Etablissements délivrant la formation :

Cogmaster – Université Paris Descartes, département de Psychologie

Ecole Normale Supérieure de Cachan, département de Biologie

Rapport de stage de Master 2

Sciences cognitives, spécialité Neurosciences

Effects of high-frequency motor cortex repetitive

transcranial magnetic stimulation (rTMS) on hand

representation in primary somatosensory cortex.

A source modelling study of somatosensory evoked

potentials.

Par Claire Bradley,

sous la direction du Dr Garcia-Larrea,

et du Dr Magnin.

Laboratoire d'accueil :

Intégration centrale de la douleur (NeuroPain),

Inserm U1028; Lyon Neuroscience Research Center,

Lyon, F-69000, France

1

Table of ContentsIntroduction..........................................................................................................................................3Materials and methods..........................................................................................................................5

What methods and why ? ................................................................................................................5Subjects............................................................................................................................................7Experiment time-course...................................................................................................................7Spatially guided rTMS.....................................................................................................................8Somatosensory evoked potentials (SEP).........................................................................................8

Optimising the methods : choice criteria....................................................................................8SEP recording..............................................................................................................................9SEP analysis..............................................................................................................................10Base-line correction and physiological constraints...................................................................11SEP source reconstruction.........................................................................................................11

Coordinates, measures and projection on anatomical image.........................................................12Sham stimulation...........................................................................................................................12Statistical analysis..........................................................................................................................12

Results................................................................................................................................................13A. Searching for the best possible parameters for source modelling.............................................13

I. Data pre-processing..............................................................................................................13 II. Type of model......................................................................................................................14 III. Analysis time-window .......................................................................................................15

B. Source modelling of SEPs.........................................................................................................16 I. Estimating hand cortical representation before rTMS..........................................................16 II. Changes in source localisation after rTMS.........................................................................18 III. Source modelling of PES during rTMS.............................................................................20

Discussion...........................................................................................................................................23Modelling methods........................................................................................................................23Mechanisms of rTMS-evoked plasticity........................................................................................25Functional relevance of S1 plasticity.............................................................................................26Lack of on-line assessment of ongoing plasticity..........................................................................28

Conclusion..........................................................................................................................................30References..........................................................................................................................................31

2

Introduction

Repetitive transcranial magnetic stimulation (rTMS) is a powerful tool that allows non-

invasive stimulation of the cortex. It has been shown to have physiological effects (modulation of

cortical excitability, oscillatory activity and of the network of brain regions involved in a specific

task), sometimes -but not always- associated to behavioural consequences (see Ziemann et al., 2008

for a review on motor cortex rTMS-evoked plasticity). Since its after-effects can outlast the

stimulation itself (Siebner & Rothwell, 2003; Thut & Pascual-Leone, 2010), rTMS is being actively

tested as a potential non-invasive therapeutic tool for neurological and psychiatric disorders

(Ridding & Rothwell, 2007). In the treatment of pharmaco-resistant neuropathic pain (i.e. pain

resulting from a lesion or a disease of the somatosensory system, Treede et al., 2008), rTMS might

become a non-invasive alternative to motor cortex epidural stimulation (MCS). Although its effects

are weaker and of shorter duration than those of MCS, rTMS of the motor cortex M1 has

demonstrated a significant potential for pain relief (André-Obadia et al., 2006; André-Obadia et al.,

2008; Hosomi et al., 2008; Lefaucheur, 2006; Lefaucheur, 2011; Leo & Latif, 2007). This effect is

reached only when rTMS is delivered to the motor cortex at high frequencies (5 to 20 Hz), rather

than low frequencies (0.1 to 1 Hz) (André-Obadia et al., 2006; Lefaucheur, 2011).

How high-frequency rTMS pain relief is brought about remains speculative. In any rTMS

protocol, stimulation is directed at one given cortical area. However, there is evidence that the

effects can spread to other connected brain regions and affect distant sites (Bestmann et al., 2003;

Gerschlager et al., 2001; Münchau et al., 2002; Sack et al., 2007). Of the many areas connected to

the motor cortex (Fox et al., 1997), somatosensory cortex S1 has been shown to display somatotopic

plasticity that correlates with neuropathic pain development (Flor et al., 1995; Wrigley et al., 2009).

This “maladaptive plasticity” seems to be reversed by successful treatment of the pain (Birbaumer

et al., 1997; Huse et al., 2001; Lotze et al., 1999; Napadow et al., 2007). Effective treatment of

neuropathic pain by rTMS of the motor cortex might therefore be correlated to S1 map plasticity.

This hypothesis is further supported by studies showing that rTMS can indeed induce changes in

cortical sensori-motor representations (Lee et al., 2003; Siebner & Rothwell, 2003; Tegenthoff et

al., 2005). However, somatotopic reorganisation of S1 following high-frequency motor cortex rTMS

has not yet been formally explored. The main aim of this study is to investigate this question in

healthy subjects, as a previous step to a more comprehensive study addressed to patients with drug-

resistant neuropathic pain.

As current rTMS protocols are approaching a limit, especially in their clinical applications,

more and more efforts are allocated to increasing both size and duration of effects (Thut & Pascual-

3

Leone, 2010). Among the possible parameters to be modified, duration of stimulation, number of

TMS pulses and their timing have been investigated. Unfortunately, the relationship between these

temporal parameters and the actual after-effects is far from simple. For example, Gamboa et al.,

(2010) showed that the after-effects of theta-burst stimulation (a variation of classical rTMS

protocols) could be reversed just by doubling the duration of stimulation. This suggests that at some

point during the stimulation, a boundary was reached, that determined the direction of after-effects

(increase or decrease of excitability). Whether or not such a boundary exists and has been reached

-or even exceeded- in other rTMS protocols is an open question. In support of the existence of

temporal discontinuities in the establishment of rTMS effects within a session, Gerschlager et al.,

(2001) described effects of 1Hz rTMS over pre-motor cortex as being maximised after 3/5 of the

total stimulation. However, such studies are scarce and more information is needed about the time-

course of the effects of rTMS during stimulation. This is a challenging task: data acquisition is

limited in time by the parameters of the stimulation and the effects to detect are much smaller in

magnitude than the change observed from before to after the stimulation. Taking into account this

double difficulty, the second aim of this study is to evaluate the possibility of uncovering the time-

course of putative S1 somatotopic plasticity induced by high-frequency rTMS of the motor cortex in

healthy subjects. Optimisation of the methods will be pursued to come as close as possible to this

goal.

In this study, we asked two novel questions : (a) Can a session of high-frequency rTMS to

the motor cortex induce plasticity in somatosensory representations in healthy subjects? (b) If so, is

it possible to determine a time-course for this plasticity during the session? To answer these

questions, we used a 128-electrode cap to record N20 activity in response to ulnar nerve and thumb

stimulation. The cortical representation of these two points is spatially separated, the ulnar border of

the hand being posterior and superior (dorso-caudal) relative to the position of the thumb in human

S1. Modelling the cortical sources of these two hand regions allowed us to assess the extent of hand

representation in somatosensory primary cortex (S1) and to study its evolution following 20Hz

rTMS of the hand motor cortex.

4

Materials and methods

What methods and why ?

To ensure the relevance of this work both to fundamental and to clinical research, we chose

to evaluate the effects of a high-frequency (20Hz) rTMS protocol which is currently used to treat

patients with neuropathic pain (Hosomi et al., 2008; Lefaucheur, 2011; Leo & Latif, 2007). High-

frequency stimulation of motor cortex is also studied in other contexts (Azila Noh & Fuggetta,

2011; Cosentino et al., 2012). Stimulation was delivered to the hand motor cortex for several

reasons. First, an important number of patients treated by rTMS for neuropathic pain receive

stimulation to the superior limbs or to the face motor areas. Indeed, the motor representation of the

lower half of the body is difficult to stimulate because it is deeper located in inter-hemispheric fold.

Second, of the superior limbs and face somatosensory representations, hand representation in S1 has

been the most studied. It displays a robust somatotopic organisation of the fingers, with the thumb

located significantly more ventral than the fifth digit within area S1. Third, this somatosensory

representation is susceptible to plasticity, and changes in its somatotopic organisation have been

described after amputation or deafferentation (Flor et al., 2006), anaesthesia (Rossini et al., 1994;

Waberski et al., 2003) but also attention (Iguchi et al., 2001), or behavioural experience

(Schwenkreis et al., 2007; Weibull, et al., 2011; Xerri, 2008).

Finger representation in S1 has been assessed by functional magnetic resonance imaging

(Kurth et al., 1998; Maldjian et al., 1999; van Westen et al., 2004), magnetoencephalography

(Baumgartner et al., 1991; Hari et al., 1993; Nakamura et al., 1998) and electroencephalography

(Baumgartner et al., 1991; Baumgartner et al., 1993; Schaefer, et al., 2002a). Of all these

techniques, we chose electroencephalography because it is the most flexible to use in combination

with TMS, especially in the event of studying effects during stimulation (Siebner et al., 2009).

Indeed, MEG imaging does not allow concurrent rTMS delivery. fMRI would, but it requires

special a-magnetic equipment (carbon electrodes, special coil...) and does not allow adjustments to

the position of the coil in case the subject moves. In contrast, EEG is easily interleaved with TMS

(free coil placement, semi-specific equipment...), can be repositioned online (in real time) in case of

head movements, and can be used to successfully uncover finger somatotopy by modelling the

cortical sources of early somatosensory evoked potentials (SEPs) (Buchner et al., 1995; Houzé, et

al., 2011). SEPs are an electrical activity recorded at the scalp that results from the nervous system

in response to peripheral stimulation. Summing the EEG activity over numerous stimulations allows

the specific response of the nervous system to emerge from other ongoing EEG activity (« noise »).

Like other event-related potentials, SEPs can be broken down into discrete « components », which

5

are identified by their polarity (N for negative and P for positive) and their latency relative to the

stimulation, as well as their distribution on the scalp (scalp topography). Non-pathological SEPs can

be described as follows (Mauguière, 1995). A first diffuse positivity, P14, generated at a sub-

thalamic level is followed by a diffuse negativity, N18, that peaks 18ms after stimulation, but

persists and decays slowly. These subcortical components are followed by N20/P20, a parietal

negativity and a frontal positivity contralateral to the stimulus occurring 20ms after stimulation.

N20/P20 is the first cortical component ; there is strong evidence that it arises from area 3b of

somatosensory cortex (Allison et al., 1989; Papadelis et al., 2011). It might sometimes overlap with

P22, a radial activity whose origin is still debated (pre-central area 4 or post central area 1).

Following components include P25/N25, P27, and P45, which have a post-central origin

(subdivisions of area S1). Up until 50ms, activity seems to be restricted to primary sensory-motor

cortices (Allison et al., 1989).

