alpha ERD o e ime (p= 7.8e-4 o low; p= 0.0053
o medium). In con as , high in ensi y caused a signi ican
enhancemen o be a ERD o e ime (p= 5.12e-5). Fu he mo e,
as can be seen in Figu e 6, we obse ed ha in e e y
block (sepa a ed by yellow e ical lines), he e is a educ ion
o he ERD (inc ease o powe plo ed as linea eg ession
in Supplemen a y Figu e 4) p og essi ely induced pe ial
om he i s o he las ial. The i s ial o he
new block p esen ed a la ge ERD (dec ease o powe ) in
compa ison o he ERD o he las ial o he p e ious block
(i espec i e o he s imula ion in ensi y and o de o he block
wi hin he session).
DISCUSSION
This s udy demons a ed he signi ican e ec s o a i ac s,
in ensi y, and dose o neu omuscula elec ical s imula ion
(NMES) o he uppe limb on he ongoing b ain oscilla o y
ac i i y eco ded using EEG. Fi s o all, we deal wi h he
issue o a i ac emo al o allow accu a ely es ima ing he
co ical oscilla o y ac i i y. Reco dings o b ain ac i i y a e easily
pollu ed, especially when elec ical s imula ion in e ac ing wi h
he ne ous sys em is concu en ly used. This con amina ion
can nega i ely a ec he signal- o-noise a io, co e ing he b ain
F on ie s in Neu oscience | www. on ie sin.o g 8Janua y 2021 | Volume 14 | A icle 593360
nins-14-593360 Janua y 15, 2021 Time: 14:6 # 9
Insaus i-Delgado e al. NMES Pa ame e s In luence Co ical Exci abili y
Low in ensi y
0
20
-20
40
-40
ERD/ERS (%)
0504030102
Numbe o ial
y isne nihgiHy isne nimuideM
Alpha band
(7-13 Hz)
0
20
-20
40
-40
ERD/ERS (%)
0504030102
Numbe o ial
0
20
-20
40
-40
ERD/ERS (%)
0504030102
Numbe o ial
0
20
-20
40
-40
ERD/ERS (%)
0504030102
Numbe o ial
0
20
-20
40
-40
ERD/ERS (%)
0504030102
Numbe o ial
0
20
-20
40
-40
ERD/ERS (%)
0504030102
Numbe o ial
Be a band
(14-30 Hz)
R = -0.5222, m = -0.2374, p = 5.12e-05R = -0.1305, m = -0.0687, p = 0.3469R = 0.2189, m = 0.1025, p = 0.1116
R = 0.4433, m = 0.4564, p = 7.8e-04 R = 0.3740, m = 0.3294, p = 0.0053 R = 0.0931, m = 0.0610, p = 0.5030
FIGURE 6 | Compa ison o co ical ac i a ion o e ials o alpha and be a bands. The co ical ac i i y du ing NMES pe iod [(0.5, 2.5) s] quan i ied as ERD/ERS o e
he 54 ials o each s imula ion in ensi y, di ided in o blocks by e ical yellow lines, and a e aged o all he pa icipan s. The pe cen age o ERD/ERS is calcula ed
acco ding o he baseline (−2.5, −1.5) s. Di e en in ensi ies a e compa ed in columns: low (le ), medium (middle), and high ( igh ). Alpha (uppe ow) and be a
(lowe ow) equency bands a e desc ibed. Signi ican co ela ions be ween ERD/ERS and sequence o ials o e session a e ep esen ed wi h black solid linea
eg essions.
ac i i y. Ou indings e idenced ha he median il e enhanced
he de ec ion o senso imo o oscilla o y ac i i y a e emo ing
s imula ion a i ac s.
A e he EEG da a was cleaned, especially o NMES-
induced a i ac s, we analyzed he modula ion o alpha and be a
oscilla ions p oduced by he s imula ion. Powe supp ession,
o desynch oniza ion, o hese equencies has been associa ed
wi h co ical exci a ion, whe eas synch oniza ion e lec s a
s a e o inhibi ion (Klimesch e al., 2007). Du ing high-
in ensi y NMES, he induced desynch oniza ion in alpha
and be a was signi ican ly la ge han du ing s imula ion
a low o medium in ensi ies o no s imula ion. While
below-mo o - h eshold s imula ion in ensi ies only ac i a e
cu aneous mechano ecep o s (e.g., Pacinian co puscles and
Me kel disks) and senso y axons, s imula ion abo e he
mo o h eshold also ec ui s p op iocep i e ecep o s (e.g.,
muscle spindles, Golgi endon o gans, and join a e en s)
(Ma iule i e al., 2008;Be gquis e al., 2011;Golaszewski
e al., 2012). I has been p oposed ha muscle spindles
ansmi inpu s o he spinal co d and can di ec ly in luence
he mo o co ex (M1) h ough he a ea 3a, while he
p ojec ions om a ea 3b ac i a ed by cu aneous eedback
o M1 a e less likely o happen (Ca son and Buick, 2020).
We can he e o e assume ha high-in ensi y NMES leads
o highe neu al exci a ion, p obably by ec ui ing a la ge
numbe o ecep o s de i ed om muscle con ac ions, in
addi ion o he cu aneous and senso y ibe a e ence ha is
also engaged in senso y- h eshold s imula ion, and ha he
ec ui men o muscle spindles esul s in ac i a ion o M1 ia
a ea 3a o he soma osenso y co ex (S1) (Schab un e al.,
2012;Ca son and Buick, 2020). These esul s o in ensi y-
dependen b ain ac i a ion a e in line wi h co icomuscula
esponses (Sasaki e al., 2017), me abolic esponses eco ded
by unc ional magne ic esonance imaging ( MRI) (Backes
e al., 2000;Smi h e al., 2003) and nea -in a ed spec oscopy
(NIRS) (Schü holz e al., 2012), which demons a ed a di ec
quan i a i e associa ion wi h s imula ion in ensi y. No ewo hy,
ou esul s showed ha he co ical ac i i y measu ed wi h EEG,
quan i ied as e en - ela ed (de)synch oniza ion (ERD/ERS),
du ing low and medium in ensi ies was no signi ican ly
di e en o no s imula ion. This sugges s ha below-mo o -
h eshold NMES migh no ec ui enough a e en ibe s o
induce signi ican co ical modula ion and ha mo e a e ence
(p obably h ough muscle con ac ion, p op iocep ion due
o he mo emen , and a la ge numbe o senso y ibe s
ec ui ed) is equi ed o ansmi mo e in o ma ion ha
eaches he b ain and is measu able in he EEG a he
analyzed equencies.
I is well known ha du ing olun a y mo emen , a
s onge co ical ac i a ion is seen in alpha han in be a
(López-La az e al., 2014;Ramos-Mu guialday and Bi baume ,
2015). Such modula ion in alpha and be a co ical ac i i ies
has been ela ed o he con ol o op-down and bo om-
up neu al p ocesses, sugges ing i s ole in he in eg a ion
F on ie s in Neu oscience | www. on ie sin.o g 9Janua y 2021 | Volume 14 | A icle 593360