Abstract—Electrical cortical stimulation (CS) of the auditory
cortices has been shown to reduce the severity of debilitating
tinnitus in some patients. In this study, we performed MEG
source imaging of spontaneous brain activity during concurrent
CS of the left secondary auditory cortex of a volunteer
suffering from right unilateral tinnitus. CS produced MEG
artifacts which were successfully sorted and removed using a
combination of sensor and source level signal separation and
classification techniques. This contribution provides the first
proof of concept reporting on analysis of MEG data with
concurrent CS. Effects of CS on ongoing brain activity were
revealed at the MEG sensor and source levels and indicate CS
significantly reduced ongoing brain activity in the lower
frequency range (<40Hz), and emphasized its higher (>40Hz),
gamma range components. Further, our results show that CS
increased the spectral correlation across multiple frequency
bands in the low and high gamma ranges, and between the
alpha and beta bands of the MEG. Finally, MEG sources
localized in the auditory cortices and nearby regions exhibited
abnormal spectral activity that was suppressed by CS. These
results provide promising evidence in favor of the
Thalamocortical Dysrhytmia (TCD) hypothesis of tinnitus, and
suggest that CS may prove to be an effective treatment of
tinnitus when targeted to brain regions exhibiting abnormal
INNITUS is the perception of an auditory stimulus in
the absence of external environmental sounds.
Subjective tinnitus is a subtype generated in the
sensorineural auditory system and cannot be heard by an
examiner. About 15% of adults have prolonged tinnitus
requiring medical evaluation .
Subjective tinnitus has been hypothesized to originate in
many areas of the central nervous system, including the
cochlea, eighth cranial nerve, brain stem, or cortex, but in
practice, there are presently no means to reliably
discriminate between these possible origins in a patient.
Though the pathophysiology of tinnitus remains poorly
understood, there is indirect evidence to suggest that the
majority of subjective tinnitus cases are associated with
Manuscript received April 23, 2009
All authors are with the Medical College of Wisconsin, Milwaukee, WI
S. Baillet, C.R. Butson and R.R. Ramirez are with the Dept of Neurology
(phone: 414-805-1104; fax: 414-805-1103; e-mail: email@example.com,
B.H. Kopell is with the Dept of Neurosurgery (e-mail: firstname.lastname@example.org).
W. Gaggl is with the Dept of Radiology (e-mail: email@example.com).
D.R. Friedland is with the Dept of Otolaryngology (e-mail:
partial deafferentation of the auditory system. This typically
occurs via cochlea hair cell loss due to aging, noise, or
ototoxic medications. A secondary effect of peripheral
deafferentation is central auditory compensation. It is
hypothesized that plastic changes within central auditory
nuclei try to compensate for reduced peripheral input but
secondarily lead to spontaneous neural activity and the
percept of sound (i.e., tinnitus).
Although there have been many approaches to treating
tinnitus, none have been effective in meaningfully
alleviating symptoms. A variety of drugs have been tried on
tinnitus patients with poor results. Masking devices and
behavioral training techniques (e.g., biofeedback and
cognitive therapy) also have had limited success. Electrical
stimulation of the cortex has been used recently to treat
tinnitus symptoms. DeRidder et al. have extended their
seminal findings using this technique to a 12-patient study
with over 50% of the patients benefiting from the therapy
, . Cortical stimulation (CS) was found to be more
effective at reducing pure-tone, as opposed to white-noise,
tinnitus symptoms in patients with predominantly unilateral
tinnitus. In another study, Fenoy et al. reported on a patient
with bilateral tinnitus who experienced acute suppression of
his tinnitus during subdural electrical stimulation of the right
auditory cortex . Our group investigated the feasibility of
cortical stimulation in 8 subjects and found no immediate
responses. Rather, we identified the likelihood that chronic
electrical stimulation could induce cortical changes that
ameliorate the tinnitus percept in some individuals . It is
the findings from this study that prompt the current study to
refine the identification of central targets using multimodal
neuroimaging for tinnitus treatment.
