Pallidal Recordings in Chronically Implanted Dystonic Patients: Mitigation of Tremor-Related Artifacts [original]

Citation: Del Vecchio Del Vecchio, J.;

Hanafi, I.; Pozzi, N.G.; Capetian, P.;

Isaias, I.U.; Haufe, S.; Palmisano, C.

Pallidal Recordings in Chronically

Implanted Dystonic Patients:

Mitigation of Tremor-Related

Artifacts. Bioengineering 2023,10, 476.

https://doi.org/10.3390/

bioengineering10040476

Academic Editors: Christina

Zong-Hao Ma and Hong Fu

Received: 13 March 2023

Revised: 31 March 2023

Accepted: 1 April 2023

Published: 15 April 2023

Licensee MDPI, Basel, Switzerland.

This article is an open access article

distributed under the terms and

conditions of the Creative Commons

Attribution (CC BY) license (https://

creativecommons.org/licenses/by/

4.0/).

bioengineering

Article

Pallidal Recordings in Chronically Implanted Dystonic Patients:

Mitigation of Tremor-Related Artifacts

Jasmin Del Vecchio Del Vecchio 1,*, Ibrahem Hanafi 1, NicolóGabriele Pozzi 1, Philipp Capetian 1,

Ioannis U. Isaias 1,2 , Stefan Haufe 3,4,5,† and Chiara Palmisano 1,†

1Department of Neurology, University Hospital of Würzburg and Julius-Maximilian-University Würzburg,

97080 Würzburg, Germany; hanafi_i@ukw.de (I.H.); pozzi_n2@ukw.de (N.G.P.); capetian_p@ukw.de (P.C.);

isaias_i@ukw.de (I.U.I.); palmisano_c@ukw.de (C.P.)

2Centro Parkinson e Parkinsonismi, ASST G. Pini-CTO, 20122 Milano, Italy

3Uncertainty, Inverse Modeling and Machine Learning Group, Technische Universität Berlin,

10623 Berlin, Germany; [email protected]

4Physikalisch-Technische Bundesanstalt Braunschweig und Berlin, 10587 Berlin, Germany

5Berlin Center for Advanced Neuroimaging, Charité—Universitätsmedizin Berlin, 10117 Berlin, Germany

*Correspondence: delvecchio_j@ukw.de

† These authors contributed equally to this work.

Abstract:

Low-frequency oscillatory patterns of pallidal local field potentials (LFPs) have been

proposed as a physiomarker for dystonia and hold the promise for personalized adaptive deep

brain stimulation. Head tremor, a low-frequency involuntary rhythmic movement typical of cervical

dystonia, may cause movement artifacts in LFP signals, compromising the reliability of low-frequency

oscillations as biomarkers for adaptive neurostimulation. We investigated chronic pallidal LFPs with

the Percept

PC (Medtronic PLC) device in eight subjects with dystonia (five with head tremors).

We applied a multiple regression approach to pallidal LFPs in patients with head tremors using

kinematic information measured with an inertial measurement unit (IMU) and an electromyographic

signal (EMG). With IMU regression, we found tremor contamination in all subjects, whereas EMG

regression identified it in only three out of five. IMU regression was also superior to EMG regression

in removing tremor-related artifacts and resulted in a significant power reduction, especially in the

theta-alpha band. Pallido-muscular coherence was affected by a head tremor and disappeared after

IMU regression. Our results show that the Percept PC can record low-frequency oscillations but also

reveal spectral contamination due to movement artifacts. IMU regression can identify such artifact

contamination and be a suitable tool for its removal.

Keywords: dystonia; tremor; local field potentials; globus pallidus; deep brain stimulation

1. Introduction

Dystonia is a movement disorder characterized by patterned, directional, and often

sustained involuntary muscle contractions that produce abnormal postures or repetitive

movements [1]. Tremor can be a basic feature of a dystonic contraction and is reported in up

to 87% of patients [

]. One of the most common dystonic tremors is head tremor in patients

with cervical dystonia (CD) [3], in which it causes a severe reduction in quality of life.

The pathophysiology of CD and dystonic tremor (DT) is not entirely known. It is still

unclear whether tremor in dystonia has its own pathophysiology unrelated to dystonia

or shares some mechanisms. It is also still debated whether DT, which is the presence of

tremor in a body part affected by dystonia, and tremor associated with dystonia (TAWD),

defined as tremor in a body part not affected by dystonia [

], are distinct entities or simi-

lar syndromes [

]. Previous studies revealed increased intermuscular coherence in DT

because of a loss of reciprocal inhibition [

–

], indicating that DT shares some dystonia

pathophysiology [

–

]. Functional imaging studies further showed the involvement of

Bioengineering 2023,10, 476. https://doi.org/10.3390/bioengineering10040476 https://www.mdpi.com/journal/bioengineering

Bioengineering 2023,10, 476 2 of 15

the basal ganglia, thalamus, midbrain, and sensory-motor cortex in DT, similar to dys-

tonia without tremor [

]. Recently, there has been increasing evidence of cerebellar

involvement and interactions between the cerebellum and basal ganglia in DT and dys-

tonia, and most of the literature converges in supporting the involvement of both basal

ganglia-thalamo-cortical and cerebello-thalamo-cortical pathways in dystonia [

–

]. In

particular, the cerebellum has a critical role in the generation and expression of tremor

and CD [

]. Neuroimaging studies showed greater activation of the anterior cerebellar

regions ipsilateral to the direction of head rotation and decreased activation in the posterior

cerebellar regions [

]. Patients with CD and head tremor showed higher clinical scores

of cerebellar dysfunctions, e.g., ataxia, than those without tremor [

], and dystonic tremor

improved after deep brain stimulation (DBS) of the cerebellar thalamus (ventralis interme-

diate nucleus [VIM]) [

]. On the other hand, a recent study comparing patients with CD

without tremor, CD with jerky head oscillations, and sinusoidal head oscillations showed

a distinct pallidal dysfunction in the group with sinusoidal oscillations [

]. This would

favor the hypothesis of a specific contribution of the basal ganglia to the pathophysiology

of DT [15].

Chronic DBS of the globus pallidus pars interna (GPi) is a safe and effective treatment

for advanced, disabling dystonia [

–

]. Despite successful results, pallidal stimulation

for dystonia remains a poorly standardized therapy with variable clinical outcomes [

]. A

significant challenge remains the variability of treatment benefits at an individual level. An

added complexity is that improvement after DBS is typically delayed or progressive over

months or years [35,38,39].

New implantable devices capable of chronically recording local field potentials (LFPs)

during stimulation will enable a better understanding of disease-related brain activity

patterns, their evolution over time, and their modulation in response to therapies [

This could dramatically improve tailoring treatment to each patient by adapting stimula-

tion parameters (adaptive DBS, aDBS) in response to an input signal that can represent

symptoms, motor activity, or other behavioral features [

]. In this regard, two aspects

are particularly relevant: (i) the ability to identify robust biomarkers reflecting symptoms

and their fluctuations in the context of activities of daily living and (ii) the reliability of the

device for real-time monitoring of artifact-free recordings and online adjustment of one or

more stimulation parameters [40].

One of the most promising biomarkers for dystonia is an increase in the magnitude of

oscillatory activity in the theta-alpha range (3–12 Hz) embedded in GPi-LFPs [

–

]. Neu-

mann et al. showed a correlation between these oscillations and the severity of dystonia [

but their acute suppression might not be followed by a direct change in symptom sever-

ity [

]. Additionally, the reliability of this biomarker in chronically stimulated patients

has yet to be fully explored. Piña-Fuentes et al. observed a significantly lower theta-alpha

frequency power in (still symptomatic) dystonic patients chronically treated with DBS in

comparison to newly implanted patients, even when stimulation was suspended [45].

The aim of the present study was to investigate pallidal low-frequency oscillatory

activity in our first series of dystonic patients chronically stimulated with the Percept

device (Medtronic PLC). This is one of the first commercially available devices for chronic

DBS able to continuously record LFP in real time and transmit them wirelessly to a storage

device (a tablet-user interface) [46].

In this study, we particularly focused on dystonic tremor as a possible source of arti-

factual contamination of GPi-LFP recordings. Indeed, consistent and rhythmic movements

are particularly likely to contaminate the power spectrum [

]. This would be critical in

subjects with CD as they may present head tremor at a frequency of 1–6 Hz, which falls in

the same range as the GPi-LFP’s clinically relevant spectral power [48,49].

Bioengineering 2023,10, 476 3 of 15

2. Materials and Methods

2.1. Subjects, Surgery, and Clinical Evaluation

We reviewed data collected from eight patients with idiopathic dystonia implanted in

the GPi with the Percept PC device (Medtronic, PLC), who were routinely evaluated at our

center. Demographic and clinical data are listed in Table 1. Four patients had CD, three had

myoclonus dystonia (DM), and one had segmental dystonia primarily affecting the head

and the (left and right) arm (MFD). Five of these patients showed dystonic tremors of the

head. This was defined as a spontaneous oscillatory, rhythmical head movement [1].

