300Hz subthalamic oscillations in Parkinson's disease

300-Hz subthalamic oscillations in Parkinson'sdisease

G. Foffani,1,2,4* A. Priori,1* M. Egidi,1 P. Rampini,1 F. Tamma,3 E. Caputo,3 K. A. Moxon,4

S. Cerutti2 and S. Barbieri1

1Department of Neurological Sciences, UniversitaÁ di

Milano, IRCCS Ospedale Maggiore di Milano,2Department of Biomedical Engineering, Politecnico di

Milano, 3Department of Clinical Neurology, Ospedale

San Paolo, Milano, Italy and 4School of Biomedical

Engineering, Science and Health Systems,

Drexel University, Philadelphia, PA, USA

Correspondence to: Professor Alberto Priori, Dipartimento

di Scienze Neurologiche, Clinica Neurologica, Padiglione

Ponti, Ospedale Maggiore Policlinico, Via F. Sforza 35,

Milano, 20122, Italy

E-mail: [email protected]

*These two authors equally contributed to this work.

SummaryDespite several studies and models, much remainsunclear about how the human basal ganglia operate.Deep brain stimulation (DBS) of the subthalamicnucleus (STN) is an effective treatment for complicatedParkinson's disease, but how DBS acts also remainsunknown. The clinical bene®t of DBS at frequencies>100 Hz suggests the possible importance of neuralrhythms operating at frequencies higher than the rangenormally considered for basal ganglia processing (<100Hz). The electrodes implanted for DBS also offer theopportunity to record neural activity from the humanbasal ganglia. This study aimed to assess whether oscil-lations at frequencies >100 Hz operate in the humanSTN. While recording local ®eld potentials from theSTN of nine patients with Parkinson's disease throughDBS electrodes, we found a dopamine- and movement-dependent 300-Hz rhythm. At rest, and in the absenceof dopaminergic medication, in most cases (eight out of11 nuclei) the 100±1000 Hz band showed no consistentrhythm. Levodopa administration elicited (or markedly

increased) a 300-Hz rhythm at rest [(mean 6 SD) cen-tral frequency: 319 6 33 Hz; bandwidth: 72 6 21 Hz;power increase (after medication ± before medication)/before medication: 1.30 6 1.25; n = 11, P = 0.00098].The 300-Hz rhythm was also increased by apomor-phine, but not by orphenadrine. The 300-Hz rhythmwas modulated by voluntary movement. Before levo-dopa administration, movement-related power increasein the 300-Hz rhythm was variably present in differentsubjects, whereas after levodopa it became a robustphenomenon [before 0.014 6 0.014 arbitrary units(AU), after 0.178 6 0.339 AU; n = 8, P = 0.0078]. Thedopamine-dependent 300-Hz rhythm probably re¯ects abistable compound nuclear activity and supports high-resolution information processing in the basal gangliacircuit. An absent 300-Hz subthalamic rhythm could bea pathophysiological clue in Parkinson's disease. The300-Hz rhythm also provides the rationale for an excita-toryÐand not only inhibitoryÐinterpretation of DBSmechanism of action in humans.

Keywords: deep brain stimulation; event-related synchronization; high frequency; levodopa; subthalamus

Abbreviations: DBS = deep brain stimulation; ERS = event-related synchronization; LFP = local ®eld potential; SN =

substantia nigra; STN = subthalamic nucleus; UPDRS = uni®ed Parkinson's disease rating scale

IntroductionAbout 80 years ago Kinnear Wilson referred to the basal

ganglia as `the dark basement of the brain' (Wilson, 1925).

Despite enormous progress achieved by experimental studies

and functional models (Wichmann and DeLong, 1996), still

little is known about the signals controlling information

processing and integration in the basal ganglia of living

humans.

Neuronal rhythms, i.e. oscillatory activities, play a key role

in brain information processing (Gray, 1994) at multiple

scales, from subcellular regulation (Goldbeter, 2002) to

network integration (Varela et al., 2001). The importance of

neuronal rhythms in the basal ganglia circuit has been

demonstrated in animal studies in vitro (Bevan and Wilson,

1999; Plenz and Kital, 1999) and in vivo (Bergman et al.,

Brain 126 ã Guarantors of Brain 2003; all rights reserved

DOI: 10.1093/brain/awg229 Advanced Access publication August 21, 2003 Brain (2003), 126, 2153±2163

by guest on May 24, 2014

http://brain.oxfordjournals.org/D

ownloaded from

http://brain.oxfordjournals.org/

1998). In humans, electrophysiological recordings of local

®eld potentials (LFPs) from patients implanted with elec-

trodes for deep brain stimulation (DBS) have so far revealed

the presence of various rhythms in the classic EEG frequency

range, from very low frequencies (<5 Hz) up to the gamma

range (30±80 Hz) (Brown et al., 2001; Marsden et al., 2001;

Pesenti et al., 2003). These studies showed the existence

within the basal ganglia circuit of multiple modes of

operation, observable as network activities oscillating at

different frequencies (Bevan et al., 2002; Levy et al., 2000).

The pathophysiological relevance of these network oscilla-

tions has been suggested by their dependency on dopaminer-

gic medication and on movement execution (Brown et al.,

2001; Cassidy et al., 2002; Foffani et al., 2002; Levy et al.,

2002; Priori et al., 2002). We reported that voluntary

movement modulates beta (~27 and ~17 Hz) oscillations in

the human subthalamic nucleus (STN) and globus pallidus

internus with a pattern of movement-related desynchroniza-

tion (i.e. power decrease) similar to the one observed at the

scalp over the primary motor area (Foffani et al., 2002; Priori

et al., 2002). Similar ®ndings were reported independently by

two other groups. Cassidy et al. (2002) studied the move-

ment-related changes in synchronization in the basal ganglia,

expanding the movement dependency we observed in the beta

range up to gamma frequencies (70±80 Hz); and Levy et al.

(2002) described the dependence of STN oscillations on

movement and dopamine, also showing that network oscil-

lations in LFPs are not necessarily re¯ected by the ®ring rates

of the single neurons.