When a patch of cortex (a “source”) is active and its resulting scalp activity is recorded, it is

sometimes possible to estimate the location of the source by a modelling approach. Typically, non-

moving equivalent current dipoles are created to explain activity during a given time-window, often

ideally encompassing one component (BESA® 5.3; Scherg, 1990; Scherg & Picton, 1991). The

location, orientation and magnitude of activity of each dipole are allowed to vary. These parameters

are iteratively adjusted by the software in order to minimise the discrepancy between modelled and

recorded activity, as reflected by the residual variance (RV). The resulting dipoles give an

estimation of the barycentre of the modelled activity. As a consequence, components that result

from the simultaneous activity of many different cortical areas are often difficult to model. By

studying early SEPs and focusing on the first 50ms following stimulation, one can study primary

somatosensory cortex in the absence of concurrent activity. Among the early SEP components

described above, N20/P20 (henceforth simply called “N20”) has been shown to yield the most

reliable cortical representation of the hand from stimulating the thumb and ulnar nerve (rather than

the fifth digit) (Houzé et al., 2011). Its source lies close to the cortical surface and tangential to the

scalp, which is the configuration best amenable to EEG modelling. To model N20 accurately, an

appropriate number of electrodes should be used (128 electrodes or more) in order to increase

spatial sampling and therefore spatial resolution. Overlap from other components such as P14/N18

and P22 must be reduced as much as possible (Buchner et al., 1995), which can be achieved by

using multiple dipoles over the first 50ms. The full details of these methods will now be presented.

Subjects

Thirteen subjects (9 women, 22 to 63 years old, mean age: 32.2 years, 10 right-handed) gave

their written informed consent to take part in this study and were remunerated for their

6

participation. They had no previous neurological or psychiatric history. The study was approved by

the local ethics committee (CPP Léon Bérard-Lyon; 2008-A01437-48). Ten subjects took part in the

main experiment; three of them underwent the sham experiment. Three new subjects undertook the

sham experiment only. As three subjects (2 from the main group and one from the sham group) did

not yield data of appropriate quality for modelling, they were removed from analysis. Therefore,

main experiment analysis was conducted on 8 subjects and sham experiment analysis on 5 subjects.

Experiment time-course

At the beginning of a session, subjects were fitted with a 128-electrode EEG cap. To

evaluate the cortical representation of the hand before rTMS, SEPs were recorded in response to

electrical stimulation of the first digit and ulnar nerve of the left hand. For each of the two

peripheral stimulation sites, two runs of ~1000 stimuli (ie 2000 stimuli per site) were applied at 3Hz

(Figure 1A). SEP acquisition lasted approximately 20 minutes.

The subject's left hand motor “hotspot” and motor threshold were then determined by

neuronavigated (i.e. MRI-guided) single-pulse TMS over the right primary motor cortex (Figure

1B). The rTMS session itself consisted of 20 consecutive trains of 80 stimulations, delivered at

20Hz (Figure 1C). These brief 4 seconds-long stimulation trains were separated from one another

by 84s inter-train intervals.

Figure 1: Experiment time-course: interleaved SEPs and TMS. A) The experiment started with the recording of somatosensory evoked potentials (SEPs), elicited by left hand ulnar nerve and thumb electrical stimulation. For each peripheral stimulation site, 2 consecutive runs of ~1000 stimuli each, delivered at 3Hz, were recorded. Once the pre-rTMS SEP recordings were done, B) hand motor hotspot and threshold intensity were established using single pulses of TMS to the right motor cortex. C) 20 Hz repetitive TMS was then applied for 20 minutes to the left-hand motor hotspot, at a 90% threshold intensity. rTMS pulses were delivered in short 4-seconds trains, followed by inter-train intervals of 84s. D) Inter-train intervals were used to record short runs of SEPs, interleaved to the motor cortex stimulation. An SEP run consisted of ~200 stimulations; peripheral stimulation sites (U: ulnar nerve, T: thumb) alternated at each new run. Altogether, ten short runs for each stimulation site were recorded. These SEPs recorded during rTMS were followed by E) post-rTMS SEPs using a procedure identical to the pre-rTMS SEP session.

7

Central stimulation(motor cortex)

Peripheral stimulation (hand nerves);Recording of cortical response (SEP)

T1 U2 T10

Pre rTMS During rTMS Post rTMS

t (min)

0 204s

84sA

B C

D

U1

E

Single pulseof TMS

20 Hz rTMS

Ulnar nerve

Thumb

Stimulation sitefor SEP recordings

200stimuli

2000stimuli

In order to assess hand representation plasticity during rTMS, twenty very short SEP runs

(150-200 stimuli) were recorded interleaved to the rTMS stimulation, during the 84s inter-train

intervals (Figure 1D). This procedure yielded ten measurements for the first digit, alternating with

ten measurements for the ulnar nerve, across the entire rTMS session. Left hand somatotopy was

again assessed shortly after rTMS (Figure 1E) (2 runs of ~1000 stimulations x 2 peripheral sites;

approximately 20 minutes). A typical session would last at least 3 hours.

Spatially guided rTMS

rTMS was delivered through a figure-of-eight coil (Cool-B70 butterfly coil, MagVenture®)

connected to a MagVenture stimulator (Alpine Biomed®) generating biphasic repetitive magnetic

pulses. Coil position relative to the subject's brain anatomy was monitored and guided by an MRI-

based neuronavigation system (VISOR, ANT®) in which the subject's individual 3D-MRI was

loaded.

Placement of the coil over the left-hand motor cortex was done as follows. First, with the

help of the neuronavigating system, the coil was centered over the hand “omega” (Yousry et al.,

1997) with an antero-posterior orientation, i.e., orthogonally to the central sulcus. Single pulses of

TMS were then applied with varying intensity and position, while the left hand muscle responses

were recorded by bipolar electrodes placed over the muscle abductor digiti minimi. The cortical

point allowing best motor responses with lowest TMS intensity was defined as the motor ‘hot spot’.

The rTMS session was then performed over the hot spot, at intensity equal to 90% of the motor

threshold, itself defined as the TMS intensity eliciting 5/10 motor responses with at least 50µV

amplitude at rest (Rossini et al., 1994). 1600 stimulations were delivered at 20Hz over 20 minutes.

Somatosensory evoked potentials (SEP)

Optimising the methods : choice criteria

When modelling early SEP components that obey only a few sources (such as the N20

component studied here), the quality of the processed data is determinant for the quality of the

modelling results. Indeed, these components are generated by sources of small volume, whose

activity recorded at the scalp is of very small amplitude. This makes them extremely sensitive to an

increase in the background noise : if the signal-to-noise ratio is too low, the signal will be drowned

in the noise and it may be impossible to model the desired activity. Conversely, a change in the

number of dipoles, the interval of interest and other parameters intrinsic to the modelling process

should lead to very similar solutions in terms of localisation and timing.

8

In most instances where source localisation of scalp signals is attempted, the problem is

finding a broad localisation of the most probable source, and therefore solutions providing

millimetric uncertainty may be largely sufficient. The questions asked in this study require, on the

contrary, as little incertitude as possible, as changes in cortical hand representation due to

experimental manipulations are likely to amount to some millimetres only, ie barely exceeding the

intrinsic localisation incertitude of the technique. Taking into account the above constraints, a

substantial part of this work was devoted to optimise data processing and model, which would yield

robust, accurate and well fitting solutions.

a) In order to evaluate robustness, five different time-intervals were chosen on which to fit the

N20 dipole. They varied in total duration and latency on the ascending slope of the N20

component, but could all have reasonably been chosen by an experimenter. A robust model

should resist to an operator-dependant choice of time-window, it should therefore display

little variability in response to this manipulation.

b) Accuracy was evaluated according to physiological and anatomical knowledge : the cortical

representation of the thumb should be located anterior and inferior to that of the ulnar nerve

and all dipoles should be posterior (or close to) the central sulcus.

c) Finally, residual variance indicated how well the model fitted the data. Albeit a minimal

residual variance (RV) on the N20 time-window was sought, the ability of the data and

model to uncover sources of other SEP components within S1 was also taken into account,

and a situation where one far-field (brainstem), one tangential (area 3b) and one radial

components (areas 1-2) could be modelled was preferred, as this reflected more accurately

the underlying physiology (Allison et al., 1989). To ensure a high signal-to-noise ratio, all

tests on robustness and accuracy were done on grand-average data (i.e. averaged across all

participants).

SEP recording

Subjects were comfortably seated in a quiet, semi-darkened room. They were instructed to

relax; they could close their eyes but were asked to stay awake. First, the sensory threshold was

determined by applying stimulations of increasing intensity to the first digit (ring electrode) and to

the ulnar nerve, close to the left wrist (bipolar electrode) until subjects mentioned they felt them.

Then, series of stimuli (0.2ms square waves; 3Hz) were delivered at an intensity of threefold the

sensory threshold, via an IRES-600 isolated, constant-current stimulator (Micromed®). Stimulation

was applied at only one site at a time.

A high-density 128-channel EEG cap (WaveGuard® Cap, ANT®) was used to record

9

cortical electrical activity. Sensors in the cap being ‘floating electrodes’, the skin-electrode

impedance was kept below 5kOhm using a conductive gel (Electro-Cap®). The left mastoid (M1),

ipsilateral to peripheral stimulation was chosen as reference (Cruccu et al., 2008)(Cruccu et al,

2008). The ground electrode was incorporated in the cap between Fz and Afz. An ASA amplifier

and ASA software (ANT®) were used to record the signal at a sampling rate of 2048 Hz, and with a

0.2-1024 Hz band pass filter.

SEP analysis

Electrophysiological data was analysed under BrainVision Analyzer software

(BrainProducts®). A 50Hz notch filter was applied to the raw EEG signal in order to minimise

spurious activity from the electrical network. The continuous signal was then segmented in 120ms-

long epochs, composed of 20ms pre-stimulus and 100ms post-stimulus periods. Each epoch was

detrended in order to minimise the influence of slow linear trends (using the first and the last 10ms

of the segment). A baseline correction was applied using a slightly shortened pre-stimulus interval (-

20 to -3ms), in order to exclude the edges of the stimulation artefact from the calculation of the

baseline. Semi-manual artefact rejection was performed, so as to discard any segment with activity

exceeding ±150μV. This resulted in the exclusion of ~10% of epochs. The segments were then

averaged. The resulting average was band-pass filtered between 15 and 350 Hz (Butterworth filters,

24 dB/oct). Since source localisation heavily relies on a good signal-to-noise ratio at all electrode

sites, the signal from electrodes that remained of poor quality after the processing steps described

above (detrending, baseline correction and filtering) was interpolated using spherical spline

functions. This method allows to reconstruct the signal at one electrode site by taking into account

the signal from the 128 electrodes of the whole cap (Perrin et al., 1989). The final product could

then be averaged across time (combination of two following stimulation runs) and/or averaged

across all subjects, resulting in a grand-average.