Table 1.

Demographic and clinical data. Abbreviations: T+—patients with head tremor; T

−

—patients

without head tremor; CD—cervical dystonia; MFD—multifocal dystonia; DM—dystonia with my-

oclonus; TWSTRS—Toronto Western Spasmodic Torticollis Rating Scale; DBS—Deep Brain Stimulation;

NA—not available. * Identifies patients studied at battery replacement. ** Age at battery replacement.

Patients DW01 (T+) DW02 (T+) DW06 (T+) DW07 *

(T+) DW08 (T+) DW03 *

(T−)

DW04 *

(T−)

DW05

(T−)

Sex F F F F F M M F

Age 52 44 57 50 33 74 62 65

Age at

onset,

years

38 36 42 Childhood Childhood 42 Childhood Childhood

Age at

surgery 50 43 56 48 ** 31 72 ** 62 ** 63

Disease CD CD CD MFD DM CD DM DM

TWSTRS

pre-DBS,

score

18 19 13 NA 22 22 NA 18

TWSTRS

post-DBS,

score

Stim-off 16 18 8 17 19 16 21 18

TWSTRS

post-DBS,

score

Stim-on 15 14 6 8 NA 5 13 6

All patients were implanted with standard non-directional DBS leads (3389, Medtronic,

PLC). The surgical procedure for DBS implantation has been described previously. Briefly,

patients underwent simultaneous bilateral stereotactic implantation of DBS electrodes

into the posteroventrolateral internal globus pallidus. The DBS electrode used was model

3389 (Medtronic, PLC), with four platinum-iridium cylindrical contacts of 1.5 mm each

and a contact-to-contact separation of 0.5 mm. The DBS electrodes were connected to an

implantable pulse generator (IPG) during the same or a subsequent surgery [

]. Three

out of eight patients received the Percept PC as a battery replacement. Along with GPi-DBS,

one patient with DM received DBS of the motor thalamus (i.e., VIM).

We assessed the severity of symptoms with the Toronto Western Spasmodic Torticollis

Rating Scale (TWSTRS, severity subscale [51,52]). The evaluations were performed before

and after DBS implantation (i.e., on the day of GPi-LFP recordings). After DBS, evaluations

were performed in both stimulation-off (stim-off) and stimulation-on (stim-on) conditions

with the clinically optimized stimulation parameters.

The local Institutional Review Board approved the study and waived review for the

data collection. Informed consent was obtained from all subjects involved in the study

according to the Declaration of Helsinki.

2.2. Experimental Setup and Recordings

We recorded GPi-LFPs from the chronically implanted electrodes in stim-off condition

at least six months after DBS implant. The stimulation was paused for at least 30 min before

Bioengineering 2023,10, 476 4 of 15

the experiment. During the recordings, patients were at rest, comfortably sitting in a chair.

They were asked to keep their eyes open without speaking or performing any voluntary

movement. The average (

standard deviation) recording length was 233.62 s (

141.68 s),

ranging from 66.8 s (patient DW02) to 451.48 s (patient DW03). According to the clinical

evaluation of a neurologist expert in movement disorders (N.P.), the subject was considered

a tremor (T+) or non-tremor (T

−

) patient. The presence of tremor was further confirmed

by inspecting the video recordings of the experiment (VIXTA, BTS).

For one T+ patient (DW02), we performed three additional recordings aimed at in-

vestigating the effect of head tremor on LFPs. First, during the execution of an alleviating

maneuver (sensory trick or geste antagoniste) [

], namely, a light touch of the face per-

formed with the left hand. This is a peculiar feature of dystonia that enables a temporary

relief of dystonic muscle contractions upon sensory stimulation [

]. Second, during

voluntary alternating (left to right and back) rhythmic movements of the head with small

and, third, large amplitude. We limited this second set of recordings to only patient DW02

as she was the only one among the T+ patients with a clinically effective sensory trick.

GPi LFPs were recorded bilaterally from all non-adjacent contact pairs (i.e., 0–3, 0–2,

1–3, where 0 is the lowermost and 3 the uppermost contact, respectively) at a sampling

frequency of 250 Hz (indefinite streaming mode) [46].

Head tremor was recorded bilaterally with surface electromyography (FREEEMG,

BTS) of the sternocleidomastoid and trapezius muscles at a sampling frequency of 1000

Hz. We chose these two muscles because they were affected by tremor in all patients

and could be easily recorded. Additionally, we placed one EMG probe on the left chest

to record heart activity and remove cardiac artifacts from the LFPs [

] and one on the

neck close to the cable connecting the implantable pulse generator (IPG) with the DBS

electrodes for synchronization purposes [

]. The method for synchronizing LFP and

EMG recordings has previously been described in [

]. Briefly, a transcutaneous

electrical nerve stimulation (TENS) burst was delivered at the level of the neck (f = 80 Hz)

at the beginning and at the end of each recording session. We used the abrupt drop-off of

the TENS artifact simultaneously recorded by the DBS electrodes and EMG probe to align

the two signals.

In the T+ group, head tremor was recorded with a triaxial inertial measurement unit

(IMU) (Opal, APDM) placed on the forehead at a sampling frequency of 128 Hz. IMU and

EMG signals were synchronized by aligning the data with respect to the rising edge of a

transistor-transistor logic (TTL) signal going from 0 to 5 V.

2.3. Data Analysis

2.3.1. Data Preprocessing

All data were imported to MATLAB (R2022b, The Mathworks, Natick, MA, USA) and

analyzed offline using custom codes.

EMG recordings were down-sampled to the LFP sampling frequency, i.e., 250 Hz. Al-

though the IPG was located in the right subclavicular region in five out of eight patients [

all LFP signals were contaminated by cardiac activity. The cardiac artifacts were removed

from the LFP signals by means of singular value decomposition (SVD). We first detected

the cardiac QRS peaks as recorded by the EMG probe placed on the chest. We then divided

the LFP signals into epochs centered on each QRS peak. For each epoch, we computed

the SVD of the LFP signals and visually identified all the eigenvectors corresponding to

the ECG artifact. We then reconstructed the cardiac artifact from these components and

subtracted it from the raw LFP [55].

To characterize tremor, EMGs were band-pass filtered in the band 1–120 Hz (second

order IIR filter). We used the absolute value of the Hilbert transform of the EMG signals for

the subsequent analysis. The IMU signals were high-pass filtered at 1 Hz (second-order

Butterworth filter).

Figure 1shows an example of the synchronized recordings of EMG, IMU, and LFP in

one patient (DW08).

Bioengineering 2023,10, 476 5 of 15

Bioengineering 2023, 10, x FOR PEER REVIEW 5 of 15

Figure 1. Patient DW08: time series of 20 s of recording of the right sternocleidomastoid muscle (first

row), IMU (second row), and LFP (third row) during rest. The IMU time series were averaged over

the three axes (x, y, z), and the LFP time series were averaged over the six contact pairs (0–3, 0–2, 1–

3, left and right). The patient showed a dystonic tremor during the acquisition, as also confirmed by

clinical notes and a video evaluation of the recording. The waveform of the tremor peaks captured

by the IMU can also be observed in the LFPs.

2.3.2. Spectral Analyses

The power spectral density (PSD) of all signals was computed using Welch’s method

with 1-s windows and a 50% overlap.

The LFP power spectra were characterized by the presence of an aperiodic part fol-

lowing a 1/f power law. Following Donoghue et al. [55,59], we modeled the observed PSD

as the sum of putative, periodic oscillatory components parameterized by their center fre-

quency, power, and bandwidth, as measured from Gaussian model fits, plus an aperiodic

component parameterized by the offset and slope of an exponentially decaying (1/f

shaped) function. Following the parameterization procedure, we removed the 1/f aperi-

odic component from the PSD to analyze only the true oscillatory components. The result-

ing LFP power spectra were then normalized to the standard deviation of the spectrum in

the range 5–95 Hz to de-emphasize spectral features prone to movement artifacts [42,60].

LFP PSDs were averaged over all contact pairs and over trials.

IMU and EMG power spectra were normalized to the standard deviation of the spec-

tra in the range 1–64 Hz and 1–125 Hz, respectively. EMG power spectra were averaged

over the four recorded muscles and over trials. IMU power spectra were averaged over

the three recorded axes and over trials.

2.3.3. Assessment of Tremor-Artifacts in LFP Recording and Comparison of IMU vs.

EMG Regression Analysis

We investigated eventual dystonic tremor contamination in the LFP by (i) comparing

PSDs across modalities and (ii) regressing multiple time-lagged copies of the EMG and

the IMU out of the LFP and evaluating the effect on LFP–PSDs.

With regards to PSD comparison, individual averaged power spectra were visually

inspected for peaks in the theta-alpha band (3–12 Hz), which corresponds to the range of

tremor [61]. We focused on the theta-alpha frequency peaks as computed on LFP, EMG,

and IMU PSDs. LFP spectra were considered contaminated by dystonic tremor when the

peak in theta-alpha power aligned across LFPs, EMGs, and IMUs.

