Mas e ’s Thesis
Mas e in Neu oenginee ing and Rehabili a ion
Imp o ing he quali y o combined EEG-TMS neu al
eco dings: a i ac emo al and ime analysis
REPORT
Au ho : Gema Mijancos Ma ínez
Di ec o : Alejand o Bachille Ma a anz
Joan F ancesc Alonso López
Call: Ap il 2022
Escola Tècnica Supe io
d’Enginye ia Indus ial de Ba celona
Pàg. 2 Repo
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg.3
Resum
La combinació de la TMS (es imulació magnè ica ansc anial, pe les se es sigles en anglès)
i l’elec oence alog ama pe me una a aluació uncional di ec a de egions co icals d’una
mane a con olada, no in asi a i sense necessi a de eali za una asca pe pa del subjec e
d’es udi.
És amb la combinació d’aques es dues ècniques que es ol ca ac e i za les senyal ce eb als
de pe sones amb esquizo ènia (16) i pe sones sanes (15), ob enin els po encials e oca s pel
TMS (TEP). S’a aluen dos p o ocols: SICI (sho -in e al in aco ical inhibi ion) i LICI (long-
in e al in aco ical inhibi ion), que ac i en di e en s ecep o s de ipus inhibi o i (GABA-A i
GABA-B, espec i amen ). Dins d’aques s p o ocols, hi ha senyals amb un sol pols de TMS
(SP) o amb dos polsos (PP). Així doncs, a pa de la ca ac e i zació de la senyal, ambé es
compa en els esul a s en e els di e en s ipus de polsos (1 ó 2), els p o ocols i els dos g ups
d’es udi.
La me odologia seguida és la ípica pe aques ipus de senyals: eliminació del pols o polsos;
aplicació d’ICA (anàlisis de componen s independen s), pe al d’elimina aquelles
componen s a e ac uades i/o so olloses; econs ucció de la senyal amb les componen s
bones i eliminació de canals i expe imen s so ollosos. Un cop s’ha e el p e-p ocessamen de
la senyal i aques a es à ne a, es busquen els po encials e oca s pel pols(os) del TMS. Pe
busca les di e ències en e PP i SP, es es a la p ime a senyal a la segona i es calcula un
a i de modulació.
S’han oba els TEPs P25, N30, P50, N100, P140, N160 i P200. Re e en a les di e ències
en e ipus de polsos, s’obse a que les ampli uds dels TEPs és meno en LICI PP que en
LICI SP. Tanma eix, aques a di e encia no es eu an cla a en el p o ocol SICI. Pel que a les
di e ències en e subjec es, sembla que els con ols (les pe sones sanes) enen més inhibició
que no pas els pacien s.
En conclusió, els esul a s ob ingu s són bas an p ope s als espe a s. Degu a la mida eduïda
de la mos a i a que es ac a d’un camp enca a poc es udia , els esul a s no són exac amen
els de la li e a u a pe ò sí que an en la ma eixa línia.
Pàg. 4 Repo
Resumen
La combinación de la TMS (es imulación magné ica ansc anial, po sus siglas en inglés) y el
elec oence alog ama pe mi e una a aluación uncional di ec a de egiones co icales de una
mane a con olada, no in asi a y sin necesidad de ealiza una a ea po pa e del suje o de
es udio.
Es con la combinación de ambas écnicas que se quie e ca ac e iza las señales ce eb ales
de pe sonas con esquizo enia (16) y pe sonas sanas (15), ob eniendo los po enciales
e ocados po el TMS (TEPs). Se e alúan dos p o ocolos: SICI (sho -in e al in aco ical
inhibi ion) y LICI (long-in e al in aco ical inhibi ion), que ac i an di e en es ecep o es de ipo
inhibi o io (GABA-A y GABA-B; espec i amen e). Den o de es os p o ocolos, hay señales
con un solo pulso de TMS (SP) o con dos pulsos (PP). Así pues, a pa e de la ca ac e ización
de la señal, ambién se compa an los esul ados en e los di e en es ipos de pulsos, los
p o ocolos y los dos g upos de es udio.
La me odología seguida es la ípica pa a es e ipo de señales: eliminación del pulso o pulsos;
aplicación de ICA (análisis de componen es independien es), con el in de elimina las
componen es a e ac uadas y/o uidosas; econs ucción de la señal con las componen es
buenas y eliminación de canales y expe imen os uidosos. Una ez hecho el p e-p ocesado
de la señal y és a es á limpia, se buscan los TEPs. Pa a busca las di e encias en e PP y SP,
se es a la p ime a señal a la segunda y se calcula una a io de modulación.
Se han encon ado los TEPs P25, N30, P50, N100, P140, N169 y P200. Re e en e a las
di e encias en e ipos de pulsos, se obse a que las ampli udes de los TEPs son meno es en
LICI PP que en LICI SP. Sin emba go, es a di e encia no se e an cla a en SICI. En cuan o a
las di e encias en e suje os, pa ece se que los con oles (pe sonas sanas) ienen más
inhibición que los pacien es.
En conclusión, los esul ados son muy simila es a los espe ados. Debido al amaño educido
de la mues a y a que es un campo poco es udiado, los esul ados no son exac amen e iguales
a los de la li e a u a, pe o sí que an en la misma línea.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg.5
Abs ac
The combina ion o TMS ( ansc anial magne ic s imula ion) and EEG
(elec oencephalog aphy) allows a unc ional assessmen o co ical egions in a con olled,
non-in asi e way and wi hou he need o ha ing subjec s unde s udy pe o m a ask.
Thanks o his combina ion, a cha ac e iza ion o he ce eb al signals o schizoph enic pa ien s
(16) and heal hy con ols (15), h ough he TMS-e oked po en ials (TEPs), will be pe o med.
Two p o ocols will be e alua ed: SICI (sho -in e al in aco ical inhibi ion) and LICI (long-
in e al in aco ical inhibi ion), which ac i a e di e en inhibi o y- ype ecep o s (GABA-A and
GABA-B, espec i ely). In bo h p o ocols, he e a e signals wi h a single TMS-pulses (SP) o
wi h pai ed-pulses ( wo pulses, PP). Thus, besides he cha ac e iza ion, he di e en esul s
ob ained will be compa ed be ween ype o pulses, p o ocols and g oups o s udy.
The me hodology ollowed is he ypical one o his ype o signals: emo al o he TMS-
pulse(s); ICA (independen componen analysis) applica ion o dele e he a e ac ua ed o
noisy componen s; signal econs uc ion wi h he good componen s and bad channel and bad
ial ejec ion. Once he p e-p ocessing is inished and he signal is clean, he TMS-e oked
po en ials a e ob ained. Fo he pu pose o inding he di e ences be ween PP and SP, he
o me signal is sub ac ed om he la e and a modula ion a io is compu ed.
The TEPs ound a e P25, N30, P50, N100, P140, N160 and P200. Rega ding he di e ences
be ween ypes o pulses, he obse ed ha he TEP ampli udes o LICI PP a e lowe han he
ones o LICI SP. Howe e , his di e ence is no clea ly seen in SICI p o ocol. As o he
di e ences be ween subjec s, i seems ha he con ols ha e g ea e inhibi ion han he
pa ien s.
In conclusion, he esul s a e ema kably simila o he ones expec ed. Due o he small sample
size and he ac ha i is s ill a poo ly s udied iled, some esul s do no ma ch comple ely wi h
he ones in he li e a u e, bu hey ollow he same endency.

Pàg. 6 Repo
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg.7
Summa y
SUMMARY ___________________________________________________ 7
1. GLOSSARY ______________________________________________ 9
2. PREFACE _______________________________________________ 13
2.1. O igin o he p ojec ...................................................................................... 13
2.2. Mo i a ion ..................................................................................................... 13
2.3. P e ious equi emen s ................................................................................. 14
3. INTRODUCTION __________________________________________ 15
3.1. Objec i es ..................................................................................................... 15
4. THEORETICAL FRAMEWORK ______________________________ 16
4.1. Schizoph enia ............................................................................................... 16
4.1.1. Symp oms ....................................................................................................... 16
4.1.2. B ain a eas in ol ed in schizoph enia ............................................................. 17
4.2. Non-in asi e b ain s imula ion...................................................................... 18
4.2.1. CS .................................................................................................................. 18
4.2.2. TMS ................................................................................................................ 19
4.3. EEG .............................................................................................................. 21
4.3.1. EEG p ocessing .............................................................................................. 21
4.3.2. TMS-EEG ....................................................................................................... 22
4.4. TMS-EEG and schizoph enia ....................................................................... 28
4.4.1. Cha ac e iza ion .............................................................................................. 28
4.4.2. Clinical applica ion .......................................................................................... 31
5. METHODOLOGY _________________________________________ 35
5.1. Signal acquisi ion and da ase ...................................................................... 35
5.2. Algo i hm/Pipeline......................................................................................... 38
5.2.1. P e-p ocessing ................................................................................................ 40
5.2.2. P ocessing ...................................................................................................... 55
5.2.3. S a is ical analysis ........................................................................................... 58
6. RESULTS AND DISCUSSION _______________________________ 59
6.1. Clean signal .................................................................................................. 59
6.2. TMS-e oked po en ials (TEPs) .................................................................... 61
6.2.1. All subjec s ...................................................................................................... 61
6.2.2. Compa ison be ween ype o subjec s and pulse ype .................................... 65
6.3. Single-pulse signal s pai ed-pulse signal .................................................... 67
Pàg. 8 Repo
6.3.1. Tempo al signal ............................................................................................... 67
6.3.2. Spa ial dis ibu ion o TEPs .............................................................................. 69
6.3.3. Modula ion a io (PP/SP) .................................................................................. 75
7. PROJECT’S SCHEDULE ___________________________________ 77
8. FUTURE WORK __________________________________________ 78
9. ENVIRONMENTAL IMPACT ANALYSIS _______________________ 79
10. ECONOMICAL ANALYSIS __________________________________ 80
CONCLUSIONS ______________________________________________ 83
ACKNOWLEDGEMENTS _______________________________________ 87
BIBLIOGRAPHY ______________________________________________ 89
Re e ences ............................................................................................................ 89
Complemen a y bibliog aphy ................................................................................. 92
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 9
1. GLOSSARY
AEP: Audi o ily e oked po en ial; po en ial igge ed by a sound.