A quantitative EEG study by Ashton et al. found discrete
localized unilateral foci of high-frequency oscillatory brain
activity in the gamma range (>40-80Hz) over the auditory
cortex specifically in 8 out of 8 patients experiencing
unilateral tinnitus during EEG recording . Interestingly,
these “hot spots” were identified in either the left or right
auditory cortices irrespectively of the laterality of the
An MEG case study by Llinas et al. using magnetic
source imaging techniques showed that the patient’s
unilateral tinnitus percept was accompanied by spontaneous
abnormal low and high frequency electromagnetic activity
that localized to the contralateral auditory cortex . This
abnormal activity was suppressed by auditory stimulation
that masked the subjective tinnitus percept and produced
Neuromagnetic Source Imaging of Abnormal Spontaneous Activity
in Tinnitus Patient Modulated by Electrical Cortical Stimulation
Rey René Ramírez, Brian Harris Kopell, Christopher R. Butson,
Wolfgang Gaggl, David R. Friedland, and Sylvain Baillet
31st Annual International Conference of the IEEE EMBS
Minneapolis, Minnesota, USA, September 2-6, 2009
978-1-4244-3296-7/09/$25.00 ©2009 IEEE
residual inhibition. Consistent with these results, Weisz et al.
found that the spontaneous electromagnetic brain activity in
tinnitus patients was reduced in the alpha range (8-12Hz)
and increased in the delta range (1.5-4Hz) in temporal brain
areas  and that the residual inhibition of tinnitus intensity
produced by masking stimuli was correlated with a reduction
of delta power, also in temporal regions . Moreover, a
neurofeedback study showed that reducing delta (0.5-4Hz)
power and enhancing tau (8-12Hz) power using EEG-based
neurofeedback abolished the tinnitus sensation in patients
that were able to modify their brain oscillations . Other
MEG studies have also suggested that tinnitus patients have
abnormal functional source connectivity during resting and
auditory processing .
These studies reporting on global electrophysiological
measures of tinnitus-related brain activity may be interpreted
along the lines of the general assumption of thalamo-cortical
dysrhythmia (TCD) , , . TCD predicts that low-
frequency thalamocortical bursting produces a center of low-
frequency oscillatory activity in the auditory cortex (as if
this local thalamocortical circuit had fallen asleep), which in
turn laterally disinhibits the surrounding auditory cortical
areas thereby causing them to fire spontaneously in the
higher frequency range (gamma/beta) of MEG/EEG signals.
According to the TCD hypothesis, the positive symptoms of
tinnitus are produced by such spontaneous abnormal high-
frequency activity in areas of the auditory cortices mapped
to the tinnitus percept spectral contents. It is further claimed
that the tinnitus percept may be suppressed by reducing
either the center-organized
oscillations that produce the disinhibition, or the surround-
organized high frequency oscillations.
This multimodal body of evidence suggests that bilateral
and unilateral tinnitus is amenable to targeted stimulation
applied at a single cortical site. Although both unilateral and
bilateral tinnitus are often cortically-lateralized, it is not
straightforward to predict on what side of the cortex the
tinnitus activity is likely to manifest based solely on the
laterality of the tinnitus perception. Therefore, rather than
relying only on patient symptoms to inform targeting of
cortical stimulation, a multimodal brain imaging approach is
A. MEG Recordings
The MEG procedure was approved by the IRB committee
at the Medical College of Wisconsin. MEG data was
acquired using the Elekta-Neuromag VectorView MEG
system (306 sensors – 102 sensor triplets, each consisting
of 2 orthogonal planar gradiometers and 1 planar
magnetometer – that measure magnetic flux at 102
positions) at the MEG Program of Froedtert Hospital/The
Medical College of Wisconsin. The subject was a tinnitus
patient who provided fully-informed consent to participate
to the study. He had been implanted with a cortical
stimulation device (Northstar Neuroscience Inc.) in the left
secondary cortical cortex. CS consisted of 40Hz pulse with
a pulse width of 150 micro-seconds. The MEG sampling
rate was set to 2KHz. The patient was instructed to stay
still with eyes open while recorded spontaneous brain
activity was recorded with CS being either turned on or off
(2 separate 5 min recordings).