DBS of the STN is an effective treatment for advanced

Parkinson's disease (Limousin et al., 1998; Dowsey-

Limousin and Pollak, 2001). Stimulation frequencies typ-

ically >100 Hz provide therapeutic ef®cacy (Moro et al.,

2002), similar to the lesion in the STN (Bergman et al.,

1990). Even though this parallel represents the rationale for

the inhibitory/suppressive interpretation of STN stimulation

(Limousin et al., 1995), the mechanism underlying the

therapeutic action of DBS remains unclear (Lozano et al.,

2002; Vitek, 2002); indeed, excitatory effects of DBS have

been proposed recently (Windels et al., 2000; Grill and

Fig. 1 The subthalamic DBS electrode location. Left and right columns: preoperative target selection on the preoperative MRI±CT fusedscans for the right and the left STN, respectively (top and middle: horizontal plane; bottom: coronal plane). The red dot is the STNposition estimated from matching MRI±CT fused images with a digitized stereotactic atlas (see Egidi et al., 2002). Central column:postoperative DBS electrode position (black dots) on the preoperative T2-weighted MRI±CT scans (top and middle: horizonal plane;bottom: coronal plane). Note the position of the STN in preoperative scans (right and left columns): on the horizontal plane (top andmiddle) it is at the level of the anterior border of the red nucleus and lateral to it; on the coronal plane (bottom) it is supero-lateral to theSN. In both planes postoperative scans (central column) show that the DBS electrode is the black dot placed within the area correspondingto the STN in the preoperative scans.

2154 G. Foffani et al.



ownloaded from


McIntyre, 2001; Lozano et al., 2002; Vitek, 2002;

Hashimoto et al., 2003).

It is reasonable to hypothesize the existence of fast (>100

Hz) neuronal rhythms that would be driven by arti®cial

stimulation at their operating frequency. We therefore

recorded LFPs from DBS electrodes placed in the STN of

patients affected by Parkinson's disease searching for

oscillatory activities (i.e. spectral peaks using Fourier analy-

sis) in the frequency range of 100±1000 Hz. Two factors were

tested in order to assess the possible pathophysiological

relevance of these oscillations: (i) their pharmacological

modulation by levodopa (dopamine precursor), apomorphine

(dopamine agonist) or orphenadrine (anticholinergic); and (ii)

their modulation during voluntary movements.

Patients and methodsPatientsNine patients (®ve men, four women) with idiopathic

Parkinson's disease were studied after obtaining their

informed consent and local ethics committee approval. The

average age was 57 years (range 44±69), years of disease

history 16 (9±35), levodopa equivalent therapy before surgery

1455 mg (range 915±2000), Uni®ed Parkinson's Disease

Rating Scale (UPDRS) III (motor part) presurgery `OFF'

therapy 44.2 (27±63.5), `ON' therapy 6.8 (range 1±23.5),

UPDRS III 3 months after surgery OFF therapy ON

stimulation 6.6 (range 3±14; seven patients), UPDRS IV

(complications) before surgery 11.8 (range 6±17), UPDRS IV

3 months after surgery 4.3 (range 0±12; seven patients).

Neuroimaging and stereotactic targeting of theSTNThe STN was localized by direct visualization through an

MRI±CT fusion-based technique (Fig. 1) reported in detail

elsewhere (Egidi et al., 2002). The target position of the

recording electrodes was further assessed with intraoperative

microrecordings and stimulation through the probe micro-

electrodes. The 3389 electrode (Medtronic, Minneapolis, MN,

USA) was ultimately implanted for DBS. The electrode had

four cylindrical contacts (diameter 1.27 mm). Each of them

was 1.5 mm long, and was spaced 0.5 mm away from the

others (i.e. 2 mm centre to centre). Contacts were denomi-

nated 0, 1, 2 and 3 beginning from the more caudal placement.

The objective was to place contact 1 into the target position.

Double-blind assessment revealed marked clinical effects

during intra-operative monopolar macrostimulation through

contact 1. The ®nal location of the DBS electrode within the

STN was veri®ed on the postoperative T2-weighted MRI±CT

fused scan (Fig. 1). In all the patients included in this study for

STN recordings (14 sides from nine patients), contact 1 (and

consistently less contact 2) induced a remarkably greater

clinical effect (scored postoperatively by two independent

blind observers) than contacts 0 and 3. All of these procedures

taken together suggest that contact 1 was within or close to the

STN. We also included two further sides from two of the

above patients where contact 1 and 2 had little or no clinical

effectiveness and postoperative scans showed that the

electrodes were caudally shifted and their ®nal position was

in the substantia nigra (SN).

Postoperative recordingsThe postoperative recording sessions took place 2±3 days

after electrode implantation. The patients were comfortably

seated on an armchair. LFPs were recorded at rest and during

voluntary movements, 8±12 h after withdrawal of dopami-

nergic medication, before and after administration of 100±

200 mg oral fast-acting levodopa (Madopar Dispersibile;

Roche, Monza (MI), Italy) (11 STN and two SN at rest, eight

STN during movement execution) or 6 mg subcutaneous

apomorphine (Apo®n Stylo; Chiesi Farmaceutici, Parma,

Italy) [one STN at rest: AN R (AN is the patient and R

denotes right)] or 40 mg intramuscular orphenadrine (Disipal;

Yamanouchi Pharma, Carugate (MI), Italy) [two STN at rest:

PI R and PI L (PI is the patient, R denotes right and L denotes

left)]. After-medication recordings started 30±40 min after

drug administration, when the patient showed clinical

improvement. The latter was evaluated by an experienced

neurologist: levodopa and apomorphine reduced rigidity,

tremor and bradykinesia; orphenadrine improved mainly

rigidity and tremor. Most patients had some dyskinesias after

levodopa, but none before. Patients were instructed to extend

the second ®nger at the metacarpophalangeal joint from an

initial relaxed semi-¯exed position every 8±10 s as quickly as

possible [mean total movement time (6 SD) before levodopa

1.07 6 0.42 s, after levodopa 0.91 6 0.28 s]. Movements

were self-paced. One extension corresponded to one trial; the

mean (6 SD) number of artefact-free trials was 113 6 39

(range 67±165) before levodopa and 77 6 39 (range 15±133)

after. The task execution was interrupted either by the patient,

if tired, or after a block of 50 trials. Multiple blocks of trials

were usually recorded from each nucleus (range 2±3 before

levodopa, and range 1±6 after levodopa) separated by rest

periods to avoid fatigue. Dyskinesias, when present, did not

interfere with task execution. The EMG signal from the

extensor indicis muscle was recorded through a pair of

surface non-polarizable Ag/AgCl electrodes with a belly-

tendon montage. LFPs were captured from the contralateral

3389 electrode using two pairs of contact: closely spaced (1±

2) and widely spaced (0±3). Signals were pre-ampli®ed,

differentially ampli®ed and ®ltered (EMG, 20±1000 Hz;

STN, 2±1000 Hz) through a Cambridge 1902 (Cambridge

Electronic Design, Cambridge, UK), analogue/digital con-

verted (sampling rate 2500 Hz) through a Cambridge 1401

(Cambridge Electronic Design), on-line analysed on a

personal computer and stored by Signal software, version

1.80 (Cambridge Electronic Design). All further analysis was

conducted off-line with the MATLAB software, version 5.3

300-Hz STN oscillations in Parkinson's disease 2155



ownloaded from


(The Mathworks, Natick, MA, USA) with custom-written

programs described below unless otherwise speci®ed.