AMT: Ac i e mo o h eshold.
AP: Ac ion Po en ial.
AUC: A ea unde he cu e.
CS: Condi ioning s imulus.
CSP: Co ical Silen Pe iod.
dB: Decibel.
DLPFC: Do sola e al p e on al co ex.
dTMS: Deep TMS.
EEG: Elec oencephalog aphy.
EMG: Elec omyog aphy.
ERD: E en - ela ed desynch oniza ion.
ERP: E en - ela ed po en ial.
ERS: E en - ela ed synch oniza ion.
MIR: Func ional magne ic esonance imaging.
HC: Heal hy con ol.
HCUV: Hospi al Clínico Uni e si a io de Valladolid.
HEOG: Ho izon al elec ooculog aphy.
IC: Independen componen .
ICA: Independen componen analysis.
ICF: In aco ical acili a ion.
Pàg. 16 Repo
4. THEORETICAL FRAMEWORK
4.1. Schizoph enia
Schizoph enia is a ch onic b ain diso de (o men al diso de ) cha ac e ized by he al e a ion
o cogni i e p ocesses. I may esul in some combina ion o hallucina ions ( ypically hea ing
oices), delusions and ex emely diso de ed hinking and beha iou ⎯such as diso ganized
speech, social wi hd awal o lack o mo i a ion⎯ ha impai s daily unc ioning, and can be
disabling. Thus, schizoph enia is a psychosis ha a ec s 20 million people wo ldwide,
acco ding o he WHO.
E ec i e li elong ea men s a e a ailable. Wi h hese ea men s, mos symp oms o
schizoph enia imp o e g ea ly and he likelihood o ecu ence can be diminished. Simila ly o
o he cogni i e impai men s, he ou look amelio a es i he pa ien is diagnosed in an ea ly
s age.
Men a e usually diagnosed in hei la e eens o ea ly wen ies, while women a e diagnosed in
hei la e wen ies and hi ies, hough i can occu a any age. Mo eo e , i s se e i y, du a ion
and equency a y om pe son o pe son. Some may ha e only one psycho ic episode, while
o he s may ha e many episodes h oughou hei li e ime bu li ing a ela i ely no mal li e. On
he o he hand, o he symp oms migh wo sen o e ime.
4.1.1. Symp oms
Some o he ea ly symp oms a e sub le, including change in g ades ( o eens), social
wi hd awal, ouble concen a ing, empe la es and di icul y speaking.
In addi ion, he symp oms can be classi ied as ollows:
⎯ Diso ganized symp oms: con used and diso de ed hinking and speech, ouble wi h
logical hinking, p oblems making sense o e e yday sigh s, sounds, and eelings and
some imes abno mal mo emen s, among o he s.
⎯ Cogni i e symp oms: p oblems in a en ion, concen a ion and memo y. Fo ins ance:
o ge ing o losing i ems, no ha ing wo king memo y (using he in o ma ion
immedia ely a e lea ning i ) o being unable o decide due o di icul y in p ocessing
in o ma ion. They some imes migh su e om anosognosia (being unawa e o hei
own men al heal h condi ion o no pe cei ing i accu a ely).

Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 17
⎯ Nega i e symp oms: ela ed o hose ha diminish a pe son’s abili ies. Some o hem
a e lack o mo i a ion, di icul ies in planning, beginning and sus aining ac i i ies,
educed exp ession o emo ions ia acial exp ession o oice one, educed speaking,
lack o he abili y o expe ience pleasu e and poo hygiene and g ooming habi s.
⎯ Posi i e symp oms: e e ing o added hough s o ac ions ha a e no based in eali y:
hallucina ions (as commen ed be o e, usually hea ing oices), delusions (belie s no
suppo ed by objec i e ac s, such as pa anoia), ca a onia (whe e he pe son may s op
speaking and hei body may be ixed in a single posi ion o a e y long ime) and
exagge a ed o dis o ed pe cep ions, belie s and beha iou s.
4.1.2. B ain a eas in ol ed in schizoph enia
Schizoph enic pa ien s appea o ha e al e ed connec i i y (hypo and hype ) and ci cui y,
caused a leas by he suscep ibili y genes and a bi a ed h ough a pe u ba ion o b ain
de elopmen and synap ic plas ici y [1], [2]. Thus, he ongoing neu oplas ic changes play a
ole in he cou se o he disease and he neu opa hology in ol es in e connec ed ci cui s and
associa ed b ain a eas [2]. I also has been p o en ha he e is no gliosis (p oli e a ion and
hype ophy o he suppo ing cells o he b ain) in SCZ, hus he diso de is likely o be
neu ode elopmen al in o igin [1].
Me a-analysis o s uc u al magne ic esonance imaging (MRI) e eals g ey ma e olume
de ici s in he empo al lobe, he pa ie al co ex and he halamus. Pos -mo em s udies ha e
shown olume loss in he hippocampus [3]. Besides, hey a e in acco dance wi h he esul s
o he MRI s udies, as hey also showed a change o olume o he empo al lobe as well as a
change in he numbe and size o neu ons and he densi y [4]. This change in size, making
hem smalle , sugges s less ex ensi e axonal and dend i ic ees and hence, he neu ons may
be making ewe connec ions [1]. Rega ding he on al lobe, which has also a educed olume,
he cell densi y and hei size ha e been ound educed along wi h a dec eased in e media e
neu on, which seems o be ela ed o he low ac i i y o he on al lobe seen in unc ional
s udies [4].
Di usion enso imaging s udies demons a ed ha he myelin o he memb anes was educed
(which may lead o impai ed ne e cell p opaga ion o in o ma ion) as well as he whi e ma e
aniso opy in he deep le p e on al and empo al co ex [3]. Func ional magne ic esonance
imaging s udies ha e shown dis u bed connec i i y in complex hippocampal, p e on al and
ce ebella - halamic-p e on al ne wo ks [3].
Rega ding gene ics, a dec eased exp ession o glu ama e gic and gamma-amino-bu y ic acid
(GABA)e gic synap ic p o eins and i s consecu i e dis u bance o mic oconnec i i y has been
epo ed [3]. Mo eo e , mu a ions o he No ch4 gene, which con ols unc ions such as
Pàg. 18 Repo
inc eases, di e en ia ion and mig a ion o ne e s em cells, ha e been epo ed. O he
candida es genes o be suscep ible a e i) DISC1, ha ha e a ole o adjus men o ne ous
sys em mig a ion and ex ension o ne ous sys em p ojec ions du ing he neu onal
de elopmen pe iod [4]; ii) he eelin gene ( ela ed o he o ma ion o neu onal connec ions),
since abno mal egula ion and exp essions ha e been epo ed [1] and iii) he BDNF (b ain-
de i ed neu o ophic ac o ) gene, which encodes he exp ession o he BDNF o ab ineu in, a
neu o ophic ac o essen ial o egula ing he su i al, di e en ia ion, mo phology and
synap ic emodelling o neu ons du ing de elopmen [4].
4.2. Non-in asi e b ain s imula ion
Non-in asi e b ain s imula ion (NIBS o NBS) is a se o di e en echnologies and echniques
ha allow he s imula ion o he b ain by al e ing he b ain ac i i y h ough he ce eb al co ex,
wi hou he necessi y o conduc ing a su gical p ocess o en e ing any de ice in o he head o
he subjec [5].
The e a e wo di e en echniques: ansc anial cu en s imula ion ( CS) and ansc anial
magne ic s imula ion (TMS). One o he di e ences be ween hem is ha TMS p o okes he
depola iza ion o he cells, which leads o he c ea ion o he ac ion po en ial (AP). Howe e ,
CS simply induces memb ane depola iza ion, modula ing he exci abili y o a egion and
changing he p obabili y o ha ing an AP [5-6].
These echniques can be used ei he o esea ch o diagnosis, as hey a e use ul o obse e
disease- ela ed changes in b ain ac i a ion, inhibi ion o connec i i y. Mo eo e , i also can be
used o ea ing di e en b ain diso de s (e.g. dep ession) and in neu o ehabili a ion, o
ins ance a e a s oke o a spinal co d inju y [7].
4.2.1. CS
CS, which da es back o he 19 h cen u y, in ol es di ec ( DCS) and al e na ing ( ACS) cu en
s imula ion. Fo he o me , a low-ampli ude (0.5-4 mA) di ec cu en is applied o he scalp
using elec odes o some seconds o e en minu es [8]. The elec ic cu en lows om he
ca hode o he anode pene a ing he skull and modi ying he neu onal ansmemb ane
po en ials (inc easing o dec easing he exci abili y) in he cu en pa h. Usually, he exci abili y
o he egion unde he ca hode dec eases and he one in he anode inc eases [5].
On he o he hand, ACS uses a cu en ha al e na es a a equency speci ied by he ope a o
o di ec ly in e e e wi h ongoing b ain oscilla ions [9].
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 19
4.2.2. TMS
The ansc anial magne ic s imula ion was i s p esen ed in 1985 by An hony Bake and co-
wo ke s om he Uni e si y o She ield. Based on Fa aday’s Law, when a pulse o elec ic
cu en passes h ough a coil, i gene a es a apidly changing magne ic pulse, i he cu en
has su icien s eng h and sho du a ion. Being he coil placed in he subjec ’s head he
magne ic pulse gene a ed pene a es his/he scalp and skull o each he co ex wi h negligible
a enua ion. Then, he pulse o he magne ic ield induces a seconda y ionic cu en in he
b ain, which can igge ac ion po en ials in co ical neu ons [10]. I can be applied epea edly,
p oducing changes in he co ex ha las mo e han he s imula ion.
Among o he b ain s imula ion me hods, TMS is unique since i ac i a es all i s p ima y a ge
neu ons a he same ime [11].
4.2.2.1. De ailed s imula ion
The cu en lows sho ly in he coil and gene a es he magne ic ield ha induces an elec ic
cu en in he unde lying issues, e oking mul iple esponses in he co icospinal neu ons.
Depending on he si e o he s imula ion (whe e on he co ex i is applied), he e will be di e en
beha iou al esul s. Fo ins ance, applying he magne ic s imula ion on he mo o co ex a
mo emen in he con ala e al muscles migh be induced; o applying i on he isual co ex
migh induce he pe cep ion o a beam o ligh .