Base-line correction and physiological constraints

While the quality of the data was made as high as possible by the use of a high-density

electrode cap (128 electrodes) and by classical SEP processing, a difficulty specifically linked to the

modality of SEPs was encountered in this study. The electrical stimulation of the peripheral nerves

used to trigger the SEP was recorded at the scalp, resulting in a big stimulation artefact which

disrupted the previously stable base-line (Figure 2A). Given that, when using a cephalic reference,

no physiological activity should be recorded at the scalp before 14ms post-stimulation (Mauguière,

1995), the situation called for a re-definition of the base-line. A last processing step was therefore

10

taken : base-line correction using a post-stimulus interval (5 to 13ms) was applied to segments

before source reconstruction.

SEP source reconstruction

Models of early SEPs described in the literature typically used between one to three dipoles,

describing N20 activity, and sometimes P14 or P22 activities (Baumgartner et al., 1993; Buchner et

al., 1995; Pleger et al., 2001). Other studies have suggested modelling the noise in order to remove

it from the physiological model (Godinho et al., 2006; Houzé et al., 2011). In an attempt to reflect

this diversity, three different types of models were fitted to the data. Four dipoles were used to

describe SEP activity over the first 50ms : (1) a dipole accounting for subcortical activity previous

to N20, (2) a dipole accounting for cortical N20 activity, (3) a dipole accounting for later activity

(up to 50ms) and (4) a noise dipole explaining noise at the same time as the N20 dipole. Model M1

included dipoles n°1 and n°2; model M2 complemented model M1 by the addition of dipole n°3;

finally, model M3 comprised all four dipoles.

Global Field Power (GFP) profile, in combination to recorded data, was used to determine

the time-windows during which dipoles should be reconstructed (see Figure 4, top for an

illustration). GFP corresponds to the spatial standard deviation in a series of maps of potentials

(Skrandies, 1990); a peak of GFP over time corresponds to a series of fields with high similarity. It

helps to determine the latency of evoked-potentials components (Hamburger & Van der Burgt,

1991). P14 and N20 dipoles were fitted over intervals of varying lengths that covered the ascending

phase and/or the peak of the GFP curve at their respective latencies. The third dipole was fitted from

the end of the N20 dipole time-window to the end of the epoch. It accounted mostly for the radial

activity elicited by P22, N30 and P45 components. The noise dipole was added last.

Solutions with a residual variance (RV) less than 10% and compatible with physiological

knowledge were considered reliable. The model was first created on grand-average data (averaged

across participants). It was then adjusted to individual data; the time-window, dipole number and

dipole order could be modified if justified (artefact in the middle of a time-window, difference in

latency compared to the grand-average...). A four-shell ellipsoidal head-model was used (default

adult parameters values); regularisation constant was set at 1% (default value).

Coordinates, measures and projection on anatomical image

Dipole position was defined by its coordinates in the Talairach system of reference, whose

origin is the anterior commissure, with X for the lateral medial axis, with X=0 being the coordinate

of the sagittal inter-hemispheric plane; Y for the rostro-caudal (anterior-posterior) axis, Y=0 being

11

the coordinate of the vertical anterior commissure (VAC) plane and Z for the inferior-superior axis,

Z=0 being the coordinate of the horizontal AC-PC plane orthogonal to the mid-sagittal vertical

plane and passing through the anterior and posterior commissures. According to their respective Y

and Z coordinates, dipoles were located on the surface of an individual Talairach-normalised brain

(MRIcro) and 2-D euclidean distances between dipoles pre- and post-rTMS were calculated.

Coordinates and distances are expressed in millimetres.

Sham stimulation

Six subjects (3 of the main group, 3 new participants) received “sham” rTMS, i.e. they

underwent the entire standard experimental procedure, but unbeknownst to them, no actual rTMS

was delivered. After measuring the individual hand motor hotspot and motor threshold, the

stimulating coil was switched with a sham coil (MCF-P-B65 coil, MagVenture®) identical in shape,

size and noise produced by an active coil, but not delivering any significant magnetic field. This

allowed to control for any confounding effects such as placebo, noise of the coil during stimulation,

or repeated finger stimulation for the recording of SEPs.

Statistical analysis

A Kolmogorov-Smirnov test was used to make sure that the data were normally distributed.

Paired t-tests were applied to the coordinates of thumb and ulnar nerve dipoles. Two-way repeated

measures ANOVAs, with “time” (pre- or post- stimulation) as within factor and “type of

stimulation” (active rTMS or sham) as between factor were run on the 2-D Euclidean distance

between thumb and ulnar nerve dipoles. Post-hoc paired t-tests were then applied with a Bonferroni

correction. The level for significance was set to p<0.05.

12

Results

A. Searching for the best possible parameters for source modelling.

In order to investigate the somatotopic representation of the hand in S1 and its plasticity

with a maximal precision, we sought to optimise modelling conditions. The results of the

optimisation process described in the Methods section above is presented below.

I. Data pre-processing

Changing the base-line from a pre-stimulus interval (-20 to -3ms) to a post-stimulus interval

(5 to 13ms) considerably modified scalp topography of early potentials and GFP (Global Field

Power) profiles. This allowed a better identification and differentiation of the early sub-cortical SEP

component P14 (Figure 2A, (1)).

Figure 2: Change of base-line from pre- to post-stimulus. A) Overlay of electrical signal from all 128 electrodes following stimulation of the ulnar nerve, prior to rTMS. The signal was base-line corrected with a classical pre-stimulus interval (-20 to -17ms, top), or with a post-stimulus interval (5 to 13ms, bottom). Changing the base-line from pre- to post-stimulus decreased the non-physiological scatter of activity following stimulation (red arrows) and modified early SEP components P14 (1) and N20 (2) as well as following activity. Grand-average data. B) Projection of N20 dipoles onto the lateral convexity of a Talairach-normalised brain. The data presented in A) was analysed with a four-dipole model (see methods), using 5 possible time-windows for fitting the N20 dipole (groups of points of one symbol and one colour). All N20 dipoles were located posterior to the central sulcus.

Modelling of thumb and ulnar nerve sources before rTMS yielded dipoles that were all

posterior to the central sulcus, with the thumb always anterior and inferior to the ulnar nerve dipole

(Figure 2B); RV was no more than 4%. Also, dipoles obtained for different time-intervals were less

scattered when a post-stimulus base-line was used, indicating that using the post-stimulus period as

baseline enhanced the robustness of the model. Moreover, the P14 component that was difficult to

13

model accurately in a pre-stimulus base-line condition became available to modelling when the

base-line was post-stimulus. For these reasons, a post-stimulus baseline was preferred for all

subsequent calculations.

II. Type of model

The three modelling variants described in the Methods (section SEP modelling) were

compared in terms of residual variance and agreement to human physiology. Figure 3 shows that

model M1 (two dipoles : one describing P14, the other N20 activity) successfully described more

than 90% of the variance of the N20 time-window (8.8% of residual variance). The thumb was

positioned anterior to the ulnar nerve, according to physiology. However, vertical differentiation of

the two dipoles was not possible. Model M2 (model M1 plus a dipole describing activity after N20)

partly solved this issue and marginally reduced the residual variance (RV). Further improvement,

both in physiology and RV values was achieved with M3 (model M2 plus a dipole accounting for

noise). Positions of thumb and ulnar nerve sources were separable, physiologically positioned,

located posterior to the CS and only 3.5% of variance remains unexplained by the model. The use

of model M3 was therefore preferred for the rest of the study.

Figure 3: Testing three types of models of increasing complexity. Projection on a Talairach-normalised brain of ulnar and thumb N20 dipoles, based on three different models. Model M1 comprised two dipoles: one accounting for P14 activity and the other one for N20 activity. Model M2 consisted of M1 plus a dipole characterising activity from the end of the N20 up to 50ms. M3 complemented M2 with an additional dipole fitted at the same time as N20 that could account for noise. Mean residual variance (RV) describes the percentage of data that remained unexplained by the model. Data was a grand-average; it was base-line corrected based on a post-stimulus interval.

14

A

Y (Talairach coordinates)

Z (T

alairach coordinates)

25

30

35

40

45

50

55

-45 -40 -35 -30 -25 -20 -15 -10

M1M2M3

Model Mean RV (%)8.87.83.5

Ulnar pre-rTMSThumb pre-rTMSCentral Sulcus

III. Analysis time-window

Because it selects the activity to be analysed, the choice of the analysis time-window could

potentially influence the results, to an unknown extent. Latency is often determined by the

ascending phase of the N20 component (either in the raw data or in the GFP waveform, Figure 4,

top). A later time-window might capture P22 activity. In choosing the length of the time-window,

there is a trade-off between residual variance -which gives an estimate of how well the model fits

the data- and how much representative activity is captured. A short fit interval will yield a small

residual variance but will be very sensitive to any artefact within that fit interval. Stability of source

location with changes in time window was one of the parameters used to estimate robustness of the

model.

Figure 4: Influence of the N20 analysis time-window. Five time-windows were chosen around the N20 peak of the GFP (Global Field Power) curve: a long (4ms), a short (2ms) and three 3ms-long intervals whose latency varied (top). An N20 dipole (model M3) was determined based on these different time-windows (bottom). Thumb and ulnar grand-average data were post-stimulus base-line corrected. The dotted line on the GFP curve replaces the profile of the stimulation artefact to increase readability.

15

25

30

35

40

45

50

55

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

N20 time-window

EarlyMediumLateShortLong

Ulnar pre-rTMSThumb pre-rTMSCentral Sulcus

GFP

t (ms)0 20 40

Figure 4 (bottom) shows that ulnar nerve dipole position was particularly robustly

determined, as changing the time-window did not lead to important changes in dipole position.

Thumb dipoles, however, displayed more scatter, especially following the Z axis. This slightly

increased incertitude in thumb dipole position could be explained by the quality of the signal.

Indeed, nerve trunk stimulation (ulnar nerve) as opposed to stimulation of cutaneous fibres (thumb)

activates a much greater number of peripheral axons, hence yielding signals of greater amplitude.

However, in both cases, the scatter was relatively limited and did not prevent the two dipoles from

being easily separated. The choice of the time-window was therefore not considered crucial, as long

as it was on the ascending slope of the N20.

B. Source modelling of SEPs

I. Estimating hand cortical representation before rTMS

Having determined the optimal conditions for modelling as per the results above, we

investigated whether or not 20 minutes of 20 Hz rTMS delivered to the hand motor cortex could

induce a plastic reorganisation of somatosensory S1 hand representation.

a) Grand-averaged data. When modelling N20 component using different fitting time-

windows, all dipoles were obtained with no more than 4% residual variance. When projected onto a

standard Talairach-based MRI, all groups of dipoles located posterior to the central sulcus (Figure

5). Changing the N20 analysis time-window did not affect positions much. With this reliable

modelling, the positions of ulnar and thumb dipoles were clearly different and dissociable : the

thumb was anterior and inferior to the ulnar nerve (Figure 5, blue symbols).

Figure 5: Modelling of grand-average data, before and after rTMS. Projection of N20 dipoles on a Talairach-normalised MRI. Ulnar nerve and thumb N20 sources have been modelled both before (blue) and after 20 minutes of 20Hz rTMS (red). The five points for each group (same symbol, same colour) correspond to different N20 analysis time-windows tested successively on the same data.