Rest spectral analysisThe spectra of the STN LFPs were calculated for ~1 min at rest.

The Welch's averaged, modi®ed periodogram method was

used (Welch, 1967): the signal was divided into ~150 sections

of 1024 samples, with no overlap; each section was detrended

and windowed by a Hanning window; the magnitude squared

of the discrete Fourier transforms of the sections was averaged

to estimate the power spectral density of the signal. Only

spectral frequencies ranging from 100 to 1000 Hz were

considered. To compare data before and after medication, LFPs

were normalized by subtracting the mean and dividing by the

SD of the 600±1000 Hz band-pass ®ltered signals.

Parameter extractionIn order to evaluate the effect of drugs on the STN rest

spectrum, a relative spectrum was calculated for every STN

by dividing the spectrum after medication with the spectrum

before medication. To quantify the spectral properties of the

300-Hz rhythm (central frequency, area, height and width of

the peak), a curve was ®tted to the 100±1000 Hz relative

spectrum using the equation

y = y0 + [ A/(w ´ (p/2)1/2)] ´ exp(±2 ´ (x ± x0)2/w2),

where y0 is the baseline offset, A the total area under the curve

from the baseline, x0 the centre of the peak, and w

approximately 0.849 the width of the peak at half height.

The equation represents a Gaussian curve, like the normal

probability distribution function, where x0 would be the mean

and w/2 the SD. Non-linear least squares ®tting (Dennis,

1977) was performed with the Levenberg±Marquardt

algorithm (Levenberg, 1944; Marquardt, 1963) implemented

in Origin software, version 6.0 (Microcal Software Inc.,

Northampton, MA, USA). The goodness of ®t was evaluated

by two statistics: c2 and R2. c2 is de®ned as the sum of the

squares of the residuals, i.e. of the deviations of the theoretical

curve (the Gaussian) from the experimental points (the

relative spectrum), and it is minimized by the ®tting

procedure. R2 measures how successful the ®t is in explaining

the variation of the data, and it can be interpreted as the square

of the correlation between the experimental curve values and

the theoretical curve values. Therefore, good ®tting is

indicated by a c2 value that approaches zero and an R2

value that approaches one. The values obtained for the two

statistics (average 6 SD: R2 = 0.75 6 0.15, c2 = 0.017 60.010; n = 11; see Table 1) guaranteed that a Gaussian curve

was an accurate analytical description of the relative spectrum

(after/before medication) between 100±1000 Hz. The central

frequency (x0) and bandwidth (w) of the 300-Hz rhythm were

then calculated from the ®tted Gaussian curve. The effect of

levodopa on the 300-Hz rhythm was evaluated by testing the

null hypotheses that the median height (h = y evaluated in x0)

or area (A) of the Gaussian was equal to 0 using the Wilcoxon

signed rank test. This non-parametric test was used because of

the small sample size (n = 11). The central frequency and the

bandwidth were used to de®ne the band-pass ®lter for the

movement-related analysis (see below).

Movement onset±offset detectionThe movement onsets, which triggered the analysis, and

movement offsets, which allowed the estimate of movement

Table 1 Effect of levodopa on the 300-Hz rhythm at rest evaluated by ®tting the normalized 100±1000 Hz spectrum (afterlevodopa/before levodopa) with a Gaussian curve

Patient L or Rnucleus

c2 R2 Area Centre(Hz)

Width(Hz)

Height UPDRSIII OFF

UPDRSIII ON

DeltaUPDRS III

CO R 0.0106 0.8976 147.73 231.77 115.17 1.02 62 23.5 38.5BE R 0.0278 0.9145 186.77 335.20 64.65 2.31 49 1 48BE L 0.0118 0.6512 53.78 303.64 71.30 0.60 49 1 48PA R* 0.0437 0.9393 193.41 305.75 32.70 4.72 27 4.5 22.5PA L* 0.0103 0.8139 69.38 326.32 59.81 0.93 27 4.5 22.5MA R* 0.0128 0.6547 51.67 323.82 61.23 0.67 38.5 4 34.5MA L 0.0120 0.5087 38.06 304.04 79.42 0.38 38.5 4 34.5CR R 0.0195 0.8849 157.01 338.65 86.26 1.45 29 3.5 25.5CM L 0.0111 0.8450 77.92 340.60 56.57 1.10 45.5 4 41.5TO R 0.0145 0.5511 55.43 353.79 89.55 0.49 30.5 2.5 28TO L 0.0163 0.6518 68.42 341.93 81.29 0.67 30.5 2.5 28

Mean 0.0173 0.7557 99.96 318.68 72.54 1.30 38.77 5 33.77SD 0.0102 0.1555 58.74 33.42 21.48 1.25 11.42 6.26 9.36

The c2 and R2 coef®cients describing the goodness of ®t are reported, as well as the areas, centre, width and height of the ®tted Gaussian.Preoperative motor (III) UPDRS scores `OFF' and `ON' therapy, as well as their difference (delta = OFF ± ON) are also reported. Thereis no clear relationship between the UPDRS scores and the 300-Hz descriptors. Each row corresponds to a different STN and it isindicated with the identi®er of the patient followed by left nucleus (L) or right nucleus (R). Means and SDs are reported at the bottom ofthe table. *The nuclei in which the 300-Hz was already observable before levodopa administration.




ownloaded from


duration, were evaluated on the EMG signal as follows: (i) the

EMG signal was squared to obtain power; (ii) the mean of

the resulting signal was subtracted point-by-point; (iii) the

cumulative sum of the resulting signal was calculated and

then ®ltered using a moving average forward±backward (zero

phase) ®lter; (iv) the movement onsets were de®ned as the

local minima (the signal at the previous and at the next

instants is greater than the signal at the current instant) and

the offsets as the local maxima (the signal at the previous and

at the next instants is lower than the signal at the current

instant) in the ®ltered cumulative sum; and (v) signals were

visually inspected, and only epochs free from artefacts were

used for the analysis.