The elec ic cu en ac s on neu ons, inhibi ing o s imula ing hem and a ec ing hei unc ion,
as messenge s o elec ical signals. All he e ec s (so, he ac i a ion o he co ex) a e
de e mined by he in ensi y o he magne ic ield, he equency and du a ion o he magne ic
pulses as well as by he shape, size, ype and o ien a ion o he coil - he induced ield- wi h
espec o he sulci (a dep ession o g oo e in he ce eb al co ex) being pe pendicula o he
co ical su ace he op imal di ec ion [11-12].
The a ea s imula ed is only he s a ing poin o he desi ed e ec , bu hanks o he ne o
neu al ci cui s di e en dis an a eas will be a ec ed. In addi ion, his widesp ead ne wo k
allows he TMS o gene a e neu omodula ing e ec s in deep b ain and b ains em a eas whe e
cu en canno be induced di ec ly [12]. Thus, he TMS-e oked esponse obse ed a e wa ds
is p obably due o cellula mechanisms ha we e igge ed by he pulse, no by he pulse i sel
o emnan s o accumula ed cha ge [11].
Rega ding he equency o he pulse, he TMS will ha e an inhibi o y ( o low equencies,
meaning 1 Hz o less) o exci a o y e ec ( o high equencies, i.e, g ea e han 1 Hz).
Mo eo e , he pulses can be adminis e ed as single, pai ed o in se ies, which is called
epe i i e TMS ( TMS). The i s wo ypes (single and pai ed) a e used o neu odiagnos ic
Pàg. 20 Repo
pu poses and he la e has a he apeu ic bene i in psychia ic diso de s [13].
The bes -known e ec is he mo o -e oked po en ial (MEP) measu ed om pe iphe al muscles
a e he applica ion o a TMS pulse in he mo o co ex. Mo eo e , om es ima es o synap ic
delays and axonal ansmission eloci ies, he ime o ac i a ion o o he b ain egions a e he
pulse can be p edic ed. TMS may also igge oscilla o y ac i i y o pe u b ongoing hy hms;
o ins ance, e en - ela ed synch oniza ion (ERS) o desynch oniza ion (ERD) [11].
Figu e 1. Magne ic lux o he coil and he induced cu en in he b ain. F om: Halle , M. “T ansc anial magne ic
s imula ion and he human b ain”. Na u e 406, 147–150, 2000.
4.2.2.2. Coils
As commen ed abo e, he magne ic s imula ion is p o ided by a coil inside o which he e is a
coppe conduc o wi e ha p oduces he cu en ha lows in a ci cula way gene a ing he
magne ic ield. Se e al ypes o coils can be used, ega ding how deep he s imula ion should
be: ound and Fo8 (bu e ly) o he s anda d TMS (less deep) and double cone coils, H o
HCA o he dTMS (deep TMS) [14].
Figu e 2. Bu e ly (le ) and double cone ( igh ) coils.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 21
4.2.2.3. Risks and side e ec s
I is conside ed a sa e echnique, as i is non-in asi e. Ne e heless, i can ha e some side
e ec s. The mos common ones a e headache, discom o a he s imula ion si e, dizziness
and ickling and spasms in he acial muscles. Mo eo e , he isk o su e ing om epilep ic
seizu es du ing he TMS is e y low.
The main ela i e con aindica ions o TMS a e p egnan women and child en unde 2 yea s
o age. Rega ding he absolu e con aindica ions, he e a e pa ien s wi h uncon olled epilepsy,
pa ien s wi h co po al elec onic de ices (such as pacemake s, implan able de ib illa o s,
insulin pumps…) o in ac anial e omagne ic elemen s (s en s, pla es, sc ews, e c.).
4.3. EEG
Elec oencephalog aphy (EEG) is a non-in asi e neu ophysiological echnique ha eco ds
he elec ical ac i i y o he b ain h ough he elec odes placed in he scalp o he subjec . This
elec ical ac i i y is p oduced by he ionic cu en s gene a ed du ing he neu onal ac i i y and
i was cap u ed o he i s ime by he psychia is Hans Be ge in 1928 and since hen i has
been used as a ool o diagnosis, mainly in he ields o neu ology and neu osu ge y [15].
The EEG is he empo al and spa ial sum o he pos -synap ic po en ials o he py amidal
neu ons. These neu ons gene a e ionic cu en s ha con ibu e o he gene a ion o an
elec ical ield, which is eco ded by he EEG. Ne e heless, his ield canno be cap u ed
“clean”, as i is dis o ed by b ain s uc u es such as he c anial bone o he ce eb ospinal luid.
The signal o he ield is commonly ep esen ed in a g aph o ol age s ime. The EEG signal
a ies be ween subjec s, as i is complex, a iable and non-uni o m and he loca ion o he
elec odes migh be modi ied. I usually anges be ween 10 µV and 100 µV [15].
The e a e wo main ypes o eco ding: basal o es ing-EEG and e en - ela ed po en ials
(ERP). Fo he es ing-EEG, he subjec mus be es ing and usually wi h his/he eyes closed.
On he o he hand, o cap u e he ERP, a s imulus is needed ( ypically isual o audi o y) o
examine i s esponse. In his p ojec , he esponse o a TMS pulse is s udied, being he pulse
he s imulus.
4.3.1. EEG p ocessing
The p ocessing o bio-signals and, hence, EEG p ocessing has he objec i e o ex ac ing
meaning ul in o ma ion o he signals in o de o cha ac e ize diso de s, diseases and
pa hologies as well as o acili a e he wo k o he medical s a . Wi h he cha ac e iza ion, a
be e and mo e p ecise diagnosis is possible and can be used o help he p o essionals o

Pàg. 22 Repo
make and/o o suppo a decision [16].
Al hough h ough he yea s he p ocessing has been isual, ad anced me hodologies o signal
p ocessing allow he iden i ica ion o ea u es o bioma ke s non-de ec able isually. Mo eo e ,
as i is an au oma ic (o semi-au oma ic) p ocess, i dec eases he ime needed o i s inspec ion
and inc eases he objec i i y (i is no as biased as he isual inspec ion, whe e he expe ise
o he p o essional and his/he pe sonal bias ha e an impo an impac , as well as a igue) and
uni o mi y in he diagnosis.
The EEG p ocessing can be di ided in h ee di e en s ages:
1. Reco ding o he signals and i s p e-p ocessing (mainly cleaning hem, emo ing
a e ac s and denoising hem).
2. P ocessing (segmen a ion, il e ing, ea u e calcula ion, e c).
3. Classi ica ion o he signal.
4.3.2. TMS-EEG
TMS-EEG allows he possibili y, non-in asi ely, o a di ec unc ional assessmen o speci ic
co ical egions in a con olled manne wi hou he necessi y o pe o ming a ask [11], [17].
To ob ain a good signal- o-noise a io (SNR) i is impo an ha he eco ding sys em has a low
noise le el bu a he same ime, i mus be insensi i e o i mus eco e quickly om he TMS
pulse [11]. E en wi h he bes echnology a ailable, he signal will be subjec o di e en kinds
o a e ac s. The mos common a e b ie ly desc ibed below, bu di e en TMS machines,
expe imen al a angemen s and s udy designs can al e he eco dings and he physiological
TMS-EEG a e ac p o iles, esul ing in longe o new a e ac s [18].
4.3.2.1. A e ac s
4.3.2.1.1 Ha dwa e a e ac s
S anda d EEG sys ems canno be used oge he wi h TMS because he pulse may sa u a e
adi ional ampli ie s. Ha dwa e solu ions ha e been ound o deal wi h i . Fu he mo e, he
echa ge o he TMS de ice o he nex pulse may cause a signi ican a e ac . To a oid his
p oblem, monophasic de ices mus be used o inse a echa ging delay ci cui in he
s imula ion ins umen s [11, 17].
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 23
4.3.2.1.2 Muscle ac i i y
I is he mos pe asi e a e ac and esul s om he inad e en s imula ion o scalp muscles
when he TMS pulse is igge ed. I is a peak ha appea s 5 o 8 ms a e he pulse ( he pulse
is also seen in he EEG eco ding), making impossible he eco e y o neu al ac i i y ea lie
han 15 o 20 ms. Usually, he TMS pulse and a ew milliseconds a e i a e emo ed om he
EEG, hus some muscle ac i i y emains. This is he eason why some esea che s call his
a e ac he emaining- ail end [17-18]. The a e ac can be educed (e en hough some pa
o he da a will be emo ed) by mo ing o eo ien ing he coil, by educing he TMS in ensi y o
by a combina ion o hem [11].
Ne e heless, he muscle ac i i y can also be eco ded h oughou he eco ding, as he scalp
muscles can be ac i a ed by o he easons apa om he TMS pulse. The muscles mos likely
o be ac i a ed a e he neck, he mas ica ion, acial, on al, empo al o masse e [11].
4.3.2.1.3 Decay a e ac
I is also a peak bu smalle in ampli ude han he p e ious a e ac , o en demons a es an
opposi e pola i y and has a slowe eco e y han he muscle a e ac . I is also ela ed o he
size o TMS-e oked muscle ac i i y [17].
These wo a e ac s (muscle ac i i y e oked by he TMS pulse and he decay a e ac ) eco e
wi hin 10 o 12 ms, al hough small baselines o se may s ill be p esen depending on he
con igu a ion [18].
4.3.2.1.4 Blinks and eye mo emen a e ac
The e is a s eady po en ial o se e al milli ol s, which can be la ge compa ed o he EEG
signal, ac oss each eyeball. Some subjec s pe o m sys ema ic eye mo emen s and all o hem
blink ( igge ed by he TMS pulse), i s con amina ing he signal. In he eco ding, a peak a 90
ms app oxima ely is obse ed and opog aphy concen a ed o e an e io elec odes. Thus,
blinks a e o igin and ime-locked. I s emo al has minimal e ec in he TMS-e oked ac i i y
[11, 17].