To perform the regression, all recordings were down-sampled to the lowest available

sampling frequency: 128 Hz when IMUs were used, 250 Hz when EMG signals were used.

Figure 1.

Patient DW08: time series of 20 s of recording of the right sternocleidomastoid muscle (first

row), IMU (second row), and LFP (third row) during rest. The IMU time series were averaged over

the three axes (x, y, z), and the LFP time series were averaged over the six contact pairs (0–3, 0–2, 1–3,

left and right). The patient showed a dystonic tremor during the acquisition, as also confirmed by

clinical notes and a video evaluation of the recording. The waveform of the tremor peaks captured by

the IMU can also be observed in the LFPs.

2.3.2. Spectral Analyses

The power spectral density (PSD) of all signals was computed using Welch’s method

with 1-s windows and a 50% overlap.

The LFP power spectra were characterized by the presence of an aperiodic part fol-

lowing a 1/f power law. Following Donoghue et al. [

], we modeled the observed

PSD as the sum of putative, periodic oscillatory components parameterized by their center

frequency, power, and bandwidth, as measured from Gaussian model fits, plus an aperi-

odic component parameterized by the offset and slope of an exponentially decaying (1/f

shaped) function. Following the parameterization procedure, we removed the 1/f aperiodic

component from the PSD to analyze only the true oscillatory components. The resulting

LFP power spectra were then normalized to the standard deviation of the spectrum in the

range 5–95 Hz to de-emphasize spectral features prone to movement artifacts [

]. LFP

PSDs were averaged over all contact pairs and over trials.

IMU and EMG power spectra were normalized to the standard deviation of the spectra

in the range 1–64 Hz and 1–125 Hz, respectively. EMG power spectra were averaged over

the four recorded muscles and over trials. IMU power spectra were averaged over the three

recorded axes and over trials.

2.3.3. Assessment of Tremor-Artifacts in LFP Recording and Comparison of IMU vs. EMG

Regression Analysis

We investigated eventual dystonic tremor contamination in the LFP by (i) comparing

PSDs across modalities and (ii) regressing multiple time-lagged copies of the EMG and the

IMU out of the LFP and evaluating the effect on LFP–PSDs.

With regards to PSD comparison, individual averaged power spectra were visually

inspected for peaks in the theta-alpha band (3–12 Hz), which corresponds to the range of

tremor [

]. We focused on the theta-alpha frequency peaks as computed on LFP, EMG,

and IMU PSDs. LFP spectra were considered contaminated by dystonic tremor when the

peak in theta-alpha power aligned across LFPs, EMGs, and IMUs.

To perform the regression, all recordings were down-sampled to the lowest available

sampling frequency: 128 Hz when IMUs were used, 250 Hz when EMG signals were used.

Bioengineering 2023,10, 476 6 of 15

To regress the temporal dynamics of the EMG and IMU channels capturing head-

tremor out of the LFPs, a temporal embedding of the EMG and IMU time series was

performed. To this end, each time series x

(t) was complemented by temporally shifted

versions

∼

(t)=[x

(t+

τ1

. . .

(t+

τK

)]

,m= 1,

. . .

,M, where, x

(t+

) was the

activity of the m-th EMG or IMU sensor at time t+

. In this work, we used K = 51

equally spaced shifts for the EMGs, ranging from

τ1

−

250 samples to

τK

= 250 samples

in steps of 10 samples, and K = 52 equally spaced shifts ranging from

τ1

−

128 samples

τK

= 128 samples in steps of 5 samples for the IMU. Thus, the tremor dynamics of both

EMG and IMU were extracted within a window of two seconds and regressed out of LFP.

The relation between the embedded signal of all

EMG and/or IMU sensors,

∼

x(t)=

[∼

x1(t), . . . , ∼

xM(t), 1]T

(including an offset term) and the LFP signal

y(t)

was modeled to be

linear according to the equation

y(t)=βT∼

x(t)+yclean(t)

, where

yclean(t)

denotes residual

(genuine) LFP activity not explained by EMG or IMU. The

(K·M+1)

-dimensional vector

of regression coefficients

βOLS = (∼

X∼

T)−1∼

XyT

was estimated using ordinary least-squares

(OLS) regression, where

∼

X=h∼

x(1), . . . , ∼

x(T)i

y=[y(1), . . . , y(T)]

, and Tdenoted the

number of available paired measurements of EMG/IMU and LFP activity. Using the fitted

model, the part of the LFP signal that could be predicted from EMG/IMU was obtained

y(t)=βOLST∼

x(t)

[

]. The cleaned LFP signal was then obtained as the residual

yclean(t)=y(t)−ˆ

y(t).

Cleaned LFPs were used to compute PSDs, which were compared with those obtained

before performing the regression. The reduction in theta-alpha peak power after performing

the regression was used as the criterion to assess the degree of contamination of LFPs by

head tremor.

Because we were most interested in the theta-alpha band, we also computed the

theta-alpha peak power by integrating the LFP–PSDs over a 4 Hz wide band surrounding

the patient-specific theta-alpha peak (peak frequency

2 Hz; 4 bins) (i) without regression,

(ii) with EMG regression, and (iii) with IMU regression.

2.3.4. Evaluation of Pallido-Muscular Coherence with IMU-Regression

Coherence (COH) is a frequency-domain measure of the linear phase and amplitude

relationships between signals. It can reveal spectrally specific functional connectivity, and

it is an established method in neuroscience [63].

To investigate whether EMG and IMU carry different information about tremor, we

computed COH between LFP and EMG before and after IMU regression. We postulated

that the reduction in coherence after IMU regression indicates that EMG and IMU signals

would carry to some extent overlapping information. As such, the LFP–EMG coherence

would be due to tremor contamination, thus artifactual.

Additionally, we computed (i) LFP–IMU COH without and with IMU regression and

(ii) LFP–EMG COH without and with EMG regression.

We computed coherence at frequency

COH(X,Y,f)=|SXY(f)|2./SXX(f)SYY(f)

where

SXX

SYY

represent the auto-spectrum of signals Xand Y, respectively, and

SXY

represents the cross-spectrum [

]. Note that, being already normalized, coherence spectra

do not require additional normalization for across-patient analysis and do not have a

prominent 1/f aperiodic component [42].

2.3.5. Statistical Analysis

Statistically significant differences between (i) LFPs theta-alpha peak power with and

without IMU regression, (ii) theta-alpha LFP–IMU COH with and without IMU regression,

and (iii) theta-alpha pallido-muscular COH with and without IMU or EMG regression were

assessed by means of a matched-pairs Wilcoxon signed rank test with a significance level

of 0.05.

Bioengineering 2023,10, 476 7 of 15

Significant values of pallido-muscular COH were determined by statistical comparison

with a population of 1000 surrogate COH values in which any COH was destroyed. Surrogate

data were obtained by randomly shifting the EMG signal 1000 times by a random offset of at

least 2 s before computing the LFP–EMG COH. The significance level was set to 0.05.

3. Results

3.1. Clinical and Demographic Data

All but one patient (DW08) significantly benefited from GPi-DBS. The improvement

for this patient was limited by the development of typical GPi-DBS side effects (i.e., bradyki-

nesia). The overall average benefit from GPi-DBS was about 40% (TWSTRS score stim-off:

16.62 ±3.85 and stim-on: 9.57 ±4.27, mean ±standard deviation) (Table 1).

3.2. Tremor Contamination of LFP Recordings and Comparison of IMU vs. EMG

Regression Analysis

Two patients (DW01 and DW08) showed tremor contamination of the LFP, as sug-

gested by the alignment of the frequency peaks in the PSDs around the head tremor

frequency (DW01: f = 4 Hz, DW08: f = 4 Hz) (Figure 2).

Bioengineering 2023, 10, x FOR PEER REVIEW 7 of 15

Surrogate data were obtained by randomly shifting the EMG signal 1000 times by a ran-

dom offset of at least 2 s before computing the LFP–EMG COH. The significance level was

set to 0.05.

3. Results

3.1. Clinical and Demographic Data

All but one patient (DW08) significantly benefited from GPi-DBS. The improvement

for this patient was limited by the development of typical GPi-DBS side effects (i.e., brad-

ykinesia). The overall average benefit from GPi-DBS was about 40% (TWSTRS score stim-

off: 16.62 ± 3.85 and stim-on: 9.57 ± 4.27, mean ± standard deviation) (Table 1).

3.2. Tremor Contamination of LFP Recordings and Comparison of IMU vs. EMG Regression

Analysis

Two patients (DW01 and DW08) showed tremor contamination of the LFP, as sug-

gested by the alignment of the frequency peaks in the PSDs around the head tremor fre-

quency (DW01: f = 4 Hz, DW08: f = 4 Hz) (Figure 2).

Figure 2. Spectral profiles of the LFP (red line), EMG (blue line), and IMU (yellow line) recordings

of all patients. First row: patients with tremor (T+); second row: patients without tremor (T−).