Movement-related analysisLFPs from the STN were ®rst normalized to their high-

frequency content (as described in `Rest spectral analysis')

and then ®ltered with a zero-phase ®nite impulse response

(forward±backward) band-pass ®lter. The band was de®ned

as the central frequency plus and minus half the bandwidth

calculated with the Gaussian ®tting of the relative spectra, as

described above. The ®ltered signals were then squared to

obtain power and averaged over trials. The average signals

were smoothed by subsampling to a ®nal 10-Hz sampling rate

after anti-aliasing zero-phase low-pass ®ltering. Changes in

the resulting signals were then detected by change-point

analysis (Taylor, 2000; Cassidy et al., 2002) using commer-

cial software (Change-Point Analyser 2.0 shareware program;

Taylor Enterprises, IL, USA). Only changes with a probability

of change >99% and 95% con®dence interval <800 ms were

considered, performing 10 000 bootstraps without replace-

ment. The average power before the onset change was

considered as `baseline power', and the average power after

the onset change less the baseline power was considered as

event-related synchronization (ERS) (Pfurtscheller and Lopes

da Silva, 1999) or `ERS power'. The signi®cance of levodopa

effect on baseline power, ERS power and on the ratio between

ERS and baseline powers was separately tested using the

Wilcoxon matched-pairs signed rank test. This non-para-

metric test was used because of the small sample size (n = 8).

Throughout the text, values are means 6 1 SD. A P-value

<0.05 was considered to indicate statistical signi®cance.

ResultsSearching for neural rhythms at frequencies >100 Hz, we

recorded LFPs from DBS electrodes in the human STN at rest

and during voluntary movement.

Rest spectral analysisAt rest, in the absence of dopaminergic medication, no

consistent rhythmic activity (i.e. spectral peak) was observed

between 100 and 1000 Hz (Fig. 2C and F). As well as

inducing a prompt clinical motor improvement comparable to

that before surgery (see UPDRS scores in the section

`Patients' and in Table 1), levodopa administration elicited

(in eight out of 11 nuclei, from seven patients), or markedly

increased (in three out of 11 nuclei) a subthalamic 300-Hz

rhythm at rest (Fig. 2C, G and H). Hence, in three out of 11

nuclei the 300-Hz rhythm was already present before

levodopa administration, and in the absence of dyskinesias.

The 300-Hz rhythm had a central frequency of 319 6 33 Hz

and a bandwidth of 72 6 21 Hz. Overall, the boosting effect

of levodopa on the 300-Hz rhythm in the human STN at rest

was highly signi®cant. Two parameters were used to support

this assertion: (i) the ®tted Gaussian height, which corres-

ponds to maximal percent increase spectral peak after

medication compared with before medication; and (ii) the

®tted Gaussian area, which corresponds to the integral of the

relative spectrum over the entire 100±1000 Hz spectrum, i.e.

a sort of total power increase of the 300-Hz rhythm in

arbitrary units. The average height was 1.30 6 1.25, where 0

represents the value before medication and 1 corresponds to

100% increase; the increase was signi®cantly different from

zero (Wilcoxon test; n = 11, P = 0.00098). The average area

was 99.96 6 58.73 arbitrary units (AU) and signi®cantly

differed from zero (Wilcoxon test; n = 11, P = 0.00098). The

area and the height in one nucleus were markedly higher than

the others (PA R; see Table 1). The average relative spectrum

[(after levodopa ± before levodopa)/before levodopa] is

shown in Fig. 2H. The results are detailed in Table 1. The

magnitude of the 300-Hz rhythm was compared with

frequencies <100 Hz, namely the 15±40 and 70±80 Hz

frequency bands. The peak power of the 300-Hz rhythm was

smaller than the 15±40 Hz peak power (average 300 Hz/15±

40 Hz power ratio = 0.23 6 0.19) and greater than the

70±80 Hz peak power (average 300 Hz/70±80 Hz power ratio

= 1.85 6 0.66).

To test the hypothesis that the 300-Hz rhythm was

dopamine dependent, and not exclusively levodopa depend-

ent, we studied the effect of subcutaneous apomorphine

administration (one nucleus). The drug reduced rigidity,

tremor and bradykinesia. Before apomorphine the 300-Hz

rhythm was barely detectable, after drug injection it markedly

increased: the ®tted Gaussian height was 0.53 (i.e. 53%

increase), the area was equal to 47 arbitrary units. The central

frequency of the peak was 298 Hz with a bandwidth of 71 Hz

(Fig. 2D).

To con®rm further the speci®c dopaminergic dependency

of the 300-Hz rhythm at rest, we studied changes in the

oscillatory activity induced by intramuscular administration

of orphenadrine, an anticholinergic drug (two nuclei).

Although orphenadrine improved rigidity and tremor, it did

not enhance the 300-Hz rhythm (Fig. 2E). Hence, despite the

possible variability of the responses to apomorphine and

orphenadrine and the small number of nuclei tested, the

®ndings with these two drugs support a dopaminergic

dependence of the 300-Hz rhythm.

Interestingly, the 300-Hz rhythm was always maximal in

the more closely spaced pair of contacts (1±2), but absent or




ownloaded from


Fig. 2 Pharmacological modulation of the 300-Hz rhythm at rest. (A, B) One second of subthalamic LFP (A) before and (B) afterlevodopa administration. (C) Effect of levodopa on the 100±1000 Hz spectrum in one STN (BE R). (D) Effect of apomorphine (APO) onthe 100±1000 Hz spectrum in one STN (AN R). (E) Effect of orphenadrine (ORP) on the 100±1000 Hz spectrum in one STN (PI L). (C±E) Grey and black continuous lines represent data before and after medication; grey and black dotted lines are the correspondingcon®dence intervals (c.i.). (F, G) 100±1000 Hz spectrum of all 11 nuclei superimposed (F) before and (G) after levodopa administration.The y-axis is in logarithmic scale. (H) Levodopa-dependent increment of the 100±1000 Hz spectrum {[(post-levodopa) ± (pre-levodopa)]/[pre-levodopa]} expressed as a percentage and averaged over all 11 nuclei. (I) 100±1000 Hz spectrum of one STN (CM L) after levodopaadministration evaluated in the closely spaced contacts (1±2) and in the widely spaced contacts (0±3). (J) 100±1000 Hz spectrum of oneSN (CM R) before and after levodopa administration. a.u. = arbitrary units.




ownloaded from


reduced in the wider pair (0±3) (Fig. 2I). On average, after

levodopa administration the ratio between the 300-Hz power

in the closely spaced pair and the 300-Hz power in the wider

spaced pair (1±2/0±3) was 1.96 6 0.62 (Wilcoxon test; n = 7,

P = 0.0156). In addition, when the closely spaced pairs of

contacts (1±2) were supposedly in the SN (n = 2: CR L and

CM R), and hence outside the STN, we failed to detect the

300-Hz rhythm (Fig. 2J).