4.3.2.1.5 Audi o y-e oked po en ial (AEP) and soma ic sensa ion
When he pulse is igge ed a loud click (up o 120 dB) is p oduced, which ac i a es he audi o y
sys em o he subjec . This can be obse ed in he eco ding as peaks be ween 80 and 100
ms (a e he pulse) o e he on o-cen al egions. To gua an ee he elimina ion o he
esponse, a masking sound wi h he same equency spec um as he TMS clicks ( o minimize
i s powe ) mus be used in addi ion o hea ing p o ec ion. As he subjec is al eady hea ing a
Pàg. 24 Repo
sound simila o he TMS-pulse, he pulse would be masked by i and he e will no be a huge
audi o y-e oked po en ial. I is impossible o be comple ely con iden ha emo al does no
a ec TEPs (TMS-e oked co ical po en ials) [11, 17].
Fu he mo e, TMS-elici ed scalp sensa ions p oduce e oked esponses. I is e y di icul o
e alua e he deg ee o con ibu ion o soma osenso y signals due o TMS-ac i a ed muscles
mo emen s. Howe e , esea che s ha e concluded ha i is no a majo p oblem [11].
4.3.2.1.6 Elec odes and noise- ela ed a e ac s
When he elec ode mo es wi h espec o he elec oly e, he dis ibu ion o he cha ge a he
in e aces is dis u bed and a momen a y change o he po en ial is no ed. This displacemen
can be caused by muscle mo emen s, acciden al con ac o he coil o he elec odes du ing
s imula ion o by elec omagne ic o ces. The elec ode can also be pola ized and i may ake
up o hund eds o milliseconds o e u n o he equilib ium po en ial a e he pulse [11].
In addi ion, i is impo an o check ha he impedance is always below 5 kΩ since he a e ac s
om elec ode mo emen o pola iza ion a e smalle when he esis ance is low. Fu he mo e,
he con ac be ween he TMS coil and he elec odes no only can “spoil” he eco ding bu also
hea he elec odes, wi h he isk o p o oking skins bu ns [11].
Topog aphically, he noise- ela ed a e ac s a e cen ed on a single elec ode and do no
display ac i i y ime-locked o he TMS pulse and ypically accoun o ou lie ac i i y in
indi idual channels [17].
I is impo an o bea in mind ha e en wi h s a e-o - he-a EEG sys ems, he eco ding o
TMS-EEG is s ill a challenge. The e o e, i is necessa y o sys ema ically analyse he design,
execu ion and da a analysis o TMS-EEG eco dings.
4.3.2.2. Typical pipeline o TMS-EEG acquisi ion and p e-p ocessing
4.3.2.2.1 Acquisi ion
Skin p epa a ion is essen ial o a p ope signal acquisi ion and i mus be cleaned be o ehand.
The elec ode impedance has o be kep below 5 kΩ by using enough amoun o gel bu
a oiding sho -ci cui s be ween close elec odes. The coil should be immobilized wi h espec
o he head, wi hou ouching o indi ec ly a ec ing he elec odes du ing he measu emen . To
ob ain a be e and mo e p ecise s imula ion, neu ona iga ion echniques such as MRI o MRI
o he subjec ’s head can be used. I is especially use ul when he a ge a ea o he co ex is
beha iou ally silen : i s s imula ion does no p o oke an obse able esponse [11].
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 25
4.3.2.2.2 P o ocols
The e a e di e en p o ocols and ypes o pulse depending on wha is going o be assessed
(co ical exci abili y, plas ici y o connec i i y) [19]. Focusing on he ype o pulse, i can be
single-pulse; meaning ha each ial o epoch will ha e jus one TMS pulse (a es s imulus
only, TS), o pai ed-pulse; whe e each ial will ha e wo pulses, being he i s pulse he
condi ioning s imulus (CS), which usually is se o a sub h eshold in ensi y (low-in ensi y), and
he second he es s imulus (TS), se o a sup a h eshold in ensi y (high-in ensi y) [19-20]. The
h eshold in ensi y, and hence, wha es ablishes i he in ensi y applied is ei he sub h eshold
o sup a h eshold is he es ing mo o h eshold (RMT), which is he minimum in ensi y needed
o p oduce a mo o -e oked po en ial (MEP) ha exceeds a peak- o-peak ampli ude (usually
50 µV) in a leas hal o he ials when he subjec is a es . O he h esholds can be used,
such as he ac i e mo o h eshold (AMT), de ined as he minimum in ensi y needed o p oduce
a MEP g ea e han 0.1 mV (in a leas hal o he ials) when he subjec is pe o ming an
isome ic con ac ion o he a ge muscle a abou 10 o 30% o he maximum o ce. Gene ally,
he sub h eshold is a ound 60 o 80% o he RMT o 70 o 90% o he AMT.
The pai ed-pulse TMS (ppTMS) can e lec co ical inhibi ion and acili a ion a speci ic
in e s imulus in e als (ISIs). Rega ding his in e al, he e is a s imula ion pa adigm o sho
o long-in e al in aco ical inhibi ion, SICI and LICI, espec i ely, which a e also he names o
he p o ocols. LICI occu s ollowing ppTMS wi h an in e s imulus in e al be ween 50 and 200
ms and bo h pulses a e deli e ed a a sup a h eshold in ensi y. On he o he hand, SICI occu s
in esponse o a sub h eshold condi ioning s imulus ollowed by a sup a h eshold s imulus wi h
an ISI be ween 1 and 6 ms [19, 21].
O he pa adigms exis , such as he sho -la ency a e en inhibi ion (SAI), he in aco ical
acili a ion (ICF), o he co ical silen pe iod (CSP), which will no be explained.
Figu e 3. Diag am o pai ed-pulse pa adigm.
Pàg. 32 Repo
Fu he mo e, he use o non-in asi e neu omodula ion may be mo e app op ia e han in asi e
one, as hey a e gene ally well- ole a ed and hold a conside able sa e y p o ile [2].
Up- o-da e a emp s ha e been made o e adica e and/o mi iga e posi i e, nega i e and
cogni i e symp oms h ough neu os imula ion.
4.4.2.1. Posi i e symp oms
Mos s udies ha e ocused on ea ing hese symp oms, especially audi o y hallucina ions. To
do so, low- equency TMS has been used o educe hype -ac i a ion p esen du ing speech
p ocessing, which happens in audi o y hallucina ions. The e is a high deg ee o a iabili y in
he s imula ion pa ame e s, b ain egion a ge ed, he deg ee o ea men esis ance, and
du a ion o ea men [35]. None heless, he majo i y o s udies s imula e he le empo o-
pa ie al co ex. Howe e , he e a e mixed esul s; while some a icles yielded posi i e
ou comes o he s do no . Some me a-analyses ha e been pe o med p o iding he same
in o ma ion [13, 39]. In addi ion, he s imula ion has limi ed bene i s o o he psycho ic
symp oms [35].
Due o he a iable esul s, in es iga o s ha e a emp ed o examine means o enhance he
e icacy and du abili y o he TMS e ec s by applying high- equency (ins ead o low), usage
o deep TMS o s imula ing o he egions (B oca’s a ea, DLPFC, he middle empo al gy us,
We nicke’s a ea, he supe io empo al gy us, e c.) ob aining mixed esul s [35, 39].
Likewise, a Coch ane e iew led o he same conclusions: e en hough some s udies ha e
been success ul in educing posi i e symp oms, some o he s ha e no ound enough
e idence. Fu he mo e, he quali y o hem was e y low and he sample size small [40].
4.4.2.2. Nega i e symp oms
TMS is ypically adminis e ed o enhance he “hypo on ali y” o he dominan p e on al co ex
o imp o e nega i e symp oms [35].
He e again, some s udies ha e a posi i e esul when applying high- equency TMS o SCZ
pa ien s, while o he s did no ind any subs an ial imp o emen [13]. Highe pulse equency,
s onge s imulus in ensi y, longe ea men du a ion, younge age and sho e du a ion o
illness we e associa ed wi h be e ou comes. Also, he e was a end sugges ing a wo sening
o posi i e symp oms while ying o mi iga e nega i e ones. Ne e heless, hese a e sho -
e m e ec s, om small ials and in luenced by he e ogenei y in pa ien s and TMS
adminis a ion. Mo eo e , sham s imula ion also esul ed in a signi ican wi hin-g oup
imp o emen [35].
These obse a ions ha e led o he in es iga ion o newe si es o s imula ion, ha ha e ended

Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 33
in mixed and limi ed esul s [35]. Thus, opinions a y as o he e ec i eness o high- equency
TMS in ea ing nega i e symp oms [13].
4.4.2.3. Cogni i e symp oms
TMS had limi ed de imen al e ec s on cogni ion, and could pe haps be le e aged o enhance
cogni ion when deli e ed unde s ic ly moni o ed condi ions [35]. Some o he s udies ea ing
nega i e symp oms ha e also p esen ed imp o emen in dep essi e symp oms and
neu ocogni i e de ici s [13]. A con olled ial and a me a-analysis demons a ed signi ican
imp o emen in wo king memo y. Ano he one showed ha cogni ion imp o ed only in one o
he se en domains assessed ( e bal luency). Simila imp o emen s in o e all neu ocogni i e
pe o mance we e obse ed in a ial wi h a small sample. Howe e , a la ge, mul i-cen e s udy
ound no bene i s [35].
Pàg. 34 Repo
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 35
5. METHODOLOGY
5.1. Signal acquisi ion and da ase
Signals we e acqui ed a he HCUV. They used a 64-channel EEG ampli ie (B ainVision,
ac iCHamp 64) ollowing he 10-10 in e na ional sys em. Six y-one channels o he cap we e
used (see Figu e 5) and he emaining 3 we e used o eco d he VEOG, HEOG
(elec ooculog aphy e ical and ho izon al, espec i ely) and EMG (elec omyog aphy). The
elec ode impedance was kep below 5kΩ and he equency sample was 25000 Hz. The TMS
machine was a bu e ly coil o MagP o X100 o MagVen u e, speci ically he model MCF-B70.
The signals we e eco ded wi h B ain-Vision Reco de p og amme.
Figu e 5. EEG sys em o eco ding he signals. The elec odes in o ange we e he ones used.
Figu e 6. Bu e ly TMS-coil used in he eco ding.