Regression-based removal of IMU signals from LFPs led to power reduction, espe-

cially in the theta-alpha band, in all T+ patients (Figure 3). Regression of the EMG signal

was instead not affecting LFP theta-alpha peak power except for three patients (DW02,

DW06, and DW08) (Figure 3 and Table 2).

Figure 2.

Spectral profiles of the LFP (red line), EMG (blue line), and IMU (yellow line) recordings of

all patients. First row: patients with tremor (T+); second row: patients without tremor (T−).

Regression-based removal of IMU signals from LFPs led to power reduction, especially

in the theta-alpha band, in all T+ patients (Figure 3). Regression of the EMG signal was

instead not affecting LFP theta-alpha peak power except for three patients (DW02, DW06,

and DW08) (Figure 3and Table 2).

Bioengineering 2023,10, 476 8 of 15

Bioengineering 2023, 10, x FOR PEER REVIEW 8 of 15

Figure 3. LFP spectral profiles before regression (red line), after EMG regression (blue line), and

after IMU regression (yellow line). First row: patients with tremor (T+); second row: patients without

tremor (T−). Abbreviation: reg (regression).

Table 2. Columns 2–4: theta-alpha peak power [au] without and with EMG or IMU regression. Dif-

ferences (Δ) between LFP theta-alpha peak power without and with EMG or IMU regression are

shown in brackets. Columns 5–6: averaged theta-alpha pallido-muscular COH without and with

IMU regression. The differences (Δ) between LFP and EMG COH without and with IMU regression

are shown in brackets. Abbreviations: COH—coherence; EMG—electromyographic recording;

IMU—inertial measurement unit recording; LFP—local field potential; reg—regression; NA—not

available.

Patient

LFP θ-α Power

LFP θ-α Power EMG

Reg (Δ)

LFP θ-α Power IMU Reg

(Δ)

LFP-EMG θ-α

COH

LFP-EMG θ-α COH—IMU

Reg (Δ)

DW01

17.84

17.93 (−0.09)

15.01 (2.83)

0.015

0.008 (0.007)

DW02

8.16

7.03 (1.13)

4.50 (3.66)

0.020

0.007 (0.013)

DW06

14.34

12.85 (1.49)

9.78 (4.56)

0.002

0.002 (0)

DW07

14.56

13.70 (0.86)

10.59 (3.97)

0.007

0.007 (0)

DW08

6.07

2.66 (3.41)

−2.21 (8.28)

0.010

0.002 (0.008)

DW03

16.32

16.22 (0.10)

DW04

11.06

10.88 (0.18)

DW05

7.33

8.00 (0.67)

The LFP theta-alpha peak power of each patient before and after EMG and IMU re-

gression are shown in Table 2. IMU regression reduced the theta-alpha peak power more

in comparison to EMG regression. Differences between LFPs theta-alpha peak power with

and without IMU regression were statistically significant (4.66 ± 2.12 au (mean ± standard

deviation)).

3.3. Pallido-Muscular Coherence after Cleaning

LFP–IMU and LFP–EMG COH profiles are shown in Figure 4. The COH drop after

regression analysis was significant for both LFP–IMU COH and LFP–EMG COH.

Figure 3.

LFP spectral profiles before regression (red line), after EMG regression (blue line), and after

IMU regression (yellow line). First row: patients with tremor (T+); second row: patients without

tremor (T−). Abbreviation: reg (regression).

Table 2.

Columns 2–4: theta-alpha peak power [au] without and with EMG or IMU regression. Differ-

ences (

∆

) between LFP theta-alpha peak power without and with EMG or IMU regression are shown in

brackets. Columns 5–6: averaged theta-alpha pallido-muscular COH without and with IMU regression.

The differences (

∆

) between LFP and EMG COH without and with IMU regression are shown in brackets.

Abbreviations: COH—coherence; EMG—electromyographic recording; IMU—inertial measurement

unit recording; LFP—local field potential; reg—regression; NA—not available.

Patient LFP θ-α

Power

LFP θ-α

Power EMG

Reg (∆)

LFP θ-α

Power IMU

Reg (∆)

LFP-EMG

θ-αCOH

LFP-EMG

θ-α

COH—IMU

Reg (∆)

DW01 17.84 17.93 (−0.09) 15.01 (2.83) 0.015 0.008 (0.007)

DW02 8.16 7.03 (1.13) 4.50 (3.66) 0.020 0.007 (0.013)

DW06 14.34 12.85 (1.49) 9.78 (4.56) 0.002 0.002 (0)

DW07 14.56 13.70 (0.86) 10.59 (3.97) 0.007 0.007 (0)

DW08 6.07 2.66 (3.41) −2.21 (8.28) 0.010 0.002 (0.008)

DW03 16.32 16.22 (0.10) NA NA NA

DW04 11.06 10.88 (0.18) NA NA NA

DW05 7.33 8.00 (0.67) NA NA NA

The LFP theta-alpha peak power of each patient before and after EMG and IMU

regression are shown in Table 2. IMU regression reduced the theta-alpha peak power more

in comparison to EMG regression. Differences between LFPs theta-alpha peak power with

and without IMU regression were statistically significant (4.66

2.12 au (mean

stan-

dard deviation)).

3.3. Pallido-Muscular Coherence after Cleaning

LFP–IMU and LFP–EMG COH profiles are shown in Figure 4. The COH drop after

regression analysis was significant for both LFP–IMU COH and LFP–EMG COH.

Bioengineering 2023,10, 476 9 of 15

Bioengineering 2023, 10, x FOR PEER REVIEW 9 of 15

Figure 4. LFP–IMU COH (first row) without (continuous, yellow line) and with (dotted, black line)

IMU regression in patients T+. LFP–EMG COH (second row) without (continuous, blue line) and

with (dotted, black line) EMG regression in patients T+.

To investigate the effects of tremor-related artifacts on pallido-muscular COH, we

computed LFP–EMG COH with and without regressing the IMU out of the LFP signals.

LFP–EMG COH before IMU regression was small but significant with respect to shuffled

data (Figure 5). IMU regression significantly reduced pallido-muscular COH in all but

two subjects (DW06 and DW07) (Table 2 and Figure 5).

Figure 5. Pallido-muscular COH without (continuous, light blue line) and with (dotted, light blue

line) IMU regression. Surrogate data are shown in gray as the mean (continuous line) ±2 standard

deviations (shaded area).

3.4. Sensory Trick and Voluntary Rhythmic Movement

In one patient (DW02) with a clinically effective sensory trick (i.e., maneuver allevi-

ating dystonic tremor), we computed the PSD during the sensory trick and voluntary al-

ternating head movements with small and large amplitude. For each condition, we per-

formed the regression of IMU and EMG.

The sensory trick reduced the theta-alpha peak power in the LFP–PSDs as compared

to baseline (i.e., dystonic tremor condition) (Figure 6). However, both EMG and IMU de-

tected residual tremor activity (Figure 6), which was effectively removed from LFP–PSDs

with IMU regression. Of note, the PSDs of the IMU and LFP do not show an aligned peak

(Figure 6), thus suggesting that LFPs manifest harmonics of the artifactual frequency.

When looking at voluntary movements, LFP–PSDs showed movement-related arti-

facts at the head oscillation frequency (Figure 6).

Figure 4.

LFP–IMU COH (first row) without (continuous, yellow line) and with (dotted, black line)

IMU regression in patients T+. LFP–EMG COH (second row) without (continuous, blue line) and

with (dotted, black line) EMG regression in patients T+.

To investigate the effects of tremor-related artifacts on pallido-muscular COH, we

computed LFP–EMG COH with and without regressing the IMU out of the LFP signals.

LFP–EMG COH before IMU regression was small but significant with respect to shuffled

data (Figure 5). IMU regression significantly reduced pallido-muscular COH in all but two

subjects (DW06 and DW07) (Table 2and Figure 5).

Bioengineering 2023, 10, x FOR PEER REVIEW 9 of 15

Figure 4. LFP–IMU COH (first row) without (continuous, yellow line) and with (dotted, black line)

IMU regression in patients T+. LFP–EMG COH (second row) without (continuous, blue line) and

with (dotted, black line) EMG regression in patients T+.

To investigate the effects of tremor-related artifacts on pallido-muscular COH, we

computed LFP–EMG COH with and without regressing the IMU out of the LFP signals.

LFP–EMG COH before IMU regression was small but significant with respect to shuffled

data (Figure 5). IMU regression significantly reduced pallido-muscular COH in all but

two subjects (DW06 and DW07) (Table 2 and Figure 5).

Figure 5. Pallido-muscular COH without (continuous, light blue line) and with (dotted, light blue

line) IMU regression. Surrogate data are shown in gray as the mean (continuous line) ±2 standard

deviations (shaded area).

3.4. Sensory Trick and Voluntary Rhythmic Movement

In one patient (DW02) with a clinically effective sensory trick (i.e., maneuver allevi-

ating dystonic tremor), we computed the PSD during the sensory trick and voluntary al-

ternating head movements with small and large amplitude. For each condition, we per-

formed the regression of IMU and EMG.