B. MEG Analysis
MEG data was first processed by the signal-space
separation (SSS)  technique as available from Elekta’s
MaxFilter software. SSS applied a spatial filter to the MEG
sensor data to reduce contamination from external sources
(e.g., heart, muscles, and environmental nuisances such as
the CS device). This procedure is further imposed by the
active Maxshield (Elekta) shielding compensation solution
in place at the recording site.
The data was then band-pass filtered (2-100Hz) to remove
low-frequency drifts and high frequency noise and was
further denoised using signal classification through signal-
space projection (SSP). SSP projects the data time series
away from the subspaces spanned by noisy components as
identified in the data. Both filtering and SSP were performed
using the freely available MNE software (MGH-Harvard
Medical School). Spectral analysis of the processed data was
completed using the multitaper Fourier transform to obtain
the time-frequency spectral densities, log power spectrum,
cross-spectral density matrices, and cross-frequency
correlation indices at the MEG channel level. These MEG
spectral markers have proved useful to characterize global
oscillatory activity in multiple brain disorders, including
tinnitus , , . The processed data was then
analyzed with Independent Component Analysis (ICA) .
ICA provides a decomposition of the original measurements
into maximally independent time courses. ICA was
implemented in Matlab using modified code obtained from
the EEGLAB Toolbox. The time-frequency dynamics of the
unfiltered independent source time-series were computed
using multitaper spectral analysis as for sensor signals.
Brain-related independent sources were sorted out from
correlated noise sources both automatically by using a
matched-filter approach, and manually by inspecting the
spectrum of all independent sources.
Neuromagnetic source imaging was performed with a new
robust matched filter algorithm with anatomical priors
incorporated by only allowing activity to be constrained to
the individual segmented cortical surface. MRI segmentation
was performed from the subject’s T1 scan using Freesurfer
and MNE. The MEG forward model was obtained from the
Boundary Element Method (BEM) with isolated-skull
approach as implemented in MNE. Source estimates were
computed for all independent components. The ICA mixing
matrices were column normalized prior to localization and
the activation functions were multiplied by their respective
norms. Source estimates were visualized in inflated and
folded cortical surfaces.
A. Sensor level analysis
The spectrum of the processed MEG signals indicated that
the measurements were highly contaminated by noise
especially when the cortical stimulator was turned on. When
the CS was turned off, even minor head movements
produced strong noise artifacts due to the implanted CS
device and associated wiring. When CS was turned on, CS
produced subharmonic peaks in the spectrum because of
aliasing due to the relatively limited sampling rate compared
to the CS pulse width (150 micro-seconds). Nonetheless,
careful application of SSP was able to remove both types of
The spectrum of the denoised data showed the typical pink
spectral distribution with a peak at alpha, and broad power
in the beta and gamma ranges for both the CS on and off
conditions (see Fig. 1). A Wilcoxon rank sum test showed
that the power was significantly larger for the CS off
condition than the CS on condition for the lower frequency
bands ranging from 2 to 34Hz. In contrast, power was larger
for the CS on condition compared to the CS off condition for
the higher-frequency bands ranging between 38Hz and
100Hz, except within the 58-62Hz range (consistent 60Hz
power line contamination across conditions).
The spectral power correlation analysis (see Fig. 2)
revealed increased correlation of power across a wide
gamma (30-100Hz) range while CS was on. Increased
power correlation during CS was also found within the alpha
band range with the beta, theta and CS frequency (40Hz),
and between theta and high gamma. More power correlation
was found when CS was off between alpha and low and
high gamma (31 and 73Hz), and between theta and 17Hz,
and between beta and low gamma, and between beta and
B. Source level analysis
Since sensor signals are mixtures of sources, their time-
frequency spectral content reflects the dynamics of the
mixtures and not of the underlying sources. Thus, we also
performed time-frequency analysis of the independent
source activation functions obtained from ICA.
Neuromagnetic source imaging showed that multiple
independent components localized to brain areas known to
be spontaneously active during rest. These included mu
rhythm sources localized to somato-motor cortices and alpha
sources localized to striate and extrastriate visual cortices.
We also found several tau rhythm independent components
which localized to the left and right auditory cortices.