Pág. 36 Repo
The p o ocols employed o TMS we e LICI and SICI wi h pai ed-pulses (PP) and single pulses
(SP) (mo e in o ma ion abou he pa adigms in Sec ion 4.3.2.2.2). Fo he single pulse, hey
used an in ensi y 120% o he RMT. Fo he pai ed-pulse depending on he p o ocol, LICI o
SICI, hey used 120% RMT o he condi ioning and es s imulus o 80% o he condi ioning
s imulus and 120% RMT o he es s imulus, espec i ely (see Figu e 7). The in e s imulus
in e al also a ied be ween LICI and SICI. Fo he o me , ISI had a alue o 100 ms and o
he la e , a alue o 4 ms. In bo h p o ocols, SICI and LICI, he in e als be ween pulses
(independen ly i hey a e single o pai ed) a e semi- andomized be ween 5 and 7 seconds in
o de o p e en he an icipa ion o he nex pulse.
Figu e 7. P o ocols ollowed o he acquisi ion. A) Top: SICI PP. Bo om: LICI PP. B) SP.
The i s s ep o he acquisi ion is o de e mine he RMT ( es ing mo o h eshold) o he subjec .
I is de e mined h ough he ela i e equency me hod [41]; de ined as he minimum in ensi y
needed o e oke a mo o po en ial o an ampli ude g ea e han 50 µV peak- o-peak in a leas
5 pulses o a se ies o 10 magne ic pulses. Fo i s de e mina ion, he elec odes we e placed
in he con ala e al hand (speci ically a he i s do sal in e osseous muscle) o he s imula ed
hemisphe e (le hemisphe e i he subjec was igh -handed and igh hemisphe e i he subjec
was le -handed) and he p ima y mo o co ex was s imula ed.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 37
Figu e 8. A) Righ hand wi h he i s do sal in e osseous muscle poin ed wi h an a ow. B) B ain wi h he
do sola e al p e on al co ex highligh ed in blue and poin ed wi h an a ow.
Once he RMT is de ined (as a pe cen age o he in ensi y) he LICI and SICI p o ocols a e
applied, along wi h he SHAM condi ion (ou o he scope o his p ojec ). The TMS-pulse is
igge ed be ween he F3 and F5 elec odes, which o e s a mo e p ecise s imula ion o he le
DLPFC and a low in e -subjec a iabili y in absence o MRI-guided neu ona iga ion
equipmen [42-43].
Figu e 9. Diag am o he TMS-coil placed be ween F3 and F5 elec odes o igge he pulse.
The da abase o he p ojec consis s on 16 pa ien s and 15 heal hy con ols. Six pa ien s had
had only one episode o schizoph enia ( i s -episode pa ien s) and he o he s we e ch onic
pa ien s. Demog aphic da a a e summa ized in he able below.

Pág. 38 Repo
Con ols
Pa ien s
Sex
Male
11
11
Female
18
12
Age
Mean
25.6
31.8
STD
10.5
10.5
Yea s o s udy
Mean
14.6
13.9
STD
2.2
3.1
RMT
Mean
64.7
66.7
STD
8.8
10.3
Table 1. Demog aphic in o ma ion o he da abase included on he s udy.
A he ime o he acquisi ion, only wo eco dings we e made: LICI and SICI. Fo each one,
he e we e single o pai ed-pulses igge ed andomly h oughou he eco ding. Thus, he
dis inc ion be ween PP and SP was done la e in he p e-p ocessing. A e his di e en ia ion,
ou di e en eco dings we e a ailable o each subjec : wo single-pulse TMS-EEG signals
(no ed as LICI SP o SICI SP, e en hough bo h o hem a e he same), pai ed-pulse LICI
TMS-EEG signal (LICI PP) and pai ed-pulse SICI TMS-EEG signal (SICI PP).
5.2. Algo i hm/Pipeline
The oolbox used h oughou his pipeline is FieldT ip, a collabo a i e so wa e ool based on
MATLAB, which holds unc ions o he p e-p ocessing and p ocessing o TMS-EEG signals.
One o he good poin s o Field ip is ha ( he) da a is o ganised in a s uc u e, con aining
di e en ields such as he sample equency, he labels o he channels, he ials, e c.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 39
Figu e 10. Example o how he da a is o ganised.
Roughly, he algo i hm ollows mo e o less he s uc u e desc ibed in Sec ion 4.3.2.2 (Typical
pipeline o TMS-EEG acquisi ion and p e-p ocessing), which is isually shown below.
Figu e 11. Flowcha o he TMS-EEG signal p ocessing.
Pág. 40 Repo
5.2.1. P e-p ocessing
5.2.1.1. Epoching (de ine ials)
The i s s ep, once he signal is loaded, is o de ine he ials. Raw da a is ei he one ollowing
he LICI o he SICI p o ocol and i includes bo h single-pulse and pai ed-pulse TMS-EEG
signals, as commen ed abo e.
In each eco ding he e a e 150 ials o epochs (coun ing bo h SP and PP). Thus, he e a e
225 pulses o TMS h oughou he signal (75 ials o SP plus 75 ials o PP, which a e 150
pulses o PP). Fo he pu pose o de ining he a o emen ioned ials, he name o he e en
( he igge o he pulse) has o be ound, which in his case is ma ked wi h he label “S15”.
Then, he da a be ween 1 second be o e and a e he e en is selec ed.
Once i is loca ed, i is necessa y o dis inguish be ween PP o SP, aking in o accoun he
dis ance be ween wo consecu i e e en s, he ISI. I he ime be ween he wo pulses is less
han 0.4 seconds, i is conside ed pai ed-pulse, o he wise i is conside ed single pulse and
labelled as such (0 o single-pulse and 1 o pai ed-pulse). Finally, each ial has a leng h o 2
seconds o da a (50,000 samples) wi h he TMS pulse(s) cen ed.
Figu e 12. Raw TMS-EEG signal o he LICI PP pa adigm.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 41
Figu e 13. Raw TMS-EEG signal o he SICI PP pa adigm. Le : en i e signal. Righ : Zoom in o he signal om -
0.06s o 0.06 s.
Figu e 14. Raw TMS-EEG signal o he LICI SP pa adigm.
5.2.1.2. A e ac exclusion: emo al o inging a e ac
The ollowing s ep is o dele e he pa o he signal ha holds he TMS pulse. Depending on
he TMS p o ocol and ype o pulse (SP, LICI PP o SICI PP), di e en amoun o da a is
Pág. 48 Repo
Figu e 23. Example o bad ial ha had o be emo ed be o e ICA applica ion.
In h ee di e en subjec s, he TMS machine did no wo k p ope ly in one ial o each subjec ,
igge ing mo e han one o wo pulses, depending on he ype o pulse signal. These ials
we e de ec ed and emo e be o e ICA applica ion.
Figu e 24. Example o a TMS-EEG signal wi h se e al TMS pulses due o a mal unc ion o he TMS machine.
Finally, in one subjec , a channel was pa icula ly noisy in all he ials, a ec ing oo many ICs.
The channel was de ec ed and dele ed be o e he applica ion o ICA.

Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 49
Figu e 25. Signals o he bad channel om di e en ials o he same subjec .
The e o e, ICA applica ion is no only use ul (and needed) o selec he componen s necessa y
o econs uc he signal, bu also i can be a e ision p ocess o hese ini ial s eps o he p e-
p ocessing. TMS-EEG is a no el ield o esea ch and hei p e-p ocessing guidelines a e no
comple ely desc ibed. The e o e, in his momen a semi-au oma ic p ocedu e would be he
bes op ion.
5.2.1.6. Signal econs uc ion
As soon as he componen selec ion is inished, he nex s ep is o econs uc he signal wi h
he ICs ha hold b ain signal in o ma ion. To do so, ano he FieldT ip unc ion is used, in which
he use only ha e o in oduce he index o he componen s o no include in he econs uc ion.
Then, he signal is band-pass il e ed be ween 0.5 Hz and 70 Hz and downsampled o 5000
Hz in o de o educe compu a ional ime in he ollowing s eps.
5.2.1.7. Bad channels and bad ials emo al
Now, he signal is clean enough o decide which ials o keep. Di e en common algo i hms
o ial ejec ion ha e been implemen ed in o de o de e mine which one wo ks be e . Fo he
pu pose o con i ming ha he au oma ic selec ion ( he algo i hm) wo ks p ope ly, a manual
ial selec ion was pe o med o a ew eco dings. This selec ion was, once again, made by
h ee expe ience esea che s. The ials ha we e sha ed by a leas wo o hem we e
classi ied as bad. Mo eo e , he e was a ‘doub s’ sec ion o each esea che . I he ial
appea ed in wo ‘doub s’ sec ion and he o he esea che ma ked i as bad, he ial was
classi ied as bad oo.
Pág. 50 Repo
E en hough his s ep is ypically easily implemen ed, he signals p ocessed a e no ypical-
EEG as he as majo i y o a e ac s has been emo ed by ICA. Nex , i includes a desc ip ion
o he pe o med p ocess.
Fi s , a magni ude h eshold was implemen ed, aking in o accoun he mean and he s anda d
de ia ion. None heless, a lo o “good” ials we e selec ed as bad, since he h eshold had
e y low magni ude alues.
Consul ing he li e a u e, he z-sco e used in Wu e al., [25] was ep oduced. Two di e en
h esholds o his measu e we e implemen ed: 2.5 and 3 alue o z-sco e. The numbe o ials
de ec as bad was educed, howe e hey did no ma ch wi h he ones de e mined manually.
Then, he en opy o he signal was used as a pa ame e , ollowing he code de eloped by
Tos e al. [45]. Ye again, he ials selec ed as bas we e no he same as he ones de ec ed
manually.
Las ly, he join p obabili y and he ku osis, bo h being s a is ical measu es, we e used o
c ea e a h eshold. E en hough some o he ials selec ed manually we e included, he
algo i hms did no ma ch he expec a ions.
Gi en he bad esul s, a new s a egy was de ined and implemen ed.
5.2.1.7.1 Bad channels de ec ion
Fi s o all, i was obse ed ha he ac i i y o some channels we e noisie han he es o he
channels. Then, i was decided o selec p e iously he bad channels. To do so, he ku osis
o each channel in each ial is compu ed, ob aining a ma ix wi h a size Nchannels x N ials.