16

25

30

35

40

45

50

55

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

Pre rTMSPost rTMS

UlnarThumbCentral Sulcus

b) Individual data. The model obtained in the best possible conditions (grand-average,

maximal signal-to-noise ratio) was used as an adjustable template to describe individual data

obtained from an active stimulation experiment as well as from a sham (placebo) control

experiment. For two subjects out of ten from the active group and one out of six from the sham

group, modelling was unsuccessful according to our rather strict criteria (i.e. RV was >10% and

positions of dipoles were non-realistic, with presence of clear artefacts and/or exit from the model

ellipsoid). These three subjects were removed from later analysis as well as from the grand-average

analysis. Residual variance in another subject from the control group was slightly superior to 10%

but he was kept since localisation of sources was consistent with that of the other subjects. Position

of individual dipoles reproduced in each subject the findings described above for the grand average:

i.e. all dipoles respected finger somatotopy, and were located posterior or only slightly overlapping

the central sulcus. Mean RV for ulnar nerve and thumb before rTMS was small (Table 1), and the

modelling was therefore considered reliable.

Table 1: Talairach coordinates of N20 dipoles and residual variance. Mean data, with standard error of the mean, for active rTMS stimulation (N=8) and for sham rTMS (N=5). Residual variance is expressed as a percentage.

Previous to rTMS, the position of the ulnar nerve dipoles and thumb dipoles was

significantly different on the Z axis (t(4) = 3, p<0.05), but not on the Y axis (t(4) = 0.6, p>0.05) for

sham stimulation. For active rTMS stimulation, the position of ulnar nerve dipoles (Table 1) was

significantly different from that of the thumb dipoles, both on the Y axis (paired-t-test : t(7) = 3.56,

p<0.01) and on the Z axis (paired t-test : t(7) = 3.65, p<0.01). Thus, the method used allowed to

separate significantly ulnar nerve and thumb representations and therefore to determine a cortical

17

representation of the hand before any manipulation –a necessary step to validate the rest of our

analyses (Figure 6A, blue symbols).

II. Changes in source localisation after rTMS

a) Grand-averaged data. Preliminary observation of this reliable modelling showed a

movement of dipoles after rTMS, relative to their position before rTMS, both for thumb and for

ulnar nerve. As both dipoles moved away from each other, the distance between ulnar nerve and

thumb on the grand average increased (from 15 to 18 mm, 3 mm) (Figure 5, red symbols).

a) Active high-frequency rTMS. This was confirmed by modelling of individual data. Twenty

minutes of 20 Hz rTMS applied on the motor cortex modified the hand representation: the mean

ulnar nerve dipole (average position across all subjects) displayed a posterior and superior

movement, while the mean thumb dipole showed an anterior and inferior displacement (Figure 6A,

red symbols).

Figure 6: Modelling of individual data, before and after rTMS. A) Mean position of individual N20 dipoles, for the ulnar nerve and the thumb, both before (blue) and after 20 minutes of 20Hz rTMS (red). Error bars denote the SEM (standard error of the mean); N=8. Projection on a Talairach-normalised MRI. B) Individual change in dipole position from pre- to post-rTMS in a sagittal Talairach plane. Black numbers represent individuals. Red lines correspond to the individual difference in position of N20 dipole after rTMS compared to before rTMS. All movements have been represented relative to the mean position of dipoles before rTMS (blue symbols) in order to improve readability.

18

A B

25

30

35

40

45

50

55

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

Pre rTMSPost rTMS

UlnarThumbCentral Sulcus

3

416

5

82

7

7

261

385

4

25

30

35

40

45

55

50

20

15

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

Ulnar

Thumb

Pre rTMSPost rTMS

This proved to be significant at group level, (see below paragraph c, and Figure 6B), even

though a few subjects followed only one of the trends described above. For instance, one subject's

ulnar nerve dipole moved upwards (subject 4), but towards the central sulcus; his thumb dipole

moved mainly downwards. More generally, thumb dipoles moved mainly along one axis, while the

movement of ulnar nerve dipoles often involved both Y and Z axes. Inter-individual differences

were important, for example displacement for the ulnar nerve dipole ranged from 2 to 17mm. These

modifications of dipole position in opposite directions resulted in an important enlargement of the

hand representation (10.5 mm, SEM = 3.5 mm, N = 8) (see Table 2 for sizes of hand

representation).

Table 2: Mean 2D euclidean distance between ulnar nerve and thumb dipoles for sham and active rTMS and elements of analysis (ANOVA). Distances are expressed in millimetres.

b) Sham (placebo) high-frequency rTMS. A control experiment with sham stimulation was

run to ensure that neither a placebo effect, nor the noise of the coil during stimulation, nor the

peripheral stimulation used to record SEPs were responsible for the findings described above.

Changes in location of dipoles after sham stimulation were much smaller and disorganised in space

relative to those obtained after active rTMS, (ulnar nerve dipole range : 2 to 6 mm) resulting in a

mean position of dipoles barely modified after placebo stimulation ( = 0.8 mm, SEM = 2 mm; N =

5, Table 2) (Figure 7A).

c) Comparison between active and sham rTMS. The effect of active and sham stimulation on

the distance between ulnar nerve and thumb dipoles was quantified using repeated-measures

ANOVA, with “time” (pre- or post-stimulation) as within factor and “stimulation type” (active or

sham) as between factor. The ANOVA showed no main effect of stimulation type (F(1,11) = 0.42, p

= 0.5), a main effect of time (F(1,11) = 5.64, p = 0.04), and a strong trend for interaction (F(1,11) =

4.19, p = 0.06).

19

Figure 7 : Sham stimulation compared to active rTMS. A) Mean dipolar position of individual N20 dipoles after sham stimulation (light blue symbols, N=5) or after active rTMS (red symbols, N=8). The change in position following sham rTMS has been reported relative to the mean position before active rTMS (dark blue symbols). Error bars denote SEM. Projection on a Talairach-normalised MRI. B) Mean 2D-euclidean distance between ulnar nerve and thumb N20 dipoles, before and after rTMS, either active or sham. Error bars represent SEM, ns : non significant, ** : p<0.01.

Post-hoc Bonferroni tests showed that the distance between ulnar nerve and thumb dipoles

was significantly increased after active rTMS (= 10.5 mm; 95% CI : 3 to 18 mm; p<0.01), but not

after sham stimulation (= 0.8 mm; 95% CI : -9 to 10 mm; p>0.05) (Figure 7B). Thus, real 20 Hz

rTMS delivered to the hand motor cortex did modify S1 somatotopy, enlargingthe cortical

representation of the hand, whereas sham stimulation failed to do so.

III. Source modelling of PES during rTMS

High-frequency 20 Hz rTMS to the hand motor cortex enlarged the representation of the hand

in S1. Is it possible to study the time-course of this plasticity? To answer this question, we

investigated the evolution of dipole localisation estimated from short SEP runs obtained

sequentially during the administration of rTMS (see Figure 1). Thus, short runs (~1500-200 stimuli)

of peripheral stimulations were “interleaved” between rTMS bursts, and applied during the 84

seconds that separated consecutive rTMS trains. The analysis was performed on grand-averaged

data from all subjects. Twenty small SEPs recordings spread across the rTMS session allowed the

reconstruction of 8 “consecutive” dipoles for each stimulation site, reflecting the putative evolution

of dipole position during the progression of the rTMS session (the first and last SEP recordings for

each stimulation site were discarded because they were not consistently recorded in all subjects).

Although the dipoles were obtained with acceptable RVs and all but one located posterior to

20

25

30

35

40

45

50

55

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

AUlnarThumbCS

Post shamPost rTMS

Pre rTMS

0

5

10

15

20

25B

Active rTMS Sham

Uln

ar-t

hum

b in

ter-

dipo

le d

ista

nce

(mm

) Pre

Post

ns

**

the central sulcus, not all pairs of dipoles for a given time point respected somatotopic arrangement

(Figure 8A).

Figure 8: Modelling of grand-average data during stimulation. A) N20 dipole position for small time-slots during stimulation (2 to 9, spread out across the 20 minutes of stimulation, cf Figure 1), for the ulnar nerve (U) or the thumb (T). Grand-average model of pre- and post-rTMS shows the “starting” and “finishing” points for the cortical representation. The light green ellipsoid shows the area of presence of ulnar nerve dipoles, the dark green one of thumb dipoles. B) N20 dipole position for data that has been averaged across subjects (grand-average) and across time (first 5 time-slots and 5 last time-slots of the rTMS session), for the ulnar nerve (U) and the thumb (T). Projection on a Talairach-normalised MRI.

Thumb dipoles were more scattered than

ulnar nerve dipoles, and no coherent progression

with time could be seen during the rTMS session,

especially when comparing to dipole positions

before and after rTMS. An attempt at fitting a

regression curve to the data was highly unsuccessful

for a simple linear regression model (r²<0.005)

(Figure 9).

Figure 9: Evolution of 2D inter-dipole distance during rTMS. 2D euclidean distance between ulnar nerve and thumb N20 dipoles was computed. A simple linear regression was applied. As it did not describe the data well (r²<0.005), no best-fit curve was displayed. Modelling of grand-average data (N=8).

21

25

30

35

40

45

55

50

15

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

20

25

30

35

40

45

50

55

-45 -40 -35 -30 -25 -20 -15 -10Y (Talairach coordinates)

Z (Talairach coordinates)

UlnarThumbCentral Sulcus

Pre rTMSPost rTMS

U2

U9U8

U7

U6

U5

U4

U3

T2T9

T8

T7

T6

T5

T4

T3

A BUlnarThumbCentral Sulcus

Pre rTMSPost rTMS

U1 to 5

U6 to 10

T1 to 5

T6 to 10

0

5

10

15

20

25

30

35

40

45

0 2 4 6 8

PrerTMS

PostrTMS

9753

2D u

lnar

ner

ve-t

hu

mb

dis

tan

ce (

mm

)

Time interval relative to rTMS session

During rTMSPost rTMS

Pre rTMS

r²<0.005

In view of these negative result, and in order to increase the signal-to-noise ratio, half the

grand-averages were combined together; this yielded two measurement points, one summarising all

recordings obtained during the first half of the stimulating session, the other summarising the other

half. These data were again modelled with acceptable RV. One dipole was markedly anterior to the

central sulcus, while in a pair of dipoles the thumb was posterior to the ulnar nerve (Figure 8B). All

dipoles modelled during the rTMS session located superior to dipoles modelled before and after

rTMS. These results did not allow us to draw any conclusions about the time-course of plasticity in

S1 induced by motor cortex 20 Hz rTMS.