The sensory trick reduced the theta-alpha peak power in the LFP–PSDs as compared

to baseline (i.e., dystonic tremor condition) (Figure 6). However, both EMG and IMU de-

tected residual tremor activity (Figure 6), which was effectively removed from LFP–PSDs

with IMU regression. Of note, the PSDs of the IMU and LFP do not show an aligned peak

(Figure 6), thus suggesting that LFPs manifest harmonics of the artifactual frequency.

When looking at voluntary movements, LFP–PSDs showed movement-related arti-

facts at the head oscillation frequency (Figure 6).

Figure 5.

Pallido-muscular COH without (continuous, light blue line) and with (dotted, light blue

line) IMU regression. Surrogate data are shown in gray as the mean (continuous line)

2 standard

deviations (shaded area).

3.4. Sensory Trick and Voluntary Rhythmic Movement

In one patient (DW02) with a clinically effective sensory trick (i.e., maneuver alle-

viating dystonic tremor), we computed the PSD during the sensory trick and voluntary

alternating head movements with small and large amplitude. For each condition, we

performed the regression of IMU and EMG.

The sensory trick reduced the theta-alpha peak power in the LFP–PSDs as compared

to baseline (i.e., dystonic tremor condition) (Figure 6). However, both EMG and IMU

detected residual tremor activity (Figure 6), which was effectively removed from LFP–PSDs

with IMU regression. Of note, the PSDs of the IMU and LFP do not show an aligned peak

(Figure 6), thus suggesting that LFPs manifest harmonics of the artifactual frequency.

Bioengineering 2023,10, 476 10 of 15

Bioengineering 2023, 10, x FOR PEER REVIEW 10 of 15

Figure 6. Patient DW02: time series of 5 s recordings at rest (first column), while performing the

sensory trick (second column), during voluntary alternating rhythmic movement of the head with

small amplitude (third column), and with large amplitude (fourth column), of the right sternocleido-

mastoid EMG (first row), IMU (second row), and LFP (third row) without regression. IMU and LFP

time series were averaged over channels, i.e., over the three axes (x, y, z) and over the six contact

pairs (0–3, 0–2, 1–3 left and right), respectively. The corresponding PSDs are shown in the fourth

row: LFP without regression (continuous red line), LFP with EMG regression (continuous blue line),

LFP with IMU regression (continuous yellow line), EMG (dotted blue line), and IMU (dotted yellow

line). Abbreviations: VRM—voluntary rhythmic movement.

4. Discussion

New DBS devices with sensing capabilities open new opportunities to improve the

clinical effectiveness of DBS by optimizing stimulation parameters in response to an input

signal representing symptoms, motor activity, or other behavioral characteristics [40]. Pal-

lidal theta-alpha oscillatory activity is a promising biomarker for dystonia [45] and for

future therapeutic development [46], such as improving efficacy and the timing of the

therapeutic response while reducing side effects and battery consumption.

Our results show that the Percept PC is capable of recording low-frequency oscilla-

tions in chronically stimulated patients with CD but also confirm a spectral contamination

due to movement artifacts (voluntary and pathological) [46].

Already, a visual inspection of the PSDs suggested artifact contamination in two sub-

jects with dystonic tremor, with GPi-LFPs and both EMG and IMU PSDs exhibiting spec-

tral peaks at the tremor frequency (Figure 2). When applying the IMU regression, all T+

subjects showed a reduction in power, indicating a tremor-related contamination of the

raw GPi-LFP signals (Figure 3).

In our study, we showed that IMU regression was superior to EMG regression in

removing the tremor-related artifact from the GPi-LFPs (Figure 3). This difference in ef-

fectiveness may be due to the greater capacity of IMUs to record a composite and irregular

tremor such as dystonic tremor more distinctly [64]. Furthermore, EMG signals can be

affected by skin conductance and probe placement, two aspects that increase intra- and

inter-subject variability. However, we made sure that all EMG probes were adherent to

the skin and placed near the bellies of selected muscles. Before electrode placement, the

skin was cleaned with alcohol.

Figure 6.

Patient DW02: time series of 5 s recordings at rest (first column), while performing the

sensory trick (second column), during voluntary alternating rhythmic movement of the head with

small amplitude (third column), and with large amplitude (fourth column), of the right sternoclei-

domastoid EMG (first row), IMU (second row), and LFP (third row) without regression. IMU and

LFP time series were averaged over channels, i.e., over the three axes (x, y, z) and over the six contact

pairs (0–3, 0–2, 1–3 left and right), respectively. The corresponding PSDs are shown in the fourth row:

LFP without regression (continuous red line), LFP with EMG regression (continuous blue line), LFP

with IMU regression (continuous yellow line), EMG (dotted blue line), and IMU (dotted yellow line).

Abbreviations: VRM—voluntary rhythmic movement.

When looking at voluntary movements, LFP–PSDs showed movement-related artifacts

at the head oscillation frequency (Figure 6).

4. Discussion

New DBS devices with sensing capabilities open new opportunities to improve the

clinical effectiveness of DBS by optimizing stimulation parameters in response to an input

signal representing symptoms, motor activity, or other behavioral characteristics [

Pallidal theta-alpha oscillatory activity is a promising biomarker for dystonia [

] and

for future therapeutic development [

], such as improving efficacy and the timing of the

therapeutic response while reducing side effects and battery consumption.

Our results show that the Percept PC is capable of recording low-frequency oscillations

in chronically stimulated patients with CD but also confirm a spectral contamination due

to movement artifacts (voluntary and pathological) [46].

Already, a visual inspection of the PSDs suggested artifact contamination in two

subjects with dystonic tremor, with GPi-LFPs and both EMG and IMU PSDs exhibiting

spectral peaks at the tremor frequency (Figure 2). When applying the IMU regression, all

T+ subjects showed a reduction in power, indicating a tremor-related contamination of the

raw GPi-LFP signals (Figure 3).

In our study, we showed that IMU regression was superior to EMG regression in

removing the tremor-related artifact from the GPi-LFPs (Figure 3). This difference in

effectiveness may be due to the greater capacity of IMUs to record a composite and irregular

tremor such as dystonic tremor more distinctly [

]. Furthermore, EMG signals can be

Bioengineering 2023,10, 476 11 of 15

affected by skin conductance and probe placement, two aspects that increase intra- and

inter-subject variability. However, we made sure that all EMG probes were adherent to the

skin and placed near the bellies of selected muscles. Before electrode placement, the skin

was cleaned with alcohol.

It should be noted that the EMG signal carries more information than just the kinematic

(and artifactual) component of tremor, such as cross-talk from different muscles or a

pathological dystonic activity [

]. To evaluate this issue, we computed pallido-muscular

COH to estimate the information carried by IMU and EMG signals. After IMU regression,

we observed a decrease in LFP–EMG COH at the tremor frequency in three out of five

subjects (Figure 5), thus supporting the idea, in our case, of an artifactual origin of theta-

alpha pallido-muscular COH.

Finally, we tested IMU and EMG-based multiple regression under different conditions,

namely voluntary rhythmic head movement of small and large amplitude. This additional

evaluation was performed only on the patient who showed a clinically effective sensory

trick, a maneuver able to reduce tremor severity (Figure 6). We showed that both IMU

and EMG regressions were effective in capturing voluntary head movements and led to a

correction of GPi-LFPs proportional to the movement’s amplitude. Averna et al. [

] also

showed a drastic increase in the theta frequency range during neck tilting and upper limb

movement. This might refer to the higher susceptibility of some implants to contamination.

Moreover, IMU regression captured the residual tremor activity present during the sensory

trick maneuver that was not clinically visible. The sensory trick did not affect the rhythmic

spiking of the EMG, as expected [

], but EMG regression led to a marginal correction

of GPi-LFPs (Figure 6).

This study has some limitations. First, the limited number of patients reduced the

number of recordings that were available for our analysis. Considering the low prevalence of

the disease (16.3 per 100,000 [

]), the limited number of dystonic patients implanted with

DBS [

], and the recent release of the Percept PC device on the market (2020), the sample

size is still considerable and comparable with previous studies on dystonia [

]. The

same limiting factors prevented the recruitment of patients with similar clinical presentations.

However, we do not expect clinical features other than tremor to be relevant for the technical

aims of the current work. Second, we used relatively short recordings, with lengths ranging

from one to four minutes. Nevertheless, this was considered sufficient to calculate power and

coherence measurements, as Popov et al. have observed good to excellent test-retest reliability

of resting-state power and coherence in a large sample based on recordings that were just

100 s long [

]. Potential biases of coherence on recording length [

] were not addressed

here, as all statistical comparisons were performed within subjects on data of the same length.

Third, we could not record and compare tremor-free GPi-LFP recordings for the same patients.

Nevertheless, we documented the performance of IMU regression in one patient with a

clinically effective sensory trick. Lastly, the observed theta-alpha power reduction is not

sufficient proof of the cleaning efficacy. It represents the neural activity after removal of the

component that is coherent with the head tremor, which appears to introduce artifactual

spectral content in this low frequency range. Although a possible desynchronization of a

genuine neural component associated with tremor cannot be ruled out, this interpretation

is less likely because, to the best of our knowledge, dystonic tremor has not been associated

with any oscillatory activity of the LFPs to date [47].