Then, his ma ix is so ed om he lowes o he g ea es alue. O all he da a, jus he 80%
o i is used o compu e a h eshold ee o ou lie s. The e o e he 10% o he lowes alues
and he 10% o he g ea e alues we e no used o he h eshold calcula ion.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 51
Figu e 26. Simple diag am showing he selec ion o da a o he h eshold compu a ion.
A h eshold o each ial is calcula ed wi h he ollowing equa ion, aken in o accoun he mean
and he s anda d de ia ion. This h eshold o a speci ic ial indica es he maximum alue
“allowed” o each channel in his ial.
𝑡ℎ𝑡𝑟𝑖𝑎𝑙 = 𝑚𝑒𝑎𝑛(𝑣𝑎𝑙𝑢𝑒𝑠_𝑘𝑢𝑟𝑡𝑜𝑠𝑖𝑠𝑐ℎ,𝑡𝑟𝑖𝑎𝑙𝑠) + 5 ∗ 𝑠𝑡𝑑(𝑣𝑎𝑙𝑢𝑒𝑠_𝑘𝑢𝑟𝑡𝑜𝑠𝑖𝑠𝑐ℎ,𝑡𝑟𝑖𝑎𝑙𝑠)
Equa ion 1
None heless, i a channel has a g ea alue o a g ea numbe o ials, i pe haps would no
be de ec ed as bad (because he h eshold o hose ials would be g ea as well; al hough he
g ea es alues ha e al eady been emo ed by using jus he 80% o he da a). To a oid his
om happening, a common h eshold is compu ed by calcula ing he median o all he
h esholds, ob aining one h eshold o subjec .
𝑡ℎ𝑐𝑜𝑚𝑚𝑜𝑛 = 𝑚𝑒𝑑𝑖𝑎𝑛(𝑡ℎ𝑡𝑟𝑖𝑎𝑙)
Equa ion 2
Finally, each alue o ku osis (each channel o each ial) is compa ed o he common
h eshold. I i is highe han he h eshold, he channel co esponding o he alue is classi ied
as bad. I he channel is classi ied as bad in mo e han 25% o he ials, he channel is dele ed
and in e pola ed.
Fo he in e pola ion, is impo an o bea in mind ha he neighbou s’ channels ha e o be
good o o he wise di e en a e ac s will be included in he new in e pola ed channel. Thus,
once he bad channels and hei neighbou channels a e de ec ed, i a neighbou channel is
also a bad channel, he channel is dele ed and illed wi h NaNs. Then, all he bad channels a e
in e pola ed.
As an example, in he igu e below a ial wi h 5 bad channels is displayed (blue, o ange,
pu ple, blue and magen a ones o subplo s ‘Channels 1 o 10’, ‘Channels 11 o 20’, ‘Channels
Pág. 52 Repo
31 o 40’ and ‘Channels 51 o 62’, espec i ely). The in e pola ed channels a e shown in
Figu e 28.
Figu e 27. T ial wi h bad channels.
Figu e 28. T ial wi h bad channels in e pola ed.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 53
5.2.1.7.2 Bad ials de ec ion
Once he da a has no bad channels, he ial ejec ion is pe o med. He e, wo di e en
h esholds ha e been employed: a maximum h eshold and a ange h eshold.
Bo h o hem conside he magni ude o he signal. Following he same p ocedu e as he bad
channel de ec ion, he 80% o he da a is used. The e o e, he mean and he s anda d de ia ion
o each channel o each ial is compu ed and so ed, and hen, he 80% o hem in he middle
a e selec ed. Las ly, he wo di e en h esholds a e compu ed, ollowing he equa ion below:
𝑡ℎ_𝑏𝑎𝑑_𝑡𝑟𝑖𝑎𝑙𝑐ℎ = 𝑚𝑒𝑎𝑛(𝑚𝑒𝑎𝑛_𝑣𝑎𝑙𝑢𝑒𝑠𝑐ℎ,𝑡𝑟𝑖𝑎𝑙𝑠) + 𝑘 ∗ 𝑚𝑒𝑎𝑛(𝑠𝑡𝑑𝑣𝑎𝑙𝑢𝑒𝑠𝑐ℎ,𝑡𝑟𝑖𝑎𝑙𝑠)
Equa ion 3
Fo he maximum h eshold, he k akes a alue o 7 and o he ange h eshold, a alue o 5.
When one channel in he ial su passes, in absolu e alue, he maximum h eshold, he ial
is selec ed as ‘bad ial’. Ne e heless, i a channel has a alue be ween he ange h eshold
and maximum h eshold, he ial is selec ed as ‘possible bad ial’. Subsequen ly, hese
channels ( he ones ha a e he esponsible o he ‘possible bad ial’ label) a e coun ed o
each ial. I a ial has mo e han 10% o he channels (6 channels) in his ange (be ween
maximum and ange h esholds), he ial is inally labelled as ‘bad’. All he ials selec ed by
his algo i hm ma ch wi h he ones selec ed manually.
Subjec
Numbe o channels
ha su pass he
maximum h eshold
(bad ial)
Numbe o channels
ha a e in be ween
he ange h eshold
and he maximum
(possible bad)
T ial ejec ed?
1
1
0
Yes
2
0
5
No
3
2
4
Yes
4
0
13
Yes
Table 2. Example o bad ials and possible bad ials o di e en subjec s and i s decision whe he o no he ial
is inally ejec ed.
Finally, he ‘bad’ ials a e emo ed om he da a and he signals a e comple ely clean. Some
examples o bad and good ials a e displayed below.

Pág. 54 Repo
Figu e 29. Example o bad ial.
Figu e 30. Example o bad ial.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 55
Figu e 31. Example o good ial.
Di e en clean signals o a ious channels a e shown in Sec ion 6.1.
5.2.2. P ocessing
5.2.2.1. Cha ac e iza ion
Once he signals a e clean, hey a e p ocessed in o de o cha ac e ize hem.
Since one o he objec i es is o iden i y he common TMS-e oked po en ial (TEP) peaks, i is
needed o a e age he signal ac oss ials and subjec s because hose a o emen ioned
de lec ions (TEPs) a e no easily seen in single ials o a e aging jus a ew.
Ne e heless, be o e doing he a e aging is impo an o adjus he leng h o he LICI PP signal
(which is he signal ha has he wo pa s o he ials conca ena ed ins ead o in e pola ed,
see 5.2.1.2) wi h he LICI SP signal, as a compa ison be ween pai ed-pulse and single-pulse
signals is going o be pe o med la e on.
Fo he pu pose o ha ing he same leng h in he signals om LICI p o ocol, he i s sample
ha was elimina ed in he p e-p ocessing is ound in he new signal, as he equency sample
has changed. Then, a ec o wi h NaNs ha has he same leng h as he samples dele ed in
Pág. 56 Repo
5.2.1.2 ( emo al o inging a e ac ) is inse ed jus a e he a o emen ioned sample (i.e, he
samples dele ed be o e a e subs i u ed by NaNs alues). Doing so, he LICI PP signal has an
equal leng h o LICI SP. Fu he mo e, bo h a e aligned in ime, meaning ha he signals ha e
he ime 0 a he same sample.
Howe e , he signals o he SICI PP and SICI SP signals a e no aligned in ime. The e is a
di e ence o 13 samples be ween he ime 0 o bo h signals. Thus, in o de o aligned hem,
13 samples a e emo ed a he begging o he end o he signal, depending on i s ype o pulse
(PP o SP, espec i ely).
Figu e 32. D awing o he p ocess o aligned he SICI SP and SICI PP signals.
When all signals ha e he same numbe o samples and a e aligned in ime, an a e aging
ac oss ials is done o each subjec , ob aining one signal o each subjec , p o ocol and ype
o pulse. As an example, subjec 1 will ha e 4 TMS-EEG signals, one o each p o ocol and
ype o pulse (LICI PP, LICI SP, SICI PP and SICI SP).
Then, a g and a e age is pe o med a e aging he signals ac oss subjec s. Finally, a unique
signal o each p o ocol and ype o pulse is ob ained and he common TEP peaks can be
ound. Tempo al windows whe e he peaks a e expec ed a e p ede ined so he peaks a e
sea ched on hose windows. Then, he a o emen ioned peaks a e isually inspec ed and a
na owe and mo e p ecise windows a e de ined o ob ain he peaks in he signals o e e y
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 57
subjec . In pa icula , wo di e en se o windows whe e de ined: one o he SICI p o ocol and
ano he o he LICI p o ocol aking in o accoun he SP signal.
Fo he pu pose o compa ing bo h ype o pulses, each SP signal is sub ac ed om i s
analogue PP.
Ano he measu e o compa e single-pulse and pai ed-pulse signals is compu ed. Following
he p ocedu e done in Cash e al. [46], a modula ion a io is compu ed. This consis s o he
di ision o he TEP’s ampli ude o he PP signal by he TEP’s ampli ude o he SP signal.
𝑚𝑜𝑑𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟𝑎𝑡𝑖𝑜 = 𝑇𝐸𝑃 𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 𝑇𝑀𝑆 − 𝐸𝐸𝐺 𝑃𝑃
𝑇𝐸𝑃 𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 𝑇𝑀𝑆 − 𝐸𝐸𝐺 𝑆𝑃
Equa ion 4
Mo e speci ically, he TEP ampli ude is he mean ampli ude o he signal in he ime windows
de ined a e he loca ion o he peaks in he g and a e aged signals.
TEP ampli ude we e ob ained o all channels, howe e , i could be in e es ed o g oup he
TEP ac i i y in a egion o in e es (ROI). In his wo k, i is p oposed o use a ROI ha includes
he channels on he le DLPFC co ex, which a e: FP1, AF7, AF3, F7, F5, F3, F1, FC5, FC3,
FC1 [46]. The ROI can be seen in Figu e 33.
Figu e 33. Channel dis ibu ion a ound he scalp wi h he ROI iden i ied.
The algo i hm de eloped can be ound in a Gi Hub eposi o y. The link o he eposi o y is in
Annex 2 o he Annex documen .
Pág. 64 Repo
Figu e 43. Signal a e aged o all he subjec s o LICI PP wi h he TEPs ma ked wi h e ical lines.