22

DiscussionIn this study we assessed, in healthy subjects and using somatosensory evoked potentials,

the ability of high-frequency rTMS over the motor cortex to induce plastic changes in the cortical

somatotopic representation of the hand. The results of source localisation showed the following:

d) A session of high-frequency rTMS delivered to the motor cortex did alter S1 somatotopic

maps: in our subjects, S1 hand representation was enlarged by 10.5 ±3.5 mm following 20

minutes of high-frequency rTMS.

e) A sham rTMS session with identical physical characteristics but absence of magnetic field

failed to induce any significant variation of the S1 hand representation.

f) The methodology used to derive the above results was not powerful enough to describe the

details of the time-course of S1 plastic changes, when using ultra-short SEP runs interleaved

between the rTMS trains.

In what follows, each of these relevant results obtained will be discussed, but first some preliminary

comments on the general methodology followed in this work appear necessary.

Modelling methods

Alteration of somatosensory maps was uncovered using dipole modelling of SEPs. This

method is most accurate when sources of activity lie close to the cortical convexity and their

activity is recorded by a high-density array of electrodes (Scherg, 1990). Here, the conditions were

optimal, since we studied responses originating in area 3b, which is close to the parietal convexity,

using a 128 electrodes EEG cap. Of notice, in patients with neural lesions previous descriptions of

plastic changes in cortical representations via EEG modelling used different modes of stimulation

(electrical, tactile or air-puff) and stimulation sites (digits, nerve trunks, cutaneous territories), as

well as a substantially smaller number of electrodes as compared with the present study, ranging

from 32 to 61 (Birbaumer et al., 1997; Grüsser et al., 2001; Karl, et al., 2001; Montoya et al., 1998;

Schaefer et al., 2002a; Schwenkreis et al., 2001), which may explain some disparity of results. Work

performed in our laboratory previous to this report evaluated the impact of electrode density (64 vs

128 electrodes) and territory stimulated (four territories of the hand) on the quality of modelling

data. This work strongly suggested the need of high-density electrode arrays (128 electrodes) and

stimulation of ulnar nerve vs thumb to optimise the estimation of hand cortical representation using

potentials from the primary sensory cortex (Houzé et al., 2011). The methodology used in the

present study benefited from these previous results, and can be considered on theoretical grounds as

superior to most of previously documented work in this area.

23

Our methodology was however not optimal in absolute terms: for instance, localisation

accuracy could have been improved by using realistic head models with co-registration of the

electrode positions to the individual subject MRI (Brinkmann et al., 1998). As well, position relative

to central sulcus could have been assessed according to individual anatomy, even if the main object

of this study was not the exact anatomical position of the thumb and ulnar nerve dipoles but their

relative distance. Under those conditions, localisation of sources in 3b has been shown to be highly

reliable and reproducible when tested several times in the same subject, both at short and longer

intervals (Schaefer et al., 2002a, 2002b). However, to our knowledge no systematic study has

evaluated the real improvement of modelling solutions by direct contrast of models calculated with

and without coregistration of electrode positions with MRI. In the absence of such data, the actual

benefit of coregistration on the assessment of dipole position within the cortex, although highly

probable, remains hypothetical.

Two other parameters that could have improved the efficacy of source modelling would have

been increasing the number of EEG sensors (e.g. to 250 electrodes or more) and the number of

stimuli delivered. The optimal number of sensors to resolve spatially scalp EEG signals remains a

matter of debate. The discrete sampling of the brain’s electrical field at the scalp surface with

individual recording sensors is subject to the same sampling error as the discrete sampling of the

time series with analog-to-digital conversion. However, since the skull acts as a low-pass spatial

filter of the brain electrical field, high spatial frequency information in the EEG is attenuated or

suppressed, and it may be intrinsically useless to enhance the number of electrodes beyond such

resolution capacities. Previous work in the laboratory (Houzé et al., 2011) as well as theoretical

considerations (Srinivasan et al., 1998) suggests that adequately sampling the human EEG requires

a minimum of 128 sensors; however, determining whether additional increases in sampling will

improve the spatial resolution of the scalp EEG remains debatable, since it largely depends on

accurate estimation of skull resistivity –the main factor determining signal smearing. Recent reports

have suggested that the skull may be less resistive than previously estimated (Oostendorp et al.,

2000), and this may allow higher spatial frequency of the human EEG than previously believed.

Should this be confirmed, greater spatial sampling than offered by 128-channel arrays should

further improve the quality and accuracy of cortical representation estimates, and some reports have

suggested that interelectrode distances of 1 cm —corresponding to about 500 channels— may be

required to fully represent the spatial frequency of the scalp EEG (Freeman et al., 2003; Ryynänen

et al., 2005).

Concerning the number of stimuli delivered, there is little doubt that this is a crucial factor

allowing to improve the signal-to-noise ratio and enhance model accuracy. In our case, we were

24

limited by the total time of the experiment (almost 4 hours) and the need to avoid as much as

possible the subject’s fatigue (itself a source of artefacts and degrading of the data). However,

retrospectively it is clear that modelling was easier, more robust and optimally physiological in

subjects in whom a greater number of artefact-free SEP samples could be recorded, and that 2000

stimuli is just a ‘lower limit’ to model correctly signals of low amplitude, not exceeding 0.5 to 2 μV,

as is the case of parietal N20.

As per the above considerations, efforts will be devoted in the future to (a) increase the

number of stimuli, for example by increasing the stimulus rate to 5-6 Hz, which is compatible with

stability of the N20 (Garcia Larrea et al., 1992); (b) use realistic head models with coregistration of

electrodes on the subjects’MRI, so as to ensure that the orthogonal projection of sensors over the

brain surface is not mis-estimated by the head model, and (c) enhance the number of sensors to 250,

at least when discrimination of high spatial sampling is needed.

Finally, we presented a methodology for testing the robustness of different models. Fitting

the model to different time-intervals allowed discriminating between situations where concomitant

time-intervals produced very similar solutions and others where more disparities were noted. This

allows a fast and useful quality check of the model and data set regarding the non-dispersion of

measures of a same activity of interest. This quality check could become a systematic first step to

future modelling studies undertaken in the laboratory.

Mechanisms of rTMS-evoked plasticity

Somatosensory cortex is thought to have a somatotopic organisation. That is, somatosensory

stimuli from one part of the body are detected by the receptive fields of a dedicated group of

neurons, thus creating a cortical representation of the whole body. These so-called “somatotopic

maps” are thought to be dynamically maintained entities, that are shaped by a neuronal network and

the information relayed through it (Xerri, 2008). In this study, we found that high-frequency rTMS

over the hand motor cortex was able to modify such somatotopic maps in the neighbouring sensory

cortex, hence triggering motor-to-sensory plasticity. This suggests that the “information” contained

in the 20 Hz rTMS trains was relayed from the motor cortex to the somatosensory cortex, possibly

in a somatotopic way.

The anatomical pathways mediating S1 plasticity might have been direct, as direct reciprocal

connections between sensory and motor areas in primate brain exist (Krubitzer & Kaas, 1990).

Instead, or as well, plasticity could have been mediated by more indirect cortico-thalamo-cortical

connections. Indeed, there is evidence for cortico-thalamic connections from motor cortex M1 to

ventral lateral (VL) and ventroposterolateral (VPLo) nuclei of the thalamus in primates (Kultas-

25

Ilinsky et al., 2003; Rouiller & Welker, 2000). Cortico-thalamic projections from M1 could

modulate ascending sensory signals, as can be seen in rodents (Lee et al., 2008). In support of this

hypothesis, functional imaging studies have demonstrated greater metabolic changes in the lateral

thalamus than in sensorimotor cortex following subthreshold motor cortex stimulation, either

transcranial or epidural (Bestmann et al., 2003; Garcia-Larrea & Peyron, 2007).

It should be noted that a spread of stimulation from the motor to the somatosensory cortex

might have also theoretically contributed to S1 plasticity. Indeed, it has been claimed that the

focality of TMS stimulation may have been overestimated in early models of the spread of

stimulation (Wagner et al., 2009). It appears that even though the peak of the magnetic field might

be placed over a given brain region, other regions can be affected; the geometry of this field of

influence is complex and still under investigation (Toschi et al., 2008). In the present work,

however, several steps were taken to ensure that the motor cortex, and not the somatosensory

cortex, would be directly stimulated in our study. First, magnetic stimulation was performed under

the guidance of a neuronavigation system, thus ensuring the correct positioning of the coil over the

precentral cortex (Lefaucheur, 2010). Second, the site of stimulation was always anterior to the

central sulcus on the subject's own 3D-MRI. Finally, suprathreshold stimulation always triggered

localised hand electromyographic responses and never induced paresthesiae or other somatosensory

sensations. The reasons for such care in avoiding any direct stimulation of the sensory cortex by

rTMS lie in the potentially adverse effects of S1 stimulation on pain perception, which will be

discussed later.

The plasticity observed here took place over a very short amount of time (20 minutes).

While its duration was not measured, other studies suggest that the effects of one session of rTMS

are detectable a few tens of minutes after stimulation (Eisenegger et al., 2008; Gerschlager et al.,

2001). The cellular mechanisms involved are very likely to be transient modifications of excitation

and inhibition through a change of activity in existing structures rather than structural modifications

per se (Xerri, 2008). The details of representational plasticity are however far from being

understood and extensive research in many domains is needed to draw a better picture of its

mechanisms and consequences (Siebner & Rothwell, 2003; Thut & Pascual-Leone, 2010; Xerri,

2008).

Functional relevance of S1 plasticity

It is currently admitted that somatosensory maps are continuously modified by a subject's

experiences (for a review, see Xerri, 2008); this plasticity can lead to enlargement or shrinkage of

the cortical representation of a body part. For instance, some situations, such as intensive training

26

(Elbert, et al. 1995), simple allocation of attention to a tactile stimulus (Buchner et al., 2000; Iguchi

et al., 2001; Noppeney et al., 1999) or high-frequency rTMS of S1 (Tegenthoff et al., 2005) can lead

to an expansion of a body part cortical representation. On the contrary, a diminution of body part

representation was shown to occur in some populations of patients, including those suffering

amputation and deafferentation (Flor et al., 1995), or focal dystonia (Mogilner et al., 1993). In our

study, the possibility of plasticity being caused by an attentional shift, or by simple tactile activation

was controlled for by a sham control condition, in which subjects displayed no plasticity. We found

an enlargement of hand somatosensory representation after active high-frequency motor cortex

rTMS, a procedure known to relieve neuropathic pain in patients (for a review, see Lefaucheur,

2011). Other studies found that a number of pain-relieving procedures different from rTMS were

also followed by enhancement (i.e. normalisation) of previously altered somatotopic maps in

patients suffering from phantom-limb pain (Birbaumer et al., 1997; Huse et al., 2001; Lotze et al.,

1999). Moreover, a parallel project in our laboratory (Houzé et al, submitted) performed

quantitative sensory testing of the subjects before and after rTMS. Nociceptive thresholds were

increased after the rTMS session, therefore indicating a state of hypo-algesia following motor

cortex rTMS. Some association may exist, therefore, between the expansion of S1 hand

representation following motor cortex rTMS, and a modification in pain sensitivity towards

hypoalgesia in both healthy subjects and pain patients.