5. Conclusions

LFP-based biomarker detection will become standard in clinical practice, thus en-

abling better understanding and monitoring of distinctive neural signatures associated

with specific symptoms or behaviors. Pallidal theta-alpha oscillations may be critical for

understanding the pathophysiology of dystonia. They could act as a useful biomarker

not only for programming stimulation parameters but also for adaptive DBS. However,

critical use of newly available technologies is necessary to address possible drawbacks.

Our work suggests caution when considering LFP recordings, as they may be susceptible

Bioengineering 2023,10, 476 12 of 15

to contamination by movement artifacts. This is particularly the case for head tremor in

dystonia but also applies to patients with repetitive involuntary movements such as essen-

tial tremor, Tourette syndrome, or Parkinson’s disease. Neglecting this contamination can

lead to misinterpretation or hiding significant findings. We here provide methodological

guidance on how to clean LFP recordings from head tremor artifacts. Regressing out head

motions concurrently recorded with IMU might substantially alleviate LFP contamination

and facilitate the neurophysiological interpretation of LFP analyses.

Author Contributions:

Conceptualization, J.D.V.D.V., I.U.I., S.H. and C.P.; methodology, J.D.V.D.V.,

S.H. and C.P.; software, J.D.V.D.V. and S.H.; validation, J.D.V.D.V. and I.H.; formal analysis, J.D.V.D.V.

and C.P.; investigation, J.D.V.D.V., I.H., N.G.P., P.C., I.U.I. and C.P.; resources, I.U.I.; data curation,

N.G.P., I.H. and P.C.; writing—original draft preparation, J.D.V.D.V., I.H. and N.G.P. writing—review

and editing, P.C., I.U.I., S.H. and C.P.; visualization, J.D.V.D.V. and C.P.; supervision, I.U.I., S.H. and

C.P.; project administration, I.U.I. and C.P.; funding acquisition, I.U.I. All authors have read and

agreed to the published version of the manuscript.

Funding:

The study was sponsored by the Deutsche Forschungsgemeinschaft (DFG, German Re-

search Foundation)—Project-ID 424778381-TRR 295 and the Fondazione Grigioni per il Morbo di

Parkinson. IUI was supported by a grant from the New York University School of Medicine and The

Marlene and Paolo Fresco Institute for Parkinson’s and Movement Disorders, which was made possi-

ble with support from Marlene and Paolo Fresco. This publication was supported by the Open Access

Publication Fund of the University of Würzburg. J.D.V.D.V. and I.H. received financial support from

the Graduate School of Life Sciences, University of Würzburg. I.H. was supported by a scholarship

from the German Academic Exchange Service (DAAD; Deutscher Akademischer Austauschdienst).

S.H. received funding from the European Research Council (ERC) under the European Union’s

Horizon 2020 research and innovation program (Grant agreement No. 758985).

Institutional Review Board Statement:

The study was conducted in accordance with the Declaration

of Helsinki and approved by the Ethical Committee of the University Hospital Würzburg (n. 20230120 01).

Informed Consent Statement:

Written informed consent was obtained from the patients to publish

this paper.

Data Availability Statement:

The data presented in this study are available upon request. The

data are not publicly available due to privacy reasons. Inquiries can be filed to the corresponding

author (Jasmin Del Vecchio Del Vecchio, University Hospital Würzburg, Department of Neurology;

Josef-Schneider-Straße 11, 97080 Würzburg; phone: +49-(0)931/201-23605; fax: +49-(0)931/201-24901;

E-Mail: delvecchio_j@ukw.de).

Acknowledgments:

The authors would like to thank P. Fricke, C. Matthies and R. Nickl for the

neurosurgical information and J. Volkmann for the valuable comments.

Conflicts of Interest:

J.D.V.D.V., I.H., N.G.P., P.C., I.U.I., S.H. and C.P. declare no conflicts of interest

related to this study. The Percept PC (Medtronic, PLC) and the related hardware and software for

programming and readout are commercially available. The companies had no impact on study

design, patient selection, data analysis, or reporting of the results.

References

Albanese, A.; Bhatia, K.; Bressman, S.B.; DeLong, M.R.; Fahn, S.; Fung, V.S.C.; Hallett, M.; Jankovic, J.; Jinnah, H.A.; Klein, C.;

et al. Phenomenology and Classification of Dystonia: A Consensus Update: Dystonia: Phenomenology and Classification. Mov.

Disord. 2013,28, 863–873. [CrossRef] [PubMed]

2. Pandey, S.; Sarma, N. Tremor in dystonia. Park. Relat. Disord. 2016,29, 3–9. [CrossRef] [PubMed]

Shaikh, A.G.; Beylergil, S.B.; Scorr, L.; Kilic-Berkmen, G.; Freeman, A.; Klein, C.; Junker, J.; Loens, S.; Brüggemann, N.; Münchau,

A.; et al. Dystonia and Tremor. Neurology 2021,96, e563–e574. [CrossRef] [PubMed]

Bhatia, K.P.; Bain, P.; Bajaj, N.; Elble, R.J.; Hallett, M.; Louis, E.D.; Raethjen, J.; Stamelou, M.; Testa, C.M.; Deuschl, G.; et al.

Consensus Statement on the classification of tremors. from the task force on tremor of the International Parkinson and Movement

Disorder Society: IPMDS Task Force on Tremor Consensus Statement. Mov. Disord. 2018,33, 75–87. [CrossRef]

Defazio, G.; Conte, A.; Gigante, A.F.; Fabbrini, G.; Berardelli, A. Is tremor in dystonia a phenotypic feature of dystonia? Neurology

2015,84, 1053. [CrossRef]

Panyakaew, P.; Cho, H.J.; Lee, S.W.; Wu, T.; Hallett, M. The Pathophysiology of Dystonic Tremors and Comparison With Essential

Tremor. J. Neurosci. Off. J. Soc. Neurosci. 2020,40, 9317–9326. [CrossRef]

Bioengineering 2023,10, 476 13 of 15

7. Hallett, M. Neurophysiology of dystonia: The role of inhibition. Neurobiol. Dis. 2011,42, 177–184. [CrossRef]

Atkinson-Clement, C.; Pinto, S.; Eusebio, A.; Coulon, O. Diffusion tensor imaging in Parkinson’s disease: Review and meta-

analysis. Neuroimage Clin. 2017,16, 98–110. [CrossRef]

Quartarone, A.; Ruge, D. How Many Types of Dystonia? Pathophysiological Considerations. Front. Neurol.

2018

,9, 12. [CrossRef]

[PubMed]

10.

Münchau, A.; Schrag, A.; Chuang, C.; MacKinnon, C.D.; Bhatia, K.P.; Quinn, N.P.; Rothwell, J.C. Arm tremor in cervical dystonia

differs from essential tremor and can be classified by onset age and spread of symptoms. Brain J. Neurol.

2001

,124, 1765–1776.

[CrossRef]

11.

Tinazzi, M.; Fasano, A.; Di Matteo, A.; Conte, A.; Bove, F.; Bovi, T.; Peretti, A.; Defazio, G.; Fiorio, M.; Berardelli, A. Temporal

discrimination in patients with dystonia and tremor and patients with essential tremor. Neurology 2013,80, 76. [CrossRef]

12.

Conte, A.; Ferrazzano, G.; Belvisi, D.; Manzo, N.; Battista, E.; Voti, P.L.; Nardella, A.; Fabbrini, G.; Berardelli, A. Somatosensory

temporal discrimination in Parkinson’s disease, dystonia and essential tremor: Pathophysiological and clinical implications. Clin.

Neurophysiol. 2018,129, 1849–1853. [CrossRef] [PubMed]

13.

Cerasa, A.; Nisticò, R.; Salsone, M.; Bono, F.; Salvino, D.; Morelli, M.; Arabia, G.; Quattrone, A. Neuroanatomical correlates of

dystonic tremor: A cross-sectional study. Park. Relat. Disord. 2014,20, 314–317. [CrossRef]

14.

Kirke, D.N.; Battistella, G.; Kumar, V.; Rubien-Thomas, E.; Choy, M.; Rumbach, A.; Simonyan, K. Neural correlates of dystonic

tremor: A multimodal study of voice tremor in spasmodic dysphonia. Brain Imaging Behav. 2017,11, 166–175. [CrossRef]

15.

Madelein van der Stouwe, A.M.; Nieuwhof, F.; Helmich, R.C. Tremor pathophysiology: Lessons from neuroimaging. Curr. Opin.

Neurol. 2020,33, 474–481. [CrossRef] [PubMed]

16.

Tsuboi, T.; Wong, J.K.; Eisinger, R.S.; Okromelidze, L.; Burns, M.R.; Ramirez-Zamora, A.; Almeida, L.; Wagle Shukla, A.; Foote,

K.D.; Okun, M.S.; et al. Comparative connectivity correlates of dystonic and essential tremor deep brain stimulation. Brain

2021

144, 1774–1786. [CrossRef] [PubMed]

17.