One o he objec i es o he p ojec was o loca e he TEPs o he di e en p o ocols (SICI
and LICI) and ype o pulses (SP and PP). Some main TEP componen s ha e been de ined
in he li e a u e: P25, N40, P60, N100, P185 o P200 [17, 19, 20, 36]. Table 3 depic s he
ime la ency ound o each signal (in ms):
P o ocol Name
o he peak
P25
N30
P50
N100
P140
N160
P200
SICI SP
24.6
32
50
101.8
142.2
159.2
201.4
SICI PP
18.6
27
39
96.8
123.4
151
207
LICI SP
24.2
34.6
48
99.4
137.8
163
211.2
LICI PP
-
-
46
71.6
110.6
169.2
214.6
Table 3. La encies o he peaks ound o each p o ocol and ype o pulses.
The peaks ound a e: P25, N30, P50, N100, P140, N160 and P200, which ela i ely ma ch
wi h he peaks desc ibed in he li e a u e (P25, N40, P60, N100, P185 o P200, depending on
he a icle). Howe e , he N40 and P60 peaks appea a ew milliseconds be o e. Mo eo e , in
some e e ences, he las 3 peaks (P140, N160 and P200) a e e alua ed oge he , which will

Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 65
be P185. None heless, in his p ojec hey ha e been conside ed sepa a ely since he
di e ences in ampli ude a e no iceable.
Compa ing bo h ypes o pulses o each p o ocol, i seems ha he PP signal is equi alen o
he SP signal displaced o he le , as can be seen in bo h he igu es and Table 3, obse ing
he la encies be ween ypes o pulses o each p o ocol. This beha iou may be due o he ac
ha he ce eb al co ex is mo e exci ed o he pai ed-pulse pa adigm. In PP signals, he i s
pulse exci es he co ex, he TMS-e oked po en ials s a o o m and hen a second pulse is
igge ed, exci ing e en mo e he co ex and ad ancing he a o emen ioned po en ials.
Following his ain o hough s, i is no mal ha o he LICI PP signal he appea ance o he
TEPs a e e en ea lie .
PP esponse is a composi ion o he neu al esponse o he i s TMS-pulse and he second
one. In SICI p o ocol, he ISI is low (4 ms), hence as Figu e 42 shows, he TEP componen s
a e e y simila o SICI SP ones. Ne e heless, in LICI p o ocol, he in e s imulus in e al is
highe (100 ms) and hen, he neu al esponse could be iewed as a supe posi ion o he e ec
o bo h TMS pulses. In o de o a oid his e ec and o s udy only he in luence o he i s pulse
(CS; he SP signal would be jus TS, es s imulus), P emoli e al.,[47] co ec ed he PP signal
by sub ac ing he SP signal om he PP signal aligned o he i s pulse (CS). Thus, he PP
signal has a ime equal o 0 in he i s pulse and no in he second one, which is he case o
his p ojec . Howe e , Cash e al., [46] decided no o do he sub ac ion since hey s a e ha
changes in he ampli ude o TEPs componen s a e un ela ed o he in luence o CS alone, and
his is wha has been ollowed in he p ojec . Fu u e wo k will assess he p ocedu e de eloped
by P emoli e al. [47].
6.2.2. Compa ison be ween ype o subjec s and pulse ype
In he p e ious sec ion, a g and a e age was pe o med o all subjec s. Howe e , he
da abase con ains neu ological da a o m SCZ pa ien s and om heal hy con ols (HC). Figu e
44 and Figu e 45 show he compa a i e be ween he g and a e age o he subjec s ha belong
o each g oup.
Pág. 66 Repo
Figu e 44. G and a e age o bo h ype o subjec s o SICI p o ocol and bo h ype o pulses (SP and PP).
Figu e 45. G and a e age o bo h ype o subjec s o LICI p o ocol and bo h ype o pulses (SP and PP).
Obse ing he Figu e 44, i can be seen ha he mo phology o di e en pulses o SICI p o ocol
is qui e simila o he same g oup o subjec s. Thus, he SICI SP signal is simila o he SICI
PP signal o boh, schizoph enic pa ien s as well as he same signals o heal hy con ols.
Besides, he signals o each g oup a e qui e simila be ween hem. None heless, he P140
and he N160 peaks a e mo e p onounced in he con ol g oup, aking mo e posi i e alues
o P140 TEP and less posi i e o N160 TEP.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 67
Rega ding he igu e o LICI p o ocol, he TEPs o he PP pulse in bo h g oups appea ea lie ,
as has been s a ed in he p e ious esul s. He e again, he N160 peak is no e y p onounced
in he LICI SP signal o pa ien s and he di e ence o ampli ude be ween peaks is no huge.
Howe e , he ampli ude o he TMS-e oked po en ials o he PP signal is lowe han he
ampli ude o he SP signal o he con ols. Equally o SICI signals, he ob ained TEPs o LICI
PP and LICI SP a e b oadly simila be ween diagnos ic g oups.
Acco ding o he li e a u e, he ampli ude o he TEPs ollowing he pai ed-pulse should be
lowe han he ampli ude o he TEPs ha ollows a single-pulse ( o each p o ocol and in a
heal hy subjec ). This is expec ed because he i s pulse o he pai ed-pulse ac i a es
inhibi o y- ype ecep o s (conc e ely, GABA-A ecep o s o SICI p o ocol and GABA-B o LICI
p o ocol). Howe e , as he inhibi o y mechanisms a e al e ed in SCZ, wha is expec ed o
pa ien s is a less di e ence o ampli udes o TEPs. This has been p o en o he SICI p o ocol
bu he e a e dispa a e esul s o LICI [20, 46, 47, 48].
To sum up, wha is expec ed o each p o ocol is a simila mo phology o he signals o each
g oup bu wi h di e en magni ude be ween PP and SP; wi h g ea e di e ences in he con ol
g oup han in he pa ien g oup since he inhibi o y mechanisms a e well ac i a ed o he
o me . In hese esul s, he di e ence in ampli ude aking in o accoun bo h ypes o pulses
a e clea e seen in LICI han in SICI since o he la e he only di e ence in magni ude is in
he ea lie TEPs (P50).
In he nex sec ion, he a o emen ioned di e ence be ween SP and PP pulses a e deelpy
analysed.
6.3. Single-pulse signal s pai ed-pulse signal
6.3.1. Tempo al signal
To assess he inhibi ion esponse on di e en subjec s, i is common o compu e he di e ence
be ween SP and PP. Figu e 46 and Figu e 47 show he sub ac ion o SP da a minus PP da a
o bo h p o ocols (SICI and LICI) and o bo h diagnos ics (SCZ pa ien s and HC).
Pág. 68 Repo
Figu e 46. Signals ob ained o sub ac ing SICI PP signal o SICI SP signal o con ols and pa ien s.
Figu e 47. Signals ob ained o sub ac ing LICI PP signal o LICI SP signal o con ols and pa ien s.
Obse ing Figu e 46, i is seen ha he ampli ude o bo h signals, one o he HC and o he o
SCZ pa ien s, is close o 0, meaning ha he e a e ew di e ences be ween single-pulse and
pai ed-pulse SICI. This, in e u n means ha he le el o inhibi ion is mo e o less he same o
bo h ype o pulses. Ne e heless, HC g oup shows g ea e a iabili y a he ini ial esponse
( ew milliseconds a e he s imulus). In his case, con ols’ di e ence akes alues a he om
0, which indica es ha he e is a change in ampli ude, in he milliseconds close o he s imulus
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 69
(as s a ed in he p e ious sec ion). I , in e u n, indica es mo e di e ences o inhibi ion be ween
bo h ypes o pulses in he TEPs nea e o he s imulus. Mo eo e , he inhibi ion o his TEP
(P50) is g ea e in he HC han in he SCZ pa ien s.
Rega ding LICI p o ocol (Figu e 47), he di e ence o he le el o inhibi ion be ween SP and
PP is la ge han SICI Wha is expec ed is o ob ain posi i e alues since, heo e ically, SP
should ha e a g ea e ampli ude han PP. This posi i eness is obse ed a ound 150 ms o 250
ms a e he pulse, meaning ha he inhibi ion applying a pai ed-pulse is g ea e han he one
ob ained a e a single-pulse. Howe e , bo h signals (HC and SCZ pa ien s) ake simila
alues, so he inhibi ion is mo e o less he same o bo h g oups.
6.3.2. Spa ial dis ibu ion o TEPs
In o de o compa e he TEP componen s a inie ime windows we e implemen ed o de ec
he a e aged TEP in he de ined ime window. The a e age signal o e e y ype o subjec ,
heal hy con ols and schizoph enic pa ien s, we e pe o med.
These ime windows we e de ined based on he signals o SICI SP and LICI SP, which a e he
e e ences.
Figu e 48. Signal a e aged o all subjec s o SICI SP wi h he ime windows ma ked. The exac alues o he
windows a e he ollowing. Da k blue: 19-29 ms. Da k g een: 28-37 ms. Red: 40-60 ms. Pu ple: 88-112 ms.
Yellow: 120-150 ms. Black: 185-226 ms. Ligh blue: 120-226 ms. Pink: 150-183 ms.

Pág. 70 Repo
Figu e 49. Signal a e aged o all subjec s o LICI SP wi h he ime windows ma ked. The exac alues o he
windows a e he ollowing. Da k blue: 19-29 ms. Da k g een: 29-38 ms. Red: 40-60 ms. Pu ple: 86-110 ms.
Yellow: 124-150 ms. Black: 188-227 ms. Ligh blue: 124-227 ms. Pink: 150-176 ms.
To see he ac i a ion di e ences in he scalp, opoplo s o he signal ob ained om he
sub ac ion on he second ime windows a e displayed.
Figu e 50. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he i s ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he i s ime window.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 71
Figu e 51. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he second ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he second ime window.
Figu e 52. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he hi d ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he hi d ime window.
Pág. 72 Repo
Figu e 53. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he ou h ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he ou h ime window.
Figu e 54. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he i h ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he i h ime window.
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 73
Figu e 55. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he six h ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he six h ime window.
Figu e 56. A) Topoplo and empo al signal o bo h SP-PP SICI and SP-PP LICI o he se en h ime window. B)
Tempo al signal o SICI SP and LICI SP o bo h g oup o subjec s o he se en h ime window.