However, as tempting as it might be to consider that the above findings indicate a causal

link between S1 somatotopic plasticity and pain relief, no such relationship has been formally

proven and it can be doubted on several grounds. First, a similar S1 hand representation expansion

can be evidenced after high-frequency rTMS to the somatosensory, rather than motor cortex

(Tegenthoff et al., 2005) are far from being antalgic. For instance, attention directed to a limb

enhances pain perception on that limb (Garcia-Larrea, et al.,1997; Miron et al., 1989). rTMS over

S1 for the treatment of pain has been shown to be at the best ineffective (Hirayama et al., 2006) and

may even enhance pain (Tsubokawa et al., 1991, 1993). Second, it is likely that primary

somatosensory cortex is not a crucial region for the processing of nociceptive afferents. Indeed, less

than 4% of spinothalamic system projections may reach S1, while more than 70% of them project to

the opercular and posterior insular cortices (Dum et al., 2009). In accordance, selective lesions of

S1 rarely entail deficits in pain perception in humans (Kim, 2007), while operculo-insular injury

consistently attenuates nociception (Garcia-Larrea et al., 2010; Greenspan et al., 1999). Finally, a

recent study of somato-motor cortex reorganisation in amputees reported somatosensory

reorganisation regardless of development of phantom limb pain (Simoes et al., 2012). Therefore, the

relationship between map plasticity and lowered pain threshold is less likely to be mechanistic than

27

to reflect two parallel processes elicited by motor cortex rTMS. Functional changes within S1 might

then be viewed as a marker of the ability of rTMS to exert influences over cortical networks.

The size of S1 hand representation could however be related to tactile acuity. Indeed,

enlargement of hand representation in S1 by various manipulations can sometimes be correlated

with an improvement in tactile performance (Pleger et al., 2001; Tegenthoff et al., 2005). On the

contrary, a reduced hand representation seems associated with impaired tactile performance (Sterr et

al., 1998). There is also evidence for improved tactile performance in patients treated for

neuropathic pain by epidural motor cortex stimulation (Drouot et al., 2002). This suggests that high-

frequency rTMS of the motor cortex has -at least- two types of consequences. One is the

modification of S1 somatotopy, which might impact on tactile acuity. The other is pain relief, that

could be brought about via functional changes at more distant sites. Such sites include the

parasylvian operculo-insular cortex (S2, insula) which receives direct projections from the motor

cortex (Leichnetz, 1986) and probably cortico-thalamo-cortical loops (see above). Future

investigation will therefore be oriented towards evaluating high-frequency motor cortex rTMS-

evoked plasticity in the parasylvian operculo-insular cortex. Promising technological developments

might also allow to magnetically stimulate this region and study its impact on pain perception.

Lack of on-line assessment of ongoing plasticity

The methods described above allowed us to establish that plasticity in S1 occurred after 20

Hz rTMS over the motor cortex. However, this same methodology did not allow us to unravel the

time-course of this plasticity (see Results, section B III).

The main difference between the successful and unsuccessful situations seems to lie in the

signal-to-noise ratio of the modelled SEP, the principal factor of which -as explained above- is very

likely the number of stimulations averaged in each run. Indeed, we managed to separate ulnar nerve

and thumb dipoles and to assess their movement after rTMS, when using ~2000 peripheral

stimulations at each site. On the contrary, it was impossible to model adequately SEPs obtained

from 200 stimulations. Even combining several short runs to reach 1000 stimulations did not allow

appropriate modelling of SEPs, suggesting that the limits of the modelling method were reached in

terms of how few stimulations could be used to reconstruct sources.

Few options are available to increase signal quality through number of stimulations. Because

the SEP recording time is conditioned by the rhythm of rTMS delivery, it does not seem possible to

increase recording time. Indeed, recording during 20 Hz rTMS stimulation would require special

equipment, introduce enormous artefacts (Siebner et al., 2009) and would not bring much more

information (gain of 15 stimulations over 5 seconds). The only other option would be to increase

the rate of delivery of peripheral stimuli. It has been shown that N20 is unhindered by an increase in

28

stimulation rate up to 5 Hz (Garcia Larrea et al., 1992). However, combination of 3 short runs of

SEPs would still be required to reach more than 1000 stimuli, thus limiting temporal resolution. A

few other minor adjustments could be made to increase the signal-to-noise ratio. For example,

subjects could be given a pharmacological product that reduces muscle artefacts but does not affect

somatosensory N20 (Loughnan et al., 1987), such as diazepam (Buchner et al., 1995). However, this

procedure is quite invasive and requires considerable organisation for the subject's safety. Blink

artefacts could be modelled and removed from the model instead of rejecting the contaminated

epochs (Lins et al., 1993). Finally, even though every possible care was taken to reduce electro-

magnetic noise, data could be acquired in an electrically-shielded room. But it is likely that these

last few adjustments would not impact the results as much as the number of stimulations. The

limitations found here therefore drastically reduce the interest of using this method to study the

time-course of rTMS-induced plasticity.

Even if accurate modelling of sources had been achieved during rTMS stimulation, there is

no guarantee that a coherent time-course could have been unmasked. Indeed, inter-individual

differences were marked : among our 8 participants, some showed very little plasticity while others

presented a large cortical reorganisation (Figure 6B). Important inter-individual variability in

susceptibility to plastic reorganisation of the cortex can be attributed to many factors, such as age,

attentional state, circadian rhythm, level of cortical excitability previous to stimulation (Ridding &

Ziemann, 2010; Sale et al., 2010), and maybe also genetic factors (Cheeran et al., 2008; Missitzi et

al., 2011). These differences were not problematic in determining the existence of a plasticity. But if

differences did exist not only in the extent and direction but also in the time-course of the plasticity,

any temporal discontinuities would have been masked in the grand-average data. This stresses the

importance of individual modelling. A solution to this issue could be to study thoroughly a small

number of subjects in whom plastic changes are high, or to investigate a large enough number of

subjects that could allow grouping into sub-categories. However, it is important to note that the

number of subjects participating in this study was large enough to detect rTMS-induced plasticity,

and is definitely within the range used in studies of somatosensory representation by a dipole

modelling approach (Buchner et al., 2000; Elbert et al., 1995; Noppeney et al., 1999; Sterr et al.,

1998; Ziemus et al., 2000).

29

Conclusion

High-frequency repetitive transcranial magnetic stimulation (HF-rTMS) of the motor cortex

is one of the few procedures that can bring relief to patients suffering from drug-resistant

neuropathic pain. In an effort to understand its basic mechanisms better, we investigated the effects

of a session of HF-rTMS of the motor cortex in healthy subjects, using source modelling of

somatosensory evoked potentials (SEPs). We asked whether or not somatotopic hand representation

in primary somatosensory cortex S1 would display plasticity, and if so, whether or not its time-

course could be studied using the same methods. We found that 20 minutes of HF-rTMS over the

hand motor cortex significantly increased the size of hand representation in S1. A session of sham

(placebo) stimulation failed to do so. The methods developed, although optimised for this protocol,

did not allow us to study the time-course of such plasticity. Plasticity of the somatosensory cortex

following a treatment inducing pain relief in patients has sometimes been taken as an indication that

pain relief was mediated by S1 plasticity. However, our results were obtained in healthy subjects

and extrapolation to patients suffering from chronic spontaneous pain should be done with caution.

Moreover, no causal link between S1 plasticity and pain relief was established. Rather, it is likely

that both phenomena are parallel consequences of HF-rTMS of the motor cortex. S1 plasticity

might be more strongly linked to tactile acuity, while pain relief could be mediated by structures

such as the insula-opercular cortex. Future efforts will therefore be devoted to investigating

plasticity of operculo-insular cortex following HF-rTMS to the motor cortex.

30

References

Allison, T., McCarthy, G., Wood, C. C., Darcey, T. M., Spencer, D. D., & Williamson, P. D. (1989).

Human cortical potentials evoked by stimulation of the median nerve. I. Cytoarchitectonic

areas generating short-latency activity. Journal of Neurophysiology, 62(3), 694-710.

André-Obadia, N., Mertens, P., Gueguen, A., Peyron, R., & Garcia-Larrea, L. (2008). Pain relief by

rTMS: differential effect of current flow but no specific action on pain subtypes. Neurology,

71(11), 833-40.

André-Obadia, N., Peyron, R., Mertens, P., Mauguière, F., Laurent, B., & Garcia-Larrea, L. (2006).

Transcranial magnetic stimulation for pain control. Double-blind study of different frequencies

against placebo, and correlation with motor cortex stimulation efficacy. Clinical

neurophysiology, 117(7), 1536-44.

Azila Noh, N., & Fuggetta, G. (2011). Human cortical theta reactivity to high-frequency repetitive

transcranial magnetic stimulation. Human Brain Mapping, 000.

Baumgartner, C., Doppelbauer, a, Sutherling, W. W., Zeitlhofer, J., Lindinger, G., Lind, C., &

Deecke, L. (1991). Human somatosensory cortical finger representation as studied by

combined neuromagnetic and neuroelectric measurements. Neuroscience letters, 134(1),103-8.

Baumgartner, C., Doppelbauer, A., Deecke, L., Barth, D. S., Zeitlhofer, J., Lindinger, G., &

Sutherling, W. W. (1991). Neuromagnetic investigation of somatotopy of human hand

somatosensory cortex. Experimental Brain Research, 87(3), 641-648.

Baumgartner, C., Doppelbauer, A., Sutherling, W. W., Lindinger, G., Levesque, M. F., Aull, S.,

Zeitlhofer, J., et al. (1993). Somatotopy of human hand somatosensory cortex as studied in

scalp EEG. Electroencephalography and Clinical Neurophysiology, 88(4), 271-279.

Bestmann, S., Baudewig, J., Siebner, H., Rothwell, J., & Frahm, J. (2003). Subthreshold high-

frequency TMS of human primary motor cortex modulates interconnected frontal motor areas

as detected by interleaved fMRI-TMS. NeuroImage, 20, 1685-1696.

Birbaumer, N., Lutzenberger, W., Montoya, P., Larbig, W., Unertl, K., Töpfner, S., Grodd, W., et al.

(1997). Effects of regional anesthesia on phantom limb pain are mirrored in changes in cortical

reorganization. The Journal of Neuroscience, 17(14), 5503-8.

31

Brinkmann, B. H., O’Brien, T. J., Dresner, M. a, Lagerlund, T. D., Sharbrough, F. W., & Robb, R. A.

(1998). Scalp-recorded EEG localization in MRI volume data. Brain topography, 10(4), 245-

53.

Buchner, H., Adams, L., Müller, A., Ludwig, I., Knepper, A., Thron, A., Niemann, K., et al. (1995).

Somatotopy of human hand somatosensory cortex revealed by dipole source analysis of early

somatosensory evoked potentials and 3D-NMR tomography. Electroencephalography and

clinical Neurophysiology, 96, 121-134.