Nieuwhof, F.; Panyakaew, P.; van de Warrenburg, B.P.; Gallea, C.; Helmich, R.C. The patchy tremor landscape: Recent advances

in pathophysiology. Curr. Opin. Neurol. 2018,31, 455–461. [CrossRef] [PubMed]

18.

Panyakaew, P.; Jinnah, H.A.; Shaikh, A.G. Clinical features, pathophysiology, treatment, and controversies of tremor in dystonia.

J. Neurol. Sci. 2022,435, 120199. [CrossRef]

19.

DeSimone, J.C.; Archer, D.B.; Vaillancourt, D.E.; Wagle Shukla, A. Network-level connectivity is a critical feature distinguishing

dystonic tremor and essential tremor. Brain 2019,142, 1644–1659. [CrossRef]

20.

van den Berg, K.R.E.; Helmich, R.C. The Role of the Cerebellum in Tremor–Evidence from Neuroimaging. Tremor Hyperkinetic

Mov. 2021,11, 49. [CrossRef]

21.

Ma, H.; Chen, H.; Fang, J.; Gao, L.; Ma, L.; Wu, T.; Hou, Y.; Zhang, J.; Feng, T. Resting-state functional connectivity of dentate

nucleus is associated with tremor in Parkinson’s disease. J. Neurol. 2015,262, 2247–2256. [CrossRef]

22.

Battaglia, S.; Thayer, J.F. Functional interplay between central and autonomic nervous systems in human fear conditioning. Trends

Neurosci. 2022,45, 504–506. [CrossRef]

23.

Ippolito, G.; Bertaccini, R.; Tarasi, L.; Di Gregorio, F.; Trajkovic, J.; Battaglia, S.; Romei, V. The Role of Alpha Oscillations among

the Main Neuropsychiatric Disorders in the Adult and Developing Human Brain: Evidence from the Last 10 Years of Research.

Biomedicines 2022,10, 3189. [CrossRef] [PubMed]

24.

Alongi, P.; Iaccarino, L.; Perani, D. PET Neuroimaging: Insights on Dystonia and Tourette Syndrome and Potential Applications.

Front. Neurol. 2014,5, 183. [CrossRef] [PubMed]

25.

Poston, K.L.; Eidelberg, D. Functional brain networks and abnormal connectivity in the movement disorders. NeuroImage

2012

62, 2261–2270. [CrossRef]

26.

Lehéricy, S.; Tijssen, M.A.J.; Vidailhet, M.; Kaji, R.; Meunier, S. The anatomical basis of dystonia: Current view using neuroimaging.

Mov. Disord. 2013,28, 944–957. [CrossRef] [PubMed]

27. Carbon, M.; Argyelan, M.; Eidelberg, D. Functional imaging in hereditary dystonia. Eur. J. Neurol. 2010,17, 58–64. [CrossRef]

28.

Merola, A.; Dwivedi, A.K.; Shaikh, A.G.; Tareen, T.K.; Da Prat, G.A.; Kauffman, M.A.; Hampf, J.; Mahajan, A.; Marsili, L.; Jankovic,

J.; et al. Head tremor at disease onset: An ataxic phenotype of cervical dystonia. J. Neurol. 2019,266, 1844–1851. [CrossRef]

29.

Prudente, C.N.; Stilla, R.; Singh, S.; Buetefisch, C.; Evatt, M.; Factor, S.A.; Freeman, A.; Hu, X.P.; Hess, E.J.; Sathian, K.; et al. A

Functional Magnetic Resonance Imaging Study of Head Movements in Cervical Dystonia. Front. Neurol.

2016

,7, 201. [CrossRef]

30.

Filip, P.; Gallea, C.; Lehéricy, S.; Bertasi, E.; Popa, T.; Mareˇcek, R.; Lungu, O.V.; Kašpárek, T.; Vaníˇcek, J.; Bareš, M. Disruption in

cerebellar and basal ganglia networks during a visuospatial task in cervical dystonia. Mov. Disord.

2017

,32, 757–768. [CrossRef]

31.

Tsuboi, T.; Jabarkheel, Z.; Foote, K.D.; Okun, M.S.; Wagle Shukla, A. Importance of the initial response to GPi deep brain

stimulation in dystonia: A nine year quality of life study. Park. Relat. Disord. 2019,64, 249–255. [CrossRef] [PubMed]

32.

Sedov, A.; Usova, S.; Semenova, U.; Gamaleya, A.; Tomskiy, A.; Beylergil, S.B.; Jinnah, H.A.; Shaikh, A.G. Pallidal Activity in

Cervical Dystonia with and Without Head Tremor. Cerebellum Lond. Engl. 2020,19, 409–418. [CrossRef]

33.

Kiss, Z.H.T.; Doig-Beyaert, K.; Eliasziw, M.; Tsui, J.; Haffenden, A.; Suchowersky, O. The Canadian multicentre study of deep

brain stimulation for cervical dystonia. Brain 2007,130, 2879–2886. [CrossRef]

34.

Volkmann, J.; Wolters, A.; Kupsch, A.; Müller, J.; Kühn, A.A.; Schneider, G.-H.; Poewe, W.; Hering, S.; Eisner, W.; Müller, J.-U.; et al.

Pallidal deep brain stimulation in patients with primary generalised or segmental dystonia: 5-year follow-up of a randomised

trial. Lancet Neurol. 2012,11, 1029–1038. [CrossRef]

Bioengineering 2023,10, 476 14 of 15

35.

Isaias, I.U.; Alterman, R.L.; Tagliati, M. Deep Brain Stimulation for Primary Generalized Dystonia: Long-term Outcomes. Arch.

Neurol. 2009,66, 465–470. [CrossRef] [PubMed]

36.

Kaelin-Lang, A.; You, H.; Burgunder, J.-M.; Lönnfors-Weitze, T.; Loher, T.J.; Taub, E.; Isaias, I.U.; Krauss, J.K.; Michael Schüpbach,

W.M. Bilateral pallidal stimulation improves cervical dystonia for more than a decade. Park. Relat. Disord.

2020

,81, 78–81.

[CrossRef]

37.

Tisch, S. Deep brain stimulation in dystonia: Factors contributing to variability in outcome in short and long term follow-up.

Curr. Opin. Neurol. 2022,35, 510–517. [CrossRef]

38.

Isaias, I.U.; Volkmann, J.; Kupsch, A.; Burgunder, J.-M.; Ostrem, J.L.; Alterman, R.L.; Mehdorn, H.M.; Sch?necker, T.; Krauss, J.K.;

Starr, P.; et al. Factors predicting protracted improvement after pallidal DBS for primary dystonia: The role of age and disease

duration. J. Neurol. 2011,258, 1469–1476. [CrossRef] [PubMed]

39.

Reich, M.M.; Horn, A.; Lange, F.; Roothans, J.; Paschen, S.; Runge, J.; Wodarg, F.; Pozzi, N.G.; Witt, K.; Nickl, R.C.; et al.

Probabilistic mapping of the antidystonic effect of pallidal neurostimulation: A multicentre imaging study. Brain

2019

,142,

1386–1398. [CrossRef]

40.

Vissani, M.; Isaias, I.U.; Mazzoni, A. Deep brain stimulation: A review of the open neural engineering challenges. J. Neural Eng.

2020,17, 051002. [CrossRef] [PubMed]

41.

Pozzi, N.G.; Isaias, I.U. Adaptive deep brain stimulation: Retuning Parkinson’s disease. Handb. Clin. Neurol.

2022

,184, 273–284.

42.

Neumann, W.; Horn, A.; Ewert, S.; Huebl, J.; Brücke, C.; Slentz, C.; Schneider, G.; Kühn, A.A. A localized pallidal physiomarker

in cervical dystonia. Ann. Neurol. 2017,82, 912–924. [CrossRef] [PubMed]

43.

Piña-Fuentes, D.; van Dijk, J.M.C.; Drost, G.; van Zijl, J.C.; van Laar, T.; Tijssen, M.A.J.; Beudel, M. Direct comparison of oscillatory

activity in the motor system of Parkinson’s disease and dystonia: A review of the literature and meta-analysis. Clin. Neurophysiol.

2019,130, 917–924. [CrossRef] [PubMed]

44.

Zhu, G.-Y.; Zhang, R.-L.; Chen, Y.-C.; Liu, Y.-Y.; Liu, D.-F.; Wang, S.-Y.; Jiang, Y.; Zhang, J.-G. Characteristics of globus pallidus

internus local field potentials in generalized dystonia patients with TWNK mutation. Clin. Neurophysiol.

2020

,131, 1453–1461.

[CrossRef]

45.

Piña-Fuentes, D.; Beudel, M.; Van Zijl, J.C.; Van Egmond, M.E.; Oterdoom, D.L.M.; Van Dijk, J.M.C.; Tijssen, M.A.J. Low-frequency

oscillation suppression in dystonia: Implications for adaptive deep brain stimulation. Park. Relat. Disord.