Movement-related power modulation of the 300-Hz oscillationsTo test whether voluntary movement modulated the 300-Hz

rhythm, we asked our patients to execute self-paced

contralateral second ®nger extensions during LFP recordings

from the STN (eight nuclei from six patients). When EMG

activity began, power in the 300-Hz rhythm increased; when

EMG activity ended, power returned to baseline values

(Fig. 3). The results are detailed in Table 2. The responses of

PA R are much higher than those of the other nuclei.

Before levodopa administration, the movement-related

power increase in the 300-Hz rhythm was variably present

in different subjects (signi®cant changes detected in four out

of eight nuclei); after levodopa it became a robust phenom-

enon (signi®cant changes in eight out of eight nuclei). As

shown in Table 2, levodopa administration signi®cantly

increased the baseline power of the 300-Hz rhythm (Wilcoxon

test; P = 0.0078), as well as the ERS power (Wilcoxon test;

P = 0.0078) and the ERS/baseline ratio (Wilcoxon test;

P = 0.0391). When the movement-related power increase of

the 300-Hz rhythm was already signi®cant before levodopa

administration, the phenomenon was enhanced by the

drug.

DiscussionThe spectrum of oscillatory activities in the human basal

ganglia is hence wider than known to date, spreading also to

unusually high frequencies, around 300 Hz. The 300-Hz

subthalamic oscillations in Parkinson's disease appear after

dopaminergic medication and their power increases during

voluntary movement.

300-Hz oscillations at restAt rest, the 300-Hz rhythm appears only when the patient is in

`ON' condition after levodopa or apomorphine, but not after

the anticholinergic drug. Hence dopaminergic medication in

parkinsonian patients speci®cally shifts the system to a new

dynamic equilibrium in which the human STN oscillates at a

characteristic frequency around 300 Hz. Three pieces of

evidence: (i) the fact that 15±40 Hz activity is decreased (and

not increased) by levodopa (Brown et al., 2001); (ii) the

greater peak power of the 300-Hz rhythm compared with 70±

80 Hz; and (iii) the absence of spectral peaks in harmonic

relationship around 100 or 150 Hz, argue against, though do

not exclude, the possibility of the 300-Hz rhythm being a

harmonic of lower frequencies.

Even though, for ethical and methodological reasons, we

did not systematically map the spatial distribution of the

300-Hz rhythm, the fact that 300-Hz oscillations are reduced

in the wider spaced pair of DBS contacts (0±3) compared with

the closely spaced ones (1±2) suggests that the 300-Hz

rhythm is probably located within or close to the STN. The

localization of the ®nding is also con®rmed by the fact that

300-Hz oscillations are not observable `out of target', outside

the STN (two SN). Nevertheless, the presence of 300-Hz

oscillations in other basal ganglia nuclei is possible.

Although in this paper 300 Hz is used as the reference

frequency, the range of the central frequencies of the 300-Hz

rhythm actually encompasses more than 100 Hz on the

frequency axis (from 232 to 354 Hz). This variability is not as

large as it might seem, considering that the bandwidth of the

300-Hz rhythm can be >100 Hz. The broad bandwidth of

high-frequency oscillations is in accordance with the uncer-

tainty principle expressed by the Heisenberg inequality

(Df ´ Dt > 1/4p, where Df is frequency resolution and Dt

time resolution), which states that resolution in time and

frequency can not be arbitrarily small, because their product

is lower bounded (Gabor, 1946; Heisenberg, 1949). In other

words, higher frequencies guarantee better time resolution

(i.e. faster processing), but such `speed' is paid for in term of

frequency resolution, which necessarily decreases. This

simply means that higher is the frequency of neural oscilla-

tions, wider is their bandwidth. An example of this can be

seen in the classical EEG frequency range in the STN (Levy

et al., 2002) or, more traditionally, at the scalp: the alpha or

slower rhythms are well localized in frequency, whereas the

relatively fast beta activities are more spread out on the

frequency axis.

The cellular mechanisms responsible for the 300-Hz

oscillations are unclear. Until now, no stable neural activity

oscillating as a network at such high frequencies has, to our

knowledge, been reported in any brain structure involved in

human motor control. Correlated neuronal rhythms at

frequencies >100 Hz have been observed in vivo in the

hippocampus (BuzsaÂki et al., 1992; Hirai et al., 1999). The

cellular origin of these high-frequency oscillations has been

related to electrical coupling (Draguhn et al., 1998) and

axo-axonal connections (Schmitz et al., 2001).

A synchronized, phase-shifted recruitment of STN cells,

reverberating activity of STN units due to intrinsic mem-

brane properties, subthreshold ¯uctuations determined by

coordinated pre-synaptic input and high-frequency dendro-

somatic potential uncoupling represent other possible and

not mutually exclusive mechanisms for the origin of 300-Hz

oscillations. Hence, the presence of a 300-Hz activity does

not necessarily imply either a single presynaptic 300-Hz

drive or a postsynaptic unitary activity at such high

frequencies.




ownloaded from


Movement-related modulation of the 300-HzoscillationsThe relatively slow alpha and beta EEG rhythms are

commonly interpreted as ìdling' rhythms and, accordingly,

they are lost or reduced during active processing (e.g. eye

opening, movement execution, etc.); movement-related

power decreases are often referred as event-related desyn-

chronization (Pfurtscheller and Lopes da Silva, 1999). On the

other hand, gamma oscillations (40±80 Hz) typically increase

their power in correspondence of movements or other

cognitive actions; movement-related power increases are

often referred to as ERS (Pfurtscheller and Lopes da Silva,

1999); for their characteristic ERS, gamma rhythms have

been proposed to directly re¯ect active processing

(Pfurtscheller and Lopes da Silva, 1999) and they have

been functionally related to the need for `binding' informa-

tion from different brain structures (Von der Malsburg, 1995)

to produce fast integration for higher-level processing (Singer

and Gray, 1995). Indeed, the 300-Hz rhythm we observed in

the human STN presents pharmacological and movement-

related dynamics that are very similar to those previously

reported in the same STN, coherently with the globus pallidus

internus, in the high-gamma range at 70±80 Hz (Brown et al.,

2001; Cassidy et al., 2002): they closely follow the EMG

burst, with the power rapidly increasing and decreasing back

to baseline values.