Buchner, H., Richrath, P., Grünholz, J., Noppeney, U., Waberski, T. D., Gobbelé, R., Willmes, K., et

al. (2000). Differential effects of pain and spatial attention on digit representation in the human

primary somatosensory cortex. Neuroreport, 11(6), 1289-93.

Cheeran, B., Talelli, P., Mori, F., Koch, G., Suppa, A., Edwards, M., Houlden, H., et al. (2008). A

common polymorphism in the brain-derived neurotrophic factor gene (BDNF) modulates

human cortical plasticity and the response to rTMS. The Journal of Physiology, 586, 5717-25.

Cosentino, G., Fierro, B., Paladino, P., Talamanca, S., Vigneri, S., Palermo, A., Giglia, G., et al.

(2012). Transcranial direct current stimulation preconditioning modulates the effect of high-

frequency repetitive transcranial magnetic stimulation in the human motor cortex. The

European Journal of Neuroscience, 35(1), 119-24.

Cruccu, G., Aminoff, M., Curio, G., Guerit, J., Kakigi, R., Mauguière, F., Rossini, P., et al. (2008).

Recommendations for the clinical use of somatosensory-evoked potentials. Clinical

Neurophysiology, 119, 1705-1719.

Drouot, X., Nguyen, J.P., Peschanski, M., & Lefaucheur, J.-P. (2002). The antalgic efficacy of

chronic motor cortex stimulation is related to sensory changes in the painful zone. Brain, 125,

1660-4.

Dum, R., Levinthal, D., & Strick, P. (2009). The spinothalamic system targets motor and sensory

areas in the cerebral cortex of monkeys. The Journal of Neuroscience, 29(45), 14223-14235.

Eisenegger, C., Treyer, V., Fehr, E., & Knoch, D. (2008). Time-course of “off-line” prefrontal rTMS

effects — a PET study. NeuroImage, 42, 379-384.

32

Elbert, T., Pantev, C., Wienbruch, C., Rockstroh, B., & Taub, E. (1995). Increasd Cortical

Represntation of the Fingers of the Left Hand in String Players. Science, 270(October), 305-

307.

Flor, H., Elbert, T., Knecht, S., Wienbruch, C., Pantev, C., Birbaumer, N., Larbig, W., et al. (1995).

Phantom-limb pain as a perceptual correlate of cortical reorganization following arm

amputation. Nature, 375(6531), 482-484.

Flor, H., Nikolajsen, L., & Staehelin Jensen, T. (2006). Phantom limb pain: a case of maladaptive

CNS plasticity? Nature reviews. Neuroscience, 7(11), 873-81.

Fox, P., Ingham, R., George, M., Mayberg, H., Ingham, J., Roby, J., Martin, C., et al. (1997).

Imaging human intra-cerebral connectivity by PET during TMS. NeuroReport, 8(12), 2787-

2791.

Freeman, W. J., Holmes, M. D., Burke, B. C., & Vanhatalo, S. (2003). Spatial spectra of scalp EEG

and EMG from awake humans. Clinical Neurophysiology, 114(6), 1053-1068.

doi:10.1016/S1388-2457(03)00045-2

Gamboa, O., Antal, A., Moliadze, V., & Paulus, W. (2010). Simply longer is not better: reversal of

theta burst after-effect with prolonged stimulation. Experimental Brain Research, 204, 181-

187.

Garcia-Larrea, L., Bastuji, H., & Mauguière, F. (1992). Unmasking of cortical SEP components by

changes in stimulus rate: a topographic study. Electroencephalography and Clinical

Neurophysiology, 84, 71-83.

Garcia-Larrea, L., & Peyron, R. (2007). Motor cortex stimulation for neuropathic pain: From

phenomenology to mechanisms. NeuroImage, 37 Suppl 1, S71-9.

Garcia-Larrea, L., Peyron, R., Laurent, B., & Mauguière, F. (1997). Association and dissociation

between laser-evoked potentials and pain perception. NeuroReport, 8(17), 3785-9.

Garcia-Larrea, Luis, Perchet, C., Creac’h, C., Convers, P., Peyron, R., Laurent, B., Mauguière, F., et

al. (2010). Operculo-insular pain (parasylvian pain): a distinct central pain syndrome. Brain,

133(9), 2528-2539.

Gerschlager, W., Siebner, H., & Rothwell, J. (2001). Decreased corticospinal excitability after

33

subthreshold 1 Hz rTMS over lateral premotor cortex. Neurology, 57(August (1 of 2)), 449-

455.

Godinho, F., Magnin, M., Frot, M., Perchet, C., & Garcia-Larrea, L. (2006). Emotional modulation

of pain: is it the sensation or what we recall? The Journal of Neuroscience, 26(44), 11454-61.

Greenspan, J. D., Lee, R. R., & Lenz, F. A. (1999). Pain sensitivity alterations as a function of

lesion location in the parasylvian cortex. Pain, 81(3), 273-282.

Grüsser, S. M., Winter, C., Mühlnickel, W., Denke, C., Karl, A., Villringer, K., & Flor, H. (2001).

The relationship of perceptual phenomena and cortical reorganization in upper extremity

amputees. Neuroscience, 102(2), 263-272.

Hamburger, H., & Van der Burgt, A. (1991). Global Field Power Measurement Versus Classical

Method in the Determination of the Latency of Evoked Potential Components. Brain

Topography, 3(3), 391-396.

Hari, R., Karhu, J., Hämäläinen, M., Knuutila, J., Salonen, O., Sams, M., & Vilkman, V. (1993).

Functional organization of the human first and second somatosensory cortices: a

neuromagnetic study. European Journal of Neuroscience, 5(6), 724-734.

Hirayama, A., Saitoh, Y., Kishima, H., Shimokawa, T., Oshino, S., Hirata, M., Kato, A., et al.

(2006). Reduction of intractable deafferentation pain by navigation-guided repetitive

transcranial magnetic stimulation of the primary motor cortex. Pain, 122(1-2), 22-7.

Hosomi, K., Saitoh, Y., Kishima, H., Oshino, S., Hirata, M., Tani, N., Shimokawa, T., et al. (2008).

Electrical stimulation of primary motor cortex within the central sulcus for intractable

neuropathic pain. Clinical Neurophysiology, 119, 993-1001.

Houzé, B., Perchet, C., Magnin, M., & Garcia-Larrea, L. (2011). Cortical representation of the

human hand assessed by two levels of high-resolution EEG recordings. Human brain mapping,

32(11), 1894-904.

Huse, E., Larbig, W., Flor, H., & Birbaumer, N. (2001). The effect of opioids on phantom limb pain

and cortical reorganization. Pain, 90, 47-55.

Iguchi, Y., Hoshi, Y., & Hashimoto, I. (2001). Selective spatial attention induces short-term

plasticity in human somatosensory cortex. Neuroreport, 12(14), 3133-6.

34

Karl, A., Birbaumer, N., Lutzenberger, W., Cohen, L. G., & Flor, H. (2001). Reorganization of

motor and somatosensory cortex in upper extremity amputees with phantom limb pain.

Journal of Neuroscience, 21(10), 3609-3618.

Kim, J.S. (2007). Patterns of sensory abnormality in cortical stroke: evidence for a dichotomized

sensory system. Neurology. 68:174-180.

Krubitzer, L., & Kaas, J. (1990). The Organization Marmosets and Connections of Somatosensory

Cortex in. The Journal of Neuroscience, 10(3), 952-974.

Kultas-Ilinsky, K., Sivan-Loukianova, E., & Ilinsky, I. A. (2003). Reevaluation of the primary

motor cortex connections with the thalamus in primates. Journal of Comparative Neurology,

457(2), 133-158.

Kurth, R., Villringer, K., Mackert, B.M., Schwiemenn, J., Braun, J., Curio, G., Villringer, A., &

Wolf, K.J. (1998). fMRI assessment of somatotopy in human Brodmann area 3b by electrical

finger stimulation. NeuroReport, 9(2), 207-12.

Lee, L., Siebner, H. R., Rowe, J. B., Rizzo, V., Rothwell, J. C., Frackowiak, R. S. J., & Friston, K. J.

(2003). Acute remapping within the motor system induced by low-frequency repetitive

transcranial magnetic stimulation. The Journal of Neuroscience, 23(12), 5308-18.

Lee, S., Carvell, G., & Simons, D. (2008). Motor modulation of afferent somatosensory circuits.

Nature Neuroscience, 11(12), 1430-1438.

Lefaucheur, J.P. (2006). The use of repetitive transcranial magnetic stimulation ( rTMS ) in chronic

neuropathic pain. Neurophysiologie clinique, 36, 117-124.

Lefaucheur, J.P. (2010). Why image-guided navigation becomes essential in the practice of

transcranial magnetic stimulation. Neurophysiologie clinique, 40, 1-5.

Lefaucheur, J.P. (2011). Is rTMS a therapeutic option in chronic pain syndrome ? Insights from the 

treatment of fibromyalgia. Pain, 152, 1447-1448.

Leichnetz, G. R. (1986). Afferent and efferent connections of the dorsolateral precentral gyrus (area

4, hand/arm region) in the macaque monkey, with comparisons to area 8. Journal of

Comparative Neurology, 254(4), 460-492.

35

Leo, R., & Latif, T. (2007). Repetitive Transcranial Magnetic Stimulation (rTMS) in Experimentally

Induced and Chronic Neuropathic Pain: A Review. The Journal of Pain, 8(6), 453-459.

Lins, O. G., Picton, T. W., Berg, P., & Scherg, M. (1993). Ocular artifacts in recording EEGs and

event-related potentials. II: Source dipoles and source components. Brain Topography, 6(1),

65-78.

Lotze, M., Grodd, W., Birbaumer, N., Erb, M., Huse, E., & Flor, H. (1999). Does use of a

myoelectric prosthesis prevent cortical reorganization and phantom limb pain ?   Nature

Neuroscience, 2(6), 501-502.

Loughnan, B. L., Sebel, P. S., Thomas, D., Rutherfoord, C. F., & Rogers, H. (1987). Evoked

potentials following diazepam or fentanyl. Anaesthesia, 42(2), 195-198.

Maldjian, J.A., Gottschalk, A., Patel, R.S., Detre, J.A., & Alsop, D.C. (1999). The sensory

somatotopic map of the human hand demonstrated at 4 Tesla. NeuroImage, 10, 55-62.

Mauguière F. (1995). Description and origins of SEP components. In: Binnie C, et al. editor.

Clinical neurophysiology. Vol. 1. Oxford: Butterworth-Heinemann; 1995.

Miron, D., Duncan, G. H., & Bushnell, M. C. (1989). Effects of attention on the intensity and

unpleasantness of thermal pain. Pain, 39(3), 345-352.

Missitzi, J., Gentner, R., Geladas, N., Politis, P., Karandreas, N., Classen, J., & Klissouras, V.

(2011). Plasticity in human motor cortex is in part genetically determined. The Journal of

physiology, 589(Pt 2), 297-306.