2020

,79, 105–109.

[CrossRef]

46.

Thenaisie, Y.; Palmisano, C.; Canessa, A.; Keulen, B.J.; Capetian, P.; Jiménez, M.C.; Bally, J.F.; Manferlotti, E.; Beccaria, L.; Zutt,

R.; et al. Towards adaptive deep brain stimulation: Clinical and technical notes on a novel commercial device for chronic brain

sensing. J. Neural Eng. 2021,18, 042002. [CrossRef] [PubMed]

47.

Sirica, D.; Hewitt, A.L.; Tarolli, C.G.; Weber, M.T.; Zimmerman, C.; Santiago, A.; Wensel, A.; Mink, J.W.; Lizárraga, K.J.

Neurophysiological biomarkers to optimize deep brain stimulation in movement disorders. Neurodegener. Dis. Manag.

2021

,11,

315–328. [CrossRef] [PubMed]

48.

Hess, C.W.; Pullman, S.L. Tremor: Clinical Phenomenology and Assessment Techniques. Tremor Hyperkinetic Mov.

2012

,2, 02.

[CrossRef]

49.

Lalli, S.; Albanese, A. Dystonic Tremor. In Mechanisms and Emerging Therapies in Tremor Disorders; Grimaldi, G., Manto, M., Eds.;

Springer: New York, NY, USA, 2013; pp. 203–218. ISBN 978-1-4614-4026-0.

50.

Volkmann, J.; Mueller, J.; Deuschl, G.; Kühn, A.A.; Krauss, J.K.; Poewe, W.; Timmermann, L.; Falk, D.; Kupsch, A.; Kivi, A.; et al.

Pallidal neurostimulation in patients with medication-refractory cervical dystonia: A randomised, sham-controlled trial. Lancet

Neurol 2014,13, 875–884. [CrossRef] [PubMed]

51.

Jost, W.H.; Hefter, H.; Stenner, A.; Reichel, G. Rating scales for cervical dystonia: A critical evaluation of tools for outcome

assessment of botulinum toxin therapy. J. Neural Transm. Vienna Austria 1996 2013,120, 487–496. [CrossRef]

52.

Boyce, M.J.; Canning, C.G.; Mahant, N.; Morris, J.; Latimer, J.; Fung, V.S.C. The Toronto Western Spasmodic Torticollis Rating

Scale: Reliability in neurologists and physiotherapists. Park. Relat. Disord. 2012,18, 635–637. [CrossRef] [PubMed]

53.

Masuhr, F.; Wissel, J.; Müller, J.; Scholz, U.; Poewe, W. Quantification of sensory trick impact on tremor amplitude and frequency

in 60 patients with head tremor. Mov. Disord. 2001,15, 960–964. [CrossRef]

54.

Deuschl, G.; Heinen, F.; Kleedorfer, B.; Wagner, M.; Lücking, C.H.; Poewe, W. Clinical and polymyographic investigation of

spasmodic torticollis. J. Neurol. 1992,239, 9–15. [CrossRef] [PubMed]

55.

Canessa, A.; Palmisano, C.; Isaias, I.U.; Mazzoni, A. Gait-related frequency modulation of beta oscillatory activity in the

subthalamic nucleus of parkinsonian patients. Brain Stimul. 2020,13, 1743–1752. [CrossRef] [PubMed]

56.

Canessa, A.; Pozzi, N.G.; Arnulfo, G.; Brumberg, J.; Reich, M.M.; Pezzoli, G.; Ghilardi, M.F.; Matthies, C.; Steigerwald, F.;

Volkmann, J.; et al. Striatal Dopaminergic Innervation Regulates Subthalamic Beta-Oscillations and Cortical-Subcortical Coupling

during Movements: Preliminary Evidence in Subjects with Parkinson’s Disease. Front. Hum. Neurosci. 2016,10, 611. [CrossRef]

57.

Arnulfo, G.; Pozzi, N.G.; Palmisano, C.; Leporini, A.; Canessa, A.; Brumberg, J.; Pezzoli, G.; Matthies, C.; Volkmann, J.; Isaias, I.U.

Phase matters: A role for the subthalamic network during gait. PLoS ONE 2018,13, e0198691. [CrossRef] [PubMed]

58.

Neumann, W.-J.; Memarian Sorkhabi, M.; Benjaber, M.; Feldmann, L.K.; Saryyeva, A.; Krauss, J.K.; Contarino, M.F.; Sieger, T.;

Jech, R.; Tinkhauser, G.; et al. The sensitivity of ECG contamination to surgical implantation site in brain computer interfaces.

Brain Stimul. 2021,14, 1301–1306. [CrossRef] [PubMed]

Bioengineering 2023,10, 476 15 of 15

59.

Donoghue, T.; Haller, M.; Peterson, E.J.; Varma, P.; Sebastian, P.; Gao, R.; Noto, T.; Lara, A.H.; Wallis, J.D.; Knight, R.T.; et al.

Parameterizing Neural Power Spectra into Periodic and Aperiodic Components. Nat. Neurosci. 2020,23, 1655–1665. [CrossRef]

60.

Barow, E.; Neumann, W.J.; Brücke, C.; Huebl, J.; Horn, A.; Brown, P.; Krauss, J.K.; Schneider, G.H.; Kühn, A.A. Deep brain

stimulation suppresses pallidal low frequency activity in patients with phasic dystonic movements. Brain

2014

,137, 3012–3024.

[CrossRef]

61.

Sharott, A.; Grosse, P.; Kühn, A.A.; Salih, F.; Engel, A.K.; Kupsch, A.; Schneider, G.H.; Krauss, J.K.; Brown, P. Is the synchronization

between pallidal and muscle activity in primary dystonia due to peripheral afferance or a motor drive? Brain

2008

,131, 473–484.

[CrossRef]

62.

Haufe, S.; Isaias, I.; Pellegrini, F.; Palmisano, C. Gait event prediction from surface electromyography in parkinsonian patients.

medRxiv 2023. [CrossRef]

63.

Schnitzler, A.; Gross, J. Normal and pathological oscillatory communication in the brain. Nat. Rev. Neurosci.

2005

,6, 285–296.

[CrossRef] [PubMed]

64.

Shaikh, A.G.; Jinnah, H.A.; Tripp, R.M.; Optican, L.M.; Ramat, S.; Lenz, F.A.; Zee, D.S. Irregularity distinguishes limb tremor in

cervical dystonia from essential tremor. J. Neurol. Neurosurg. Psychiatry 2008,79, 187–189. [CrossRef] [PubMed]

65.

Chen, C.C.; Kühn, A.A.; Hoffmann, K.T.; Kupsch, A.; Schneider, G.H.; Trottenberg, T.; Krauss, J.K.; Wöhrle, J.C.; Bardinet, E.;

Yelnik, J.; et al. Oscillatory pallidal local field potential activity correlates with involuntary EMG in dystonia. Neurology

2006

,66,

418–420. [CrossRef]

66.

Averna, A.; Arlotti, M.; Rosa, M.; Chabardès, S.; Seigneuret, E.; Priori, A.; Moro, E.; Meoni, S. Pallidal and Cortical Oscillations in

Freely Moving Patients With Dystonia. Neuromodulation Technol. Neural Interface

2022

, S1094715922000162. [CrossRef] [PubMed]

67.

Schramm, A.; Reiners, K.; Naumann, M. Complex mechanisms of sensory tricks in cervical dystonia. Mov. Disord.

2004

,19,

452–458. [CrossRef] [PubMed]

68.

Steeves, T.D.; Day, L.; Dykeman, J.; Jette, N.; Pringsheim, T. The prevalence of primary dystonia: A systematic review and

meta-analysis. Mov. Disord. Off. J. Mov. Disord. Soc. 2012,27, 1789–1796. [CrossRef]

69.

Ortiz, R.M.; Scheperjans, F.; Pekkonen, E. Deep brain stimulation for dystonia in Finland during 2007–2016. BMC Neurol.

2019

,19, 137.

[CrossRef]

70.

Scheller, U.; Lofredi, R.; Wijk, B.C.M.; Saryyeva, A.; Krauss, J.K.; Schneider, G.; Kroneberg, D.; Krause, P.; Neumann, W.; Kühn,

A.A. Pallidal low-frequency activity in dystonia after cessation of long-term deep brain stimulation. Mov. Disord.

2019

,34,

1734–1739. [CrossRef]

71.

Popov, T.; Tröndle, M.; Baranczuk, Z.; Pfeiffer, C.; Haufe, S.; Langer, N. Test–retest reliability of resting-state EEG in young and

older adults. Psychophysiology 2023, e14268. [CrossRef]

72.

Pellegrini, F.; Delorme, A.; Nikulin, V.; Haufe, S. Identifying best practices for detecting inter-regional functional connectivity

from EEG. bioRxiv 2022. [CrossRef]

Disclaimer/Publisher’s Note:

The statements, opinions and data contained in all publications are solely those of the individual

author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to

people or property resulting from any ideas, methods, instructions or products referred to in the content.