In Fig 3A, we show the power spectrum of a typical tau
component obtained when CS was off, which in addition to
exhibiting spontaneous oscillations in the alpha range, had
several large peaks in the theta and delta bands as expected
based on the TCD hypothesis. This source also had peaks
and broad band power in the beta and gamma bands, as
expected from a dysrhythmic source. The source estimate of
this component localized to the left auditory cortices, which
is in agreement with the subject’s experience of tinnitus to
the right side (Fig. 3B).
Fig 4A shows the spectra of four independent sources
during CS, which localized to the left auditory cortices,
insula, and nearby regions. The spectra show clear peaks in
density estimates were computed for denoised data and averaged
across 3 tapers (5s time windows with 2.5s overlap). The log mean
power was averaged across 2Hz frequency bins and across channels.
A Wilcoxon rank sum test was run to test for differences between
median log spectral power across time windows for CS off (blue) vs.
CS on (red) conditions. Plot shows medians with error bars showing
+/- standard deviations. Medians for all frequency bins were found
being significantly different (p<0.001, with Bonferroni correction)
except for bins within: [34-38], and [58-62] Hz.
Sensor level spectral analysis. Multitaper power spectral
correlation between frequency bands of the time-frequency log power
(averaged across tapers and 2Hz frequency bins) was computed for
CS on and CS off recordings. The contrast in power correlation (CS
on – CS off) is shown here thresholded following a two-tailed
permutation t-test with Bonferroni correction (p-value=8.5*10-6).
Positive correlation values (red-yellow) indicate a higher power
correlation in the CS on condition, while negative correlation values
(blue) indicate higher correlation in the CS off condition.
Sensor level power correlation analysis. The power
the alpha band that were missing from the theta/delta source
when CS was off. These sources exhibit smaller peaks in the
theta band, but these latter are clearly reduced with respect
to CS off. These sources also show power in the beta and
gamma band, and also clearly show a peak at 40Hz, the
stimulation frequency, which suggests that they might be
entrained by CS.
Finally, Fig 5 shows the location of the cortical stimulation
electrodes and the areas that were activated during an fMRI
experiment using speech babble auditory stimuli.
These early results report on both global (sensor level)
and regional (source level) effects of CS as revealed by
MEG. Abnormal theta/delta activity was found in the left
auditory cortices of the subjects when CS was off, consistent
with his experience of tinnitus on the right side and
consistent with the thalamocortical dysrhythmia hypothesis
of tinnitus. When CS was turned on, several independent
sources distributed in the left auditory cortices, insula, and
surrounding regions were found. The activity of these
sources showed spectral peaks in the alpha range suggesting
that they were tau rhythm sources, and showed decreased
theta/delta activity compared to the source identified during
no CS. This result suggests that CS can suppress oscillations
in the theta/delta bands while increasing tau oscillations. The
sources identified in the CS condition had peaks at the CS
frequency. The activity in the insular cortex may represent
the affective component that usually accompanies tinnitus.
Fig 3. Independent theta/delta auditory source during no CS. (A)
Spectrum averaged across tapers and time windows for independent
component 10 with CS turned ‘off’. (B) Source estimate of
independent component using matched filter algorithm.
Fig 4. CS-modulated independent sources during CS. (A)
Spectrum averaged across tapers and time windows for independent
components 12, 36, 71, and 81 with CS turned ‘on’. (B) Source
estimate of independent components using matched filter algorithm.
Fig 5. Location of cortical stimulation electrodes (in red) and
fMRI signal in response to speech babble (in green).
Thus, this suggests that CS may have a positive effect on
affect by reducing the theta/delta activity in the insula. One
of the sources extended to regions of the left somatosensory
cortex. This may bear some significance since the patient
reported that his tinnitus is sometimes modulated by
somatosensory stimulation of the face. Importantly, the areas
that were determined to exhibit abnormal theta/delta activity
coincide with the location of the electrodes and the areas that
were activated in the fMRI experiment. This coincides with
the MEG source imaging results reported here and suggest
that this patient responded well to CS because the
stimulating electrodes were located in thalamocortically
dysrhythmic regions of his brain.