During movement execution, levodopa signi®cantly in-

creased the 300-Hz baseline power, the ERS power and the

ERS/baseline ratio. The fact that levodopa effect is signi®cant

on the 300-Hz baseline power is obvious: the baseline power

corresponds to a `dynamic rest' between two movements,

therefore the levodopa dependent increase con®rms the

results obtained at rest. On the other hand, the fact that

levodopa effect is signi®cant on the 300-Hz ERS power

means that greater the dopamine -dependent 300-Hz activity

at rest, the greater its movement-related modulation. Finally,

the fact that effect of levodopa on the ERS/baseline ratio is

signi®cant is subtler, and it indicates that the relationship

qualitatively stated at the previous sentence is non-linear.

Intuitively the concept is that the àmount' of oscillation at

rest not only allows a certain àmount' of movement-related

modulation, but also increases the èf®ciency' of the modu-

lation itself.

Pathophysiological implicationsOne might wonder whether high-frequency oscillations in the

STN, either at 70±80 Hz (Brown et al., 2001; Cassidy et al.,

Fig. 3 Movement-related modulation of the 300-Hz rhythm. (A±H) Peri-event (0 s = EMG onset) power averages of the 300-Hz band-pass®ltered STN signals from eight single STN, before (grey line) and after (black line) levodopa administration.




ownloaded from


2002) or at 300 Hz, are not causally responsible for the

levodopa induced improvement but only its epiphenomenon.

Also, these oscillations might be, at least in part, related to

drug-induced involuntary movements. However, Cassidy

et al. (2002) reported that more than half of the patients

had no clinical evidences of dyskinesia or tremor during the

recordings and changes at 70±80 Hz started after a warning

stimulus, during the reaction time and before movement

onset. In addition, 300-Hz activity can be observed even

before levodopa administration (30% of the nuclei at rest and

50% movement related), in absence of any involuntary

movement. Finally, these high-frequency subthalamic oscil-

lations are modulated by both dopaminergic stimulation and

voluntary movement. Therefore, at least in Parkinson's

disease, the 300-Hz rhythm appears to be a determinant of

voluntary motor behaviour.

The ®nding of the 300-Hz rhythm in the human STN has

several possible implications. First, because signals at higher

frequencies can process larger amounts of information than

rhythms in the beta or even gamma range, faster rhythms

accord well with the operational complexity of the basal

ganglia circuit. Using an analogy with the classic communi-

cation theory, the STN processes phasic movement informa-

tion in amplitude modulation mode, on the 300-Hz carrier

frequency. The maximum theoretical bandwidth of the

transmission (i.e. the amount of information per time unit)

is directly proportional to the carrier frequency (Haykin,

1994). Although the information currency of the nervous

system is the action potential rather than the LFP, 300-Hz

oscillations may provide a coordinated `clock' that paces the

neurons excitability at a resolution of ~3 ms, acting somehow

similarly to the clock of a microprocessor in a personal

computer. Therefore, the 300-Hz subthalamic rhythm might

not be directly involved in individual movements or actions,

but it could be a non-speci®c regulator of neural synchrony in

order to guarantee speci®c modulation of individual move-

ments, that probably are more directly determined by lower

frequencies. The 300-Hz power increase during movement

execution would further enhance the clock precision when

fast and robust processing is needed. Hence, the loss of

300-Hz oscillations can contribute to movement abnormal-

ities in Parkinson's disease.

Secondly, from a pathophysiological point of view, the

appearance of the 300-Hz rhythm after dopaminergic medi-

cation in subjects at rest may re¯ect the shift of a bistable

neural subsystem from quiescence to an operational state that

allows the optimal information processing required for

normal motor control. Neural bistability has recently been

suggested as a key element for robust neural integration

(Koulakov et al., 2002). At a cellular level, bistable behaviour

has been experimentally observed in neurons from many

brain structures, including the striatum (Nicola et al., 2000),

depending on speci®c neurotransmitters, including dopamine

(Nicola et al., 2000; Lavin and Grace, 2001). At the network

level, the sudden appearance of synchronized neural rhythms

in correspondence with pathological or physiological events,

as, for example, in epilepsy (Manuca et al., 1998) or simply in

attention (Steinmetz et al., 2000; Salinas and Sejnowski,

2001), may be interpreted as a state transition of a complex

neural system. In Parkinson's disease, the loss of nigrostriatal

dopaminergic projections might determine an opposite tran-

sition to a desynchronized state, electrophysiologically

re¯ected by absence of the 300-Hz subthalamic rhythm and,

eventually, reversed by dopaminergic medication.

Finally, the 300-Hz rhythm could also help to explain how

STN DBS induces its bene®cial effects in Parkinson's

Table 2 Movement-related modulation of the 300-Hz rhythm before and after levodopa

Patient Side Before L-dopa After L-dopa(nucleus)

Baselinepower (AU)

Baseline+ ERSpower (AU)

ERS power(AU)

ERS/baseline(%)

Baselinepower (AU)

Baseline+ ERSpower (AU)

ERS power(AU)

ERS/baseline(%)

CO R 0.206 0.251 0.045 21.834 0.348 0.418 0.070 20.230BE R 0.126 0.141 0.014 11.401 0.264 0.336 0.072 27.128PA R 0.081 0.081 0.000 0.000 0.454 1.469 1.015 223.404PA L 0.201 0.201 0.000 0.000 0.354 0.464 0.110 31.148MA R 0.156 0.166 0.010 6.090 0.281 0.342 0.061 21.831MA L 0.158 0.177 0.019 12.167 0.236 0.259 0.023 9.754CR R 0.191 0.199 0.008 4.239 0.351 0.379 0.028 8.039CM L 0.119 0.133 0.014 11.869 0.215 0.256 0.042 19.534