Pág. 80 Repo
10. ECONOMICAL ANALYSIS
In his chap e , he budge o he p ojec is de ailed below.
Pe sonal cos s
Task
Hou s
Cos (€/h)
To al (€)
S uden
Resea ch
50
25
1250
P og amming
600
25
15000
W i ing
120
25
3000
Mee ings
30
25
750
Di ec o
Wo k e iew
100
40
4000
Mee ings
30
40
1200
Co-di ec o
Wo k e iew
30
40
1200
Mee ings
10
40
400
BIOART P o esso s
Mee ings
8
40
320
SUCEDE clinical
collabo a o s
Mee ings
8
40
320
TOTAL
27440
Table 5. S a budge .

Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 81
So wa e cos s
Uni s
To al (€)
MATLAB R2021b (S uden
license)
1
35
FieldT ip
1
0
O ice 365 Home
1
69
TOTAL
104
Table 6. So wa e budge .
Elec ici y cos s
Mon hs
Cos (€/mon h)
To al (€)
Elec ici y
8
80
640
Table 7. Elec ici y budge .
Budge
Cos (€)
Pe sonal Cos s
27440
So wa e Cos s
104
Elec ici y Cos s
640
TOTAL
28184
Table 8. Budge 's summa y wi h he o al sum.
Table 9. Budge 's summa y wi h o al sum.
Pág. 82 Repo
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 83
CONCLUSIONS
The main objec i e o his p ojec is he de elopmen o a semi-au oma ic algo i hm ha
enables he possibili y o p e-p ocess aw TMS-EEG signals ⎯ob aining a su icien ly clean
signal o i s u he p ocessing⎯ and o cha ac e ize he signal by iden i ying he common
TEPs p esen in he li e a u e. Then, a compa ison be ween ype o pulses (single o pai ed)
and be ween subjec s (schizoph enic pa ien s and heal hy con ols) has been made aken in o
accoun he a o emen ioned TEPs.
• P e-p ocessing design:
Rega ding he p e-p ocessing s age, i has ul illed he expec a ions; de eloping an
algo i hm ha has as ou pu a good clean signal. The cu en p oposal could be conside ed
as semi-au oma ic, since some s eps equi e he esea che ’s in e en ion. In pa icula ,
he manual selec ion o he independen componen s ( he ou pu o he ICA applica ion)
akes up a lo o ime. This is due o he ac ha he opinion o he h ee expe s has been
aken in o accoun , hence he o al ime equi ed o his includes he ime needed o an
indi idual selec ion and he ime o a common selec ion.
Ano he unexpec ed issue du ing he de elopmen o he p e-p ocessing algo i hm has
been he ial and channel ejec ion. Ini ially, we s a ed using a semi-au oma ic p ocedu e.
Ne e heless, he inal algo i hm allows an au oma ic ejec ion o bo h, a e ac ua ed
channels and ials, wi hou any use in e en ion.
Taking all his in o accoun , i can be concluded ha he objec i e o ha ing a good cleaning
o he da a, and hence, he aw TMS-EEG signal, has been achie ed; as he inal signals
(be o e he cha ac e iza ion) we e no noisy no a e ac ua ed.
• Cha ac e iza ion o TEP da a:
Conce ning he p ocess and cha ac e iza ion o he TMS-EEG signals, he di e en TMS-
e oked po en ials ha e been co ec ly iden i ied. As esul s sec ion shows, he TEP
componen s ob ained ma ch hose ound in p e ious s udies.
Fo he pu pose o compa ing he di e ences caused by he ype o pulses o each
p o ocol, he PP signal has been sub ac ed om he SP signal. Fo he SICI p o ocol, he
esul s a e somewha di e en om hose expec ed, as he e we e jus sligh di e ences
in ampli ude be ween bo h pulses. On he o he hand, o he LICI p o ocol, he expec ed
Pág. 84 Repo
esul s ha e been ob ained, seeing a smalle ampli ude o he TEPs in PP signal han in
SP. The same has been e i ied wi h he opog aphical igu es, as well as wi h he
modula ion a ios.
Rega ding he diagnos ic o he subjec s included in he s udy, as i could be expec ed
heal hy con ols showed mo e inhibi ion han schizoph enic pa ien s. This was obse ed
o some o he TEPs componen s and i i s wi h he hypo hesis ha es ablishes an
inhibi ion dys unc ion ela ed wi h he schizoph enia diso de .
Finally, a spa ial analysis was pe o med o localise he dis ibu ion o TEPs. The TMS-
e oked po en ials ha e been loca ed and he compa ison be ween ype o pulses and
subjec s has been pe o med. In his case, he compa ison wi h p e ious esul s is mo e
di icul since he e a e ew s udies ha ha e included opog aphical esul s. Al hough
some o he esul s we e no exac ly he expec ed, i may be because o he sample size
since hey show a endency owa d he esul s ob ained in o he s udies. Besides, wi h he
con inuous g ow h o he da abase (new TMS-EEG signals a e being p ocessed), i is
expec ed ha he esul s will mee hose o he li e a u e.
The p e-p ocessing and p ocessing o TMS-EEG signals is s ill a poo ly known ield and i
equi es g ea con ol o he di e en s eps o he da a analysis, such as il e ing and epoching
among o he s. In pa icula , a good a e ac emo al and a p ecise selec ion o he independen
componen s since a bad one migh dele e c ucial da a o he econs uc ed signal migh be oo
noisy. In his p ojec , his en i e p ocedu e has been add essed s a ing om aw da a and
ending up ob aining a good cha ac e iza ion o i . Th oughou he p ocess, e e y s ep has been
e ised so i was obus o all he da a. The e o e, he esul o he p ojec is a obus algo i hm
ha will be e y use ul o i s applica ion in new subjec s o o he s TMS-EEG da abases.
Finally, o m a pe sonal and academic poin o iew, wi h he de elopmen o he p ojec I ha e
been able o e ine and widen my p e ious knowledge o signal p ocessing. I unde s ood be e
he di e en a e ac s one migh encoun e in a TMS-EEG signal as well as I gained mo e
expe ience in he in e p e a ion o a ious plo s and igu es. Mo eo e , my skills wi h MATLAB
p og amming ha e imp o ed and I lea n how o use FieldT ip. Thus, he de elopmen o he
p ojec has en iched me as a esea che .
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 85

Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pàg. 87
Acknowledgemen s
Fi s o all, I would like o hank e y much all he help, he guidance and he suppo I ecei ed
om he di ec o o his p ojec , D . Alejand o Bachille Ma a anz, as well as he one ecei ed
om he co-di ec o , D . Joan F ancesc Alonso López. I also would like o ex end ha
app ecia ion o all he membe s o he BIOART g oup, who has gi en me ad ice o di e en
pa s o he p ojec .
This p ojec would ha e no been possible wi hou he help and he clinical iew o Inés
Fe nández and Vicen e Molina, s a o he Hosp ial Clínico Uni e si a io de Valladolid and
membe s o he SUCEDE g oup.
Finally, I wan o hank my amily and iends, wi h whom I ha e been able o disconnec om
he s ess o wo k and uni e si y.
Pág. 88 Repo
Imp o ing he quali y o combined EEG-TMS neu al eco dings: a i ac emo al and ime analysis Pág. 89
Bibliog aphy
Re e ences
[1] P. J. Ha ison, “Neu opa hology o schizoph enia,” Psychia y, ol. 7, no. 10, pp. 421–
424, 2008.
[2] H. Edemann-Callesen, C. Win e , and R. Hada , “Using co ical non-in asi e
neu omodula ion as a po en ial p e en i e ea men in schizoph enia - A e iew,” B ain
S imul., ol. 14, no. 3, pp. 643–651, 2021.
[3] A. Schmi , A. Hasan, O. G ube , and P. Falkai, “Schizoph enia as a diso de o
disconnec i i y,” Eu . A ch. Psychia y Clin. Neu osci., ol. 261, no. SUPPL. 2, pp. 150–
154, 2011.
[4] S. I i ani, “Neu opa hology o schizoph enia: A mini e iew,” Neu opa hology, ol. 27,
no. 6, pp. 604–608, 2007.
[5] A. D. Boes, M. S. Kelly, N. T. T app, A. P. S e n, D. Z. P ess, and A. Pascual-Leone,
“Nonin asi e b ain s imula ion: Challenges and oppo uni ies o a new clinical
special y,” J. Neu opsychia y Clin. Neu osci., ol. 30, no. 3, pp. 173–179, 2018.
[6] L. Gomez Fe nandez, “Es imulación ce eb al no in asi a en las en e medades
neu ológicas y psiquiá icas,” Re . Cuba. Neu ol. y Neu oci ., ol. 8, no. 2, pp. 1–17,
2019.
[7] S. L. Liew, E. San a necchi, E. R. Buch, and L. G. Cohen, “Non-in asi e b ain
s imula ion in neu o ehabili a ion: Local and dis an e ec s o mo o eco e y,” F on .
Hum. Neu osci., ol. 8, no. JUNE, pp. 1–15, 2014.
[8] T. Wagne , A. Vale o-Cab e, and A. Pascual-Leone, “Nonin asi e human b ain
s imula ion,” Annu. Re . Biomed. Eng., ol. 9, pp. 527–565, 2007.
[9] J. Vosskuhl, D. S übe , and C. S. He mann, “Non-in asi e B ain S imula ion: A
Pa adigm Shi in Unde s anding B ain Oscilla ions,” F on . Hum. Neu osci., ol. 12, no.
May, pp. 1–19, 2018.
[10] J. D. K op o , “T ansc anial Magne ic S imula ion,” in Func ional Nue oma ke s o
Psychia y, J. D. K op o , Ed. Academic P ess, 2016, pp. 281–283.
[11] R. J. Ilmoniemi and D. Kičić, “Me hodology o combined TMS and EEG,” B ain Topog .,
ol. 22, no. 4, pp. 233–248, 2010.
[12] A. Pascual-Leone and J. M. To mos-Muñoz, “Es imulación magné ica ansc aneal:
undamen os y po encial de modulación de edes neu onales especí icas,” Re .
Neu ol., ol. 46 Suppl 1, pp. S3–S10, 2008.
[13] J. C. Cole, C. G. Be nacki, A. Helme , N. Pinnin i, and J. P. O’Rea don, “E icacy o