This contribution proves the concept that simultaneous
investigations of MEG brain responses to cortical
stimulation in tinnitus is feasible and may reveal short-term
modulation of brain ongoing activity both at the global and
regional scales. Further investigations across a group of
volunteers are certainly required but the methodology
derived here opens wide, innovative perspectives towards a
better understanding of CS on neuropsychological disorders
and possibly, improved targeting and response to CS from a
increasing number of patients.
 A. Davis and A. El Refaie, "Epidemiology of
Tinnitus," in Tinnitus Handbook., R. S. Tyler, Ed. San
Diego: Singular Publishing Group, 2000.
 D. De Ridder, G. De Mulder, V. Walsh, N. Muggleton,
S. Sunaert, and M. A., "Magnetic and electrical
stimulation of the auditory cortex for intractable
tinnitus," J. Neurosurg., vol. 100, pp. 560-564, 2004.
 D. De Ridder, G. De Mulder, E. Verstraeten, K. Van
der Kelen, S. Sunaert, M. Smits, S. Kovacs, J. Verlooy,
P. Van de Heyning, A. R. Moller, and "Primary and
Secondary Auditory Cortex Stimulation for Intractable
Tinnitus," J. Otorhinolaryngol Relat Spec,, vol. 68, pp.
p. 48-54; discussion 54-5., 2006.
 A. J. Fenoy, M. A. Severson, I. O. Volkov, J. F.
Brugge, and M. A. r. Howard, "Hearing suppression
induced by electrical stimulation of human auditory
cortex," Brain Res, vol. 1118, pp. 75-83, 2006.
 G. W. Friedland DR, Runge-Samuelson C, Ulmer JL,
Kopell BH, "Feasibility of Auditory Cortical
STimulation for the Treatment of Tinnitus," Otol &
Neurotol, vol. 28, pp. 1005-12, 2007.
 H. Ashton, K. Reid, R. Marsh, I. Johnson, K. Alter, and
T. Griffiths, "High Frequency Localised "Hot Spots" in
Temporal Lobes of Patients with Intractable Tinnitus:
A Quantitative Electroencephalographic (QEEG)
Study," Neurosci Lett, vol. 426, pp. 23-8, 2007.
 R. Llinás, F. J. Urbano, E. Leznik, R. R. Ramírez, and
H. J. van Marle, "Rhythmic and dysrhythmic
thalamocortical dynamics: GABA systems and the edge
effect," Trends in Neuroscience, vol. 28, pp. 325-33,
 N. Weisz, S. Moratti, M. Meinzer, K. Dohrmann, and
T. Elbert, "Tinnitus perception and distress is related to
abnormal spontaneous brain activity as measured by
magnetoencephalography," PLoS Med. , vol. 2, pp.
 N. Kahlbrock and N. Weisz, "Transient reduction of
tinnitus intensity is marked by concomitant reductions
of delta band power," BMC Biol. , vol. 6, 2008
 K. Dohrmann, N. Weisz, W. Schlee, T. Hartmann, and
T. Elbert, "Neurofeedback for treating tinnitus," Prog
Brain Res. , vol. 166, pp. 473-85, 2007.
 W. Schlee, T. Hartmann, B. Langguth, and N. Weisz,
"Abnormal resting-state cortical coupling in chronic
tinnitus," BMC Neuroscience 2009., vol. 10:11, 2009.
 S. J. Jeanmonod D, Ramirez R, Cancro R, Lanz M,
Morel A, Magnin M, Siegemund M, Kronberg E,
Ribary U, Llinas R., "Neuropsychiatric thalamocortical
dysrhythmia: surgical implications.," Neurosurg Clin N
Am. , vol. 14, pp. 251-65, 2003
 R. R. Llinás, U. Ribary, D. Jeanmonod, E. Kronberg,
and P. P. Mitra, "Thalamocortical dysrhythmia: A
neurological and neuropsychiatric
characterized by MEG," PNAS USA. , vol. 96, pp.
 S. Taulu, M. Kajola, and J. Simola, "Suppression of
interference and artifacts by the Signal Space
Separation Method," Brain Topography. , vol. 16, pp.
 S. Makeig, T. P. Jung, A. J. Bell, D. Ghahremani, and
T. J. Sejnowski, "Blind separation of auditory event-
related brain responses into independent components,"
PNAS U S A. , vol. 94, pp. 10979-10984, 1997.