Mean 0.155 0.169 0.014 8.450 0.313 0.490 0.178 45.133SD 0.044 0.052 0.014 7.363 0.079 0.402 0.339 72.453

Baseline power (average power before the movement-related change detected by change-point analysis), baseline + ERS power (averagepower after the change), ERS power (the difference between the powers before and after the change) and ERS/baseline ratio for the eightSTN observed before left (L) and after right (R) levodopa administration. Bold numbers indicate the nuclei that had no signi®cantmovement-dependent change before levodopa, considering probability of change >99% and 95% con®dence interval <800 ms assigni®cance criteria. AU = arbitrary units.; L-dopa = levodopa.




ownloaded from


disease. The clinical ef®cacy of DBS at frequencies

>100 Hz could re¯ect its role as arti®cial (subharmonic)

drive for the physiological neural oscillations at 300 Hz

required for normal basal ganglia function. If we interpret the

basal ganglia as a complex dynamic system, dopamine may

be viewed as an external input that sets some internal

variables at the working oscillatory equilibrium required for a

correct output. When the dopaminergic system is defective as

it is in Parkinson's disease, DBS would directly drive those

internal variables into or closer to their physiological

oscillatory equilibrium. This excitatory interpretation of

DBS at a network level is in agreement with recent ®ndings

on the MPTP primate model of Parkinson's disease

(Hashimoto et al., 2003), and it does not contrast with

inhibitory effects at the cellular level that have been

originally suggested for the similar effect of STN stimulation

and its lesion. Indeed, DBS and lesions can elicit a similar

(but probably not the same) clinical improvement through

different mechanisms of actions. As shown by intracerebral

microdialysis studies in rats (Windels et al., 2000, 2002),

excitatory and inhibitory effects might coexist and both

contribute to the therapeutic effect of DBS. Hence, DBS

could act in a far more complex way than was once

thought.

AcknowledgementsThe authors wish to thank Dr G. Ardolino, Professor G.

Baselli, Ing. A. Bianchi, Dr V. Chiesa, Dr M. Locatelli and Dr

A. Pesenti, for their helpful cooperation, and S. Croci, S.

Garlaschi, G. Gherardi, A. Marsilio, B. Meda, M. Pastori and

M. Pellegrini for their technical assistance. The study was

partially supported by Associazione Amici del Centro Dino

Ferrari for Neurodegenerative Disorders, by a Health

Italian Ministry Collaborative grant with IRCCS Istituto

Neurologico Mediterraneo Neuromed, Pozzilli, Italy and by

MURST.

References

Bergman H, Wichmann T, DeLong MR. Reversal of experimental

parkinsonism by lesions of the subthalamic nucleus. Science 1990;

249: 1436±8.

Bergman H, Feingold A, Nini A, Raz A, Slovin H, Abeles M, et al.

Physiological aspects of information processing in the basal ganglia

of normal and parkinsonian primates. [Review]. Trends Neurosci

1998; 21: 32±8.

Bevan MD, Wilson CJ. Mechanisms underlying spontaneous

oscillation and rhythmic ®ring in rat subthalamic neurons. J

Neurosci 1999; 19: 7617±28.

Bevan MD, Magill PJ, Terman D, Bolam JP, Wilson CJ. Move to

the rhythm: oscillations in the subthalamic nucleus±external globus

pallidus network. [Review]. Trends Neurosci 2002; 25: 525±31.

Brown P, Oliviero A, Mazzone P, Insola A, Tonali P, Di Lazzaro V.

Dopamine dependency of oscillations between subthalamic nucleus

and pallidum in Parkinson's disease. J Neurosci 2001; 21: 1033±8.

BuzsaÂki G, HorvaÂth Z, Urioste R, Hetke J, Wise K. High-frequency

network oscillation in the hippocampus. Science 1992; 256:

1025±7.

Cassidy M, Mazzone P, Oliviero A, Insola A, Tonali P, Di Lazzaro

V, et al. Movement-related changes in synchronization in the

human basal ganglia. Brain 2002; 125: 1235±46.

Dennis JE Jr. Nonlinear least squares and equations. In: Jacob D,

editor. The state of the art of numerical analysis. Orlando:

Academic Press; 1977. p. 269±312.

Dowsey-Limousin P, Pollak P. Deep brain stimulation in the

treatment of Parkinson's disease: a review and update. [Review].

Clin Neurosci Res 2001; 1: 521±6.

Draguhn A, Traub RD, Schmitz D, Jefferys JGR. Electrical

coupling underlies high-frequency oscillations in the hippocampus

in vitro. Nature 1998; 394: 189±92.

Egidi M, Rampini P, Locatelli M, Farabola M, Priori A, Pesenti

A, et al. Visualisation of the subthalamic nucleus: a multiple

sequential image fusion (MuSIF) technique for direct stereotaxic

localisation and postoperative control. Neurol Sci 2002; 23

Suppl 2: S71±2.

Foffani G, Priori A, Rohr M, Pesenti A, Locatelli M, Caputo E, et al.

Event-related desynchronization (ERD) in the human subthalamus

and internal globus pallidus. J Physiol 2002; 539.P: 39P.

Gabor D. Theory of communication. J Inst Elect Engrs (London)

1946; 93: 429±57.

Goldbeter A. Computational approaches to cellular rhythms.

[Review]. Nature 2002; 420: 238±45.

Gray CM. Synchronous oscillations in neuronal systems:

mechanisms and functions. [Review]. J Comput Neurosci 1994; 1:

11±38.

Grill WM, McIntyre CC. Extracellular excitation of central

neurons: implications for the mechanisms of deep brain

stimulation. Thal Rel Sys 2001; 1: 269±77.

Haykin S. Communication systems. 3rd ed. New York: John Wiley;

1994.

Hashimoto T, Elder CM, Okun MS, Patrick SK, Vitek JL.

Stimulation of the subthalamic nucleus changes the ®ring pattern

of pallidal neurons. J Neurosci 2003; 23: 1916±23.

Heisenberg, W. The physical principles of quantum theory. New

York: Dover; 1949.

Hirai N, Uchida S, Maehara T, Okubo Y, Shimizu H. Enhanced

gamma (30±150 Hz) frequency in the human medial temporal lobe.

Neuroscience 1999; 90: 1149±55.

Koulakov AA, Raghavachari S, Kepecs A, Lisman JE. Model for a

robust neural integrator. Nat Neurosci 2002; 5: 775±82.