Mogilner, A, Grossman, J. A, Ribary, U., Joliot, M., Volkmann, J., Rapaport, D., Beasley, R. W., &

Llinas, R.R. (1993). Somatosensory cortical plasticity in adult humans revealed by

magnetoencephalography. Proceedings of the National Academy of Sciences of the United

States of America, 90(8), 3593-7.

Montoya, P., Ritter, K., Huse, E., Larbig, W., Braun, C., Töpfner, S., Lutzenberger, W., et al. (1998).

The cortical somatotopic map and phantom phenomena in subjects with congenital limb

atrophy and traumatic amputees with phantom limb pain. European Journal of Neuroscience,

10, 1095-1102.

Münchau, A., Bloem, B., Irlbacher, K., Trimble, M., & Rothwell, J. (2002). Functional Connectivity

36

of Human Premotor and Motor Cortex Explored with Repetitive Transcranial Magnetic

Stimulation. The Journal of Neuroscience, 22(2), 554-561.

Nakamura, a, Yamada, T., Goto, a, Kato, T., Ito, K., Abe, Y., Kachi, T., et al. (1998). Somatosensory

homunculus as drawn by MEG. NeuroImage, 7(4 Pt 1), 377-86.

Napadow, V., Liu, J., Li, M., Kettner, N., Ryan, A., Kwong, K., Hui, K., et al. (2007).

Somatosensory Cortical Plasticity in Carpal Tunnel Syndrome Treated by Acupuncture.

Human Brain Mapping, 28, 159-171.

Noppeney, U., Waberski, T. D., Gobbelé, R., & Buchner, H. (1999). Spatial attention modulates the

cortical somatosensory representation of the digits in humans. NeuroReport, 10(15), 3137-41.

Oostendorp, T. F., Delbeke, J., & Stegeman, D. F. (2000). The conductivity of the human skull:

results of in vivo and in vitro measurements. IEEE Transactions on Biomedical Engineering,

47(11), 1487-1492.

Papadelis, C., Eickhoff, S., Zilles, K., & Ioannides, A. (2011). BA3b and BA1 activate in a serial

fashion after median nerve stimulation : Direct evidence from combining source analysis of 

evoked fields and cytoarchitectonic probabilistic maps. NeuroImage, 54, 60-73.

Perrin, F., Pernier, J., Bertrand, O., Echallier, J., Inserm, U., Thomas, C. A., & France, L. (1989).

Spherical splines for scalp potential and current density mapping. Electroencephalography and

clinical Neurophysiology, 72, 184-187.

Pleger, B., Dinse, H., Ragert, P., Schwenkreis, P., Malin, J., & Tegenthoff, M. (2001). Shifts in

cortical representations predict human discrimination improvement. Proceedings of the

National Academy of Sciences of the United States of America, 98(21), 12255-12260.

Ridding, M. C., & Ziemann, U. (2010). Determinants of the induction of cortical plasticity by non-

invasive brain stimulation in healthy subjects. The Journal of physiology, 588(Pt 13), 2291-

304.

Ridding, M., & Rothwell, J. (2007). Is there a future for therapeutic use of transcranial magnetic

stimulation? Nature Reviews Neuroscience, 8(July), 559-567.

Rossini, P. M., Barker, A. T., Berardelli, A., Caramia, M. D., Caruso, G., Cracco, R. Q.,

Dimitrijević, M. R., et al. (1994). Non-invasive electrical and magnetic stimulation of the

37

brain, spinal cord and roots: basic principles and procedures for routine clinical application.

Report of an IFCN committee. Electroencephalography and Clinical Neurophysiology, 91(2),

79-92.

Rouiller, E., & Welker, E. (2000). A comparative analysis of the morphology of corticothalamic

projections in mammals. Brain Research Bulletin, 53(6), 727-741.

Ryynänen, O., Hyttinen, J., & Malmivuo, J. (2005). Effect of skull resistivity and measurement

noise on the spatial resolution of EEG. International Journal of Bioelectromagnetism, 7(1),

317-320.

Sack, A. T., Kohler, A., Bestmann, S., Linden, D. E. J., Dechent, P., Goebel, R., & Baudewig, J.

(2007). Imaging the Brain Activity Changes Underlying Impaired Visuospatial Judgments : 

Simultaneous fMRI , TMS , and Behavioral Studies, (December), 2841-2852.

Sale, M. V., Ridding, M. C., & Nordstrom, M. A. (2010). Circadian modulation of neuroplasticity in

humans and potential therapeutic implications. Reviews in the Neurosciences, 21(1), 55-66.

Schaefer, M., Mühlnickel, W., Grüsser, S. M., & Flor, H. (2002a). Reproducibility and stability of

neuroelectric source imaging in primary somatosensory cortex. Brain topography, 14(3), 179-

89.

Schaefer, M., Mühlnickel, W., Grüsser, S. M., & Flor, H. (2002b). Reliability and validity of

neuroelectric source imaging in primary somatosensory cortex of human upper limb amputees.

Brain topography, 15(2), 95-106.

Scherg, M. (1990). Fundamentals of Dipole Source Potential Analysis. Advances in Audiology, 6,

40-69.

Scherg, M., & Picton, T. W. (1991). Separation and identification of event-related potential

components by brain electric source analysis. Electroencephalography And Clinical

Neurophysiology Supplement, 42(Suppl), 24-37.

Schwenkreis, P, Witscher, K., Janssen, F., Pleger, B., Dertwinkel, R., Zenz, M., Malin, J. P., et al.

(2001). Assessment of reorganization in the sensorimotor cortex after upper limb amputation.

Clinical Neurophysiology, 112(4), 627-635.

Schwenkreis, Peter, El Tom, S., Ragert, P., Pleger, B., Tegenthoff, M., & Dinse, H. R. (2007).

38

Assessment of sensorimotor cortical representation asymmetries and motor skills in violin

players. The European journal of neuroscience, 26(11), 3291-302.

Siebner, H., Bergmann, T., Bestmann, S., Massimini, M., Johansen-Berg, H., Mochizuki, H.,

Bohning, D., et al. (2009). Consensus paper : Combining transcranial stimulation with 

neuroimaging. Brain Stimulation, 2, 58-80.

Siebner, H. R., & Rothwell, J. (2003). Transcranial magnetic stimulation: new insights into

representational cortical plasticity. Experimental Brain Research, 148(1), 1-16.

Simoes, E. L., Bramati, I., Rodrigues, E., Franzoi, A., Moll, J., Lent, R., & Tovar-Moll, F. (2012).

Functional Expansion of Sensorimotor Representation and Structural Reorganization of

Callosal Connections in Lower Limb Amputees. The Journal of Neuroscience, 32(9), 3211-

3220.

Skrandies, W. (1990). Global field power and topographic similarity. Brain topography, 3(1), 137-

41.

Srinivasan, R., Tucker, D. M., & Murias, M. (1998). Estimating the spatial Nyquist of the human

EEG. Behavior Research Methods, 30(1), 8-19.

Sterr, A., Müller, M. M., Elbert, T., Rockstroh, B., Pantev, C., & Taub, E. (1998). Perceptual

correlates of changes in cortical representation of fingers in blind multifinger Braille readers.

The Journal of Neuroscience, 18(11), 4417-23.

Tegenthoff, M., Ragert, P., Pleger, B., Schwenkreis, P., Förster, A. F., Nicolas, V., & Dinse, H. R.

(2005). Improvement of tactile discrimination performance and enlargement of cortical

somatosensory maps after 5 Hz rTMS. PLoS biology, 3(11), e362.

Thut, G., & Pascual-Leone, A. (2010). A Review of Combined TMS-EEG Studies to Characterize

Lasting Effects of Repetitive TMS and Assess Their Usefulness in Cognitive and Clinical

Neuroscience. Brain Topography, 22, 219-232.

Toschi, N., Welt, T., Guerrisi, M., & Keck, M. E. (2008). A reconstruction of the conductive

phenomena elicited by transcranial magnetic stimulation in heterogeneous brain tissue.

Physica medica, 24(2), 80-6.

Treede, R., Jensen, T., Campbell, J., Cruccu, G., Dostrovsky, J., Griffin, J., Hansson, P., et al.

39

(2008). Neuropathic pain: redefinition and a grading system for clinical and research purposes.

Neurology, 70(April), 1630-1635.

Tsubokawa, T., Katayama, Y., Yamamoto, T., Hirayama, T., & Koyama, S. (1991). Chronic motor

cortex stimulation for the treatment of central pain. Acta Neurochir Suppl, 52, 137-139.

Tsubokawa, T., Katayama, Y., Yamamoto, T., Hirayama, T., & Koyama, S. (1993). Chronic motor

cortex stimulation in patients with thalamic pain. Journal of Neurosurgery, 78(3), 393-401.

Waberski, T. ., Gobbelé, R., Kawohl, W., Cordes, C., & Buchner, H. (2003). Immediate cortical

reorganization after local anesthetic block of the thumb: source localization of somatosensory

evoked potentials in human subjects. Neuroscience Letters, 347(3), 151-154.

Wagner, T., Rushmore, J., & Eden, U. (2009). Biophysical foundations underlying TMS: setting the

stage for an effective use of neurostimulation in the cognitive neurosciences. Cortex, 45(9),

1025-1034.

Weibull, A., Flondell, M., Rosén, B., & Björkman, A. (2011). Cerebral and clinical effects of short-

term hand immobilisation. The European Journal of Neuroscience, 33(4), 699-704.

van Westen, D., Fransson, P., Olsrud, J., Rosen, B., Lundborg, G., Larsson, E.M. (2004).

Fingersomatotopy in area 3b: an fMRI-study. BMC Neuroscience 2004, 5-28.

Wrigley, P. J., Press, S. R., Gustin, S. M., Macefield, V. G., Gandevia, S. C., Cousins, M. J.,

Middleton, J. W., et al. (2009). Neuropathic pain and primary somatosensory cortex

reorganization following spinal cord injury. Pain, 141(1-2), 52-59.

Xerri, C. (2008). Imprinting of idiosyncratic experience in cortical sensory maps: neural substrates

of representational remodeling and correlative perceptual changes. Behavioural brain

research, 192(1), 26-41.

Yousry, T. a, Schmid, U. D., Alkadhi, H., Schmidt, D., Peraud, a, Buettner, a, & Winkler, P. (1997).

Localization of the motor hand area to a knob on the precentral gyrus. A new landmark. Brain,

120, 141-57.

Ziemann, U., Paulus, W., Nitsche, M., Pascual-Leone, A., Byblow, W., Berardelli, A., Siebner, H., et

al. (2008). Consensus: Motor cortex plasticity protocols. Brain Stimulation, 1, 164-182.

Ziemus, B., Huonker, R., Haueisen, J., Liepert, J., Spengler, F., & Weiller, C. (2000). Effects of

40

passive tactile co-activation on median ulnar nerve representation in human SI. Neuroreport,

11(6), 1285-8.

41