Lavin A, Grace AA. Stimulation of D1-type dopamine receptors

enhances excitability in prefrontal cortical pyramidal neurons in a

state-dependent manner. Neuroscience 2001; 104: 335±46.

Levenberg K. A method for the solution of certain problems in least

squares. Quart Appl Math 1944; 2: 164±8.




ownloaded from


Levy R, Hutchison WD, Lozano AM, Dostrovsky JO. High-

frequency synchronization of neuronal activity in the subthalamic

nucleus of parkinsonian patients with limb tremor. J Neurosci 2000;

20: 7766±75.

Levy R, Ashby P, Hutchison WD, Lang AE, Lozano AM,

Dostrovsky JO. Dependence of subthalamic nucleus oscillations

on movement and dopamine in Parkinson's disease. Brain 2002;

125: 1196±209.

Limousin P, Pollak P, Benazzouz A, Hoffmann D, Le Bas JF,

Broussolle E, et al. Effect of parkinsonian signs and symptoms

of bilateral subthalamic nucleus stimulation. Lancet 1995; 345:

91±5.

Limousin P, Krack P, Pollak P, Benazzouz A, Ardouin C, Hoffmann

D, et al. Electrical stimulation of the subthalamic nucleus in

advanced Parkinson's disease. New Engl J Med 1998; 339: 1105±11.

Lozano AM, Dostrovsky J, Chen R, Ashby P. Deep brain

stimulation for Parkinson's disease: disrupting the disruption.

[Review]. Lancet Neurol 2002; 1: 225±31.

Manuca R, Casdagli MC, Savit RS. Nonstationarity in epileptic

EEG and implications for neural dynamics. Math Biosci 1998; 147:

1±22.

Marquardt D. An algorithm for least-squares estimation of

nonlinear parameters. SIAM J Appl Math 1963; 11: 431±41.

Marsden JF, Limousin-Dowsey P, Ashby P, Pollak P, Brown P.

Subthalamic nucleus, sensorimotor cortex and muscle

interrelationships in Parkinson's disease. Brain 2001; 124:

378±88.

Moro E, Esselink RJA, Xie J, Hommel M, Benabid AL, Pollak P.

The impact on Parkinson's disease of electrical parameter settings

in STN stimulation. Neurology 2002; 59: 706±13.

Nicola SM, Surmeier DJ, Malenka RC. Dopaminergic modulation

of neuronal excitability in the striatum and nucleus accumbens.

[Review] Annu Rev Neurosci 2000; 23: 185±215.

Pesenti A, Priori A, Locatelli M, Egidi M, Rampini P, Tamma F,

et al. Subthalamic somatosensory evoked potentials in Parkinson's

disease. Mov Disord 2003; DOI: 10.1002/MDS.10519.

Pfurtscheller G, Lopes da Silva FH. Event-related EEG/EMG

synchronization and desynchronization: basic principles. [Review].

Clin Neurophysiol 1999; 110: 1842±57.

Plenz D, Kital ST. A basal ganglia pacemaker formed by the

subthalamic nucleus and external globus pallidus. Nature 1999;

400: 677±82.

Priori A, Foffani G, Pesenti A, Bianchi A, Chiesa V, Baselli G, et al.

Movement-related modulation of neural activity in human basal

ganglia and its L-DOPA dependency: recordings from deep brain

stimulation electrodes in patients with Parkinson's disease. Neurol

Sci 2002; 23 Suppl 2: S101±2.

Salinas E, Sejnowski TJ. Correlated neuronal activity and the ¯ow of

neural information. [Review]. Nat Rev Neurosci 2001; 2:

539±50.

Schmitz D, Schuchmann S, Fisahn A, Draguhn A, Buhl EH,

Petrasch-Parwez E, et al. Axo-axonal coupling: a novel

mechanism for ultrafast neuronal communication. Neuron 2001;

31: 831±40.

Singer W, Gray CM. Visual feature integration and the temporal

correlation hypothesis. [Review]. Annu Rev Neurosci 1995; 18:

555±86.

Steinmetz PN, Roy A, Fitzgerald PJ, Hsiao SS, Johnson KO, Niebur

E. Attention modulates synchronized neuronal ®ring in primate

somatosensory cortex. Nature 2000; 404: 187±90.

Taylor WA. Change-point analysis: a powerful new tool for

detecting changes. 2000. Available from: http: //www.variation.

com/cpa/tech/changepoint.html

Varela F, Lachaux JP, Rodriguez E, Martinerie J. The brainweb:

phase synchronization and large-scale integration. [Review]. Nat

Rev Neurosci 2001; 2: 229±39.

Vitek JL. Mechanisms of deep brain stimulation: excitation or

inhibition. [Review] Mov Disord 2002; 17 Suppl 3: S69±72.

VonderMalsburg C. Binding in models of perception and brain

function. [Review]. Curr Opin Neurobiol 1995; 5: 520±6.

Welch PD. The use of fast Fourier transform for the estimation of

power spectra: a method based on time averaging over short,

modi®ed periodograms. IEEE Trans. Audio Electroacoustics 1967;

AU-15: 70±3.

Wichmann T, DeLong MR. Functional and pathophysiological

models of the basal ganglia. [Review]. Curr Opin Neurobiol 1996;

6: 751±758.

Wilson K. Disorders of motility and of muscle tone, with special

reference to the corpus striatum. Lancet 1925; 2: 1±10, 53±62, 169±

78, 215±9, 268±76.

Windels F, Bruet N, Poupard A, Urbain N, Chouvet G, Feuerstein

C, et al. Effects of high frequency stimulation of subthalamic

nucleus on extracellular glutamate and GABA in substantia nigra

and globus pallidus in the normal rat. Eur J Neurosci 2000; 12:

4141±6.

Windels F, Bruet N, Poupard A, Feuerstein C, Bertrand A, Savasta

M. In¯uence of frequency on extracellular glutamate and GABA in

substantia nigra and globus pallidus during electrical stimulation of

subthalamic nucleus in rats [abstract]. Mov Disord 2002; 17 Suppl

5: S200.

Received January 8, 2003. Revised March 16, 2003.

Second revision April 16, 2003. Accepted April 21, 2003




ownloaded from


300Hz subthalamic oscillations in Parkinson's disease

Documents

Transcript of 300Hz subthalamic oscillations in Parkinson's disease