Identification of EEG Dynamics During Freezing of Gait and Voluntary Stopping in Patients with Parkinson's Disease
Zehong Cao, Alka Rachel John, Hsiang-Ting Chen, Kaylena Ehgoetz Martens, Matthew Georgiades, Moran Gilat, Hung T. Nguyen, Simon J. G. Lewis, Chin-Teng Lin
JJOURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 1
Identification of EEG Dynamics during Freezing ofGait and Voluntary Stopping in Patients withParkinson’s Disease
Zehong Cao † , Member, IEEE,
Alka Rachel John † , Hsiang-Ting Chen, Senior Member, IEEE,
Kaylena EhgoetzMartens, Matthew Georgiades, Moran Gilat, Hung T. Nguyen,
Senior Member, IEEE,
Simon J. G. Lewis *,and Chin-Teng Lin *,
Fellow, IEEE
Abstract —Mobility is severely impacted in patients withParkinson’s disease (PD), especially when they experience invol-untary stopping from the freezing of gait (FOG). Understandingthe neurophysiological difference between “voluntary stopping”and “involuntary stopping” caused by FOG is vital for thedetection and potential intervention of FOG in the daily livesof patients. This study characterised the electroencephalographic(EEG) signature associated with FOG in contrast to voluntarystopping. The protocol consisted of a timed up-and-go (TUG)task and an additional TUG task with a voluntary stoppingcomponent, where participants reacted to verbal “stop” and“walk” instructions by voluntarily stopping or walking. Event-related spectral perturbation (ERSP) analysis was used to studythe dynamics of the EEG spectra induced by different walkingphases, which included normal walking, voluntary stopping andepisodes of involuntary stopping (FOG), as well as the transitionwindows between normal walking and voluntary stopping orFOG. These results demonstrate for the first time that the EEGsignal during the transition from walking to voluntary stopping isdistinguishable from that of the transition to involuntary stoppingcaused by FOG. The EEG signature of voluntary stopping ex-hibits a significantly decreased power spectrum compared to thatof FOG episodes, with distinctly different patterns in the deltaand low-beta power in the central area. These findings suggest
Zehong Cao was with Australian Artificial Intelligence Institute, Faculty ofEngineering and Information Technology, University of Technology Sydney,NSW, Australia. He is now with the School of Information and Communica-tion Technology, University of Tasmania, TAS, Australia.Alka Rachel John is with the Australian Artificial Intelligence Institute,Faculty of Engineering and Information Technology, University of TechnologySydney, NSW, Australia.Hsiang-Ting Chen was with Australian Artificial Intelligence Institute,Faculty of Engineering and Information Technology, University of TechnologySydney, NSW, Australia. He is now with the School of Computer Science,University of Adelaide, SA, Australia.Kaylena Ehgoetz Martens was with the Parkinson’s Disease ResearchClinic, Brain and Mind Centre, University of Sydney, NSW, Australia. Sheis now with the Department of Kinesiology, University of Waterloo, Ontario,Canada.Matthew Georgiades is with the Parkinson’s Disease Research Clinic, Brainand Mind Centre, University of Sydney, NSW, Australia.Moran Gilat was with the Parkinson’s Disease Research Clinic, Brainand Mind Centre, University of Sydney, NSW, Australia. He is now withDepartment of Rehabilitation Sciences, KU Leuven, Leuven, Belgium.Hung T. Nguyen is with the Faculty of Science, Engineering and Technol-ogy, Swinburne University of Technology, VIC, Australia.Simon J. G. Lewis is with the Parkinson’s Disease Research Clinic,Brain and Mind Centre, University of Sydney, NSW, Australia. (* E-mail:[email protected]).Chin-Teng Lin is with the Australian Artificial Intelligence Institute, Fac-ulty of Engineering and Information Technology, University of TechnologySydney, NSW, Australia. (* E-mail: [email protected]). † Equal contribution. * Corresponding authors.Manuscript received xx, 2021. the possibility of a practical EEG-based treatment strategy thatcan accurately predict FOG episodes, excluding the potentialconfound of voluntary stopping.
Index Terms —EEG Dynamics, Freezing of Gait, Parkinson’sDisease, Voluntary Stopping
I. I
NTRODUCTION F REEZING of gait (FOG) is a devastating symptom ofParkinson’s disease (PD) in which patients suddenly feelas though their feet have become “stuck to the ground” [1].Approximately 80% of patients with severe PD are affected byFOG episodes, which often precipitate falls, leading to a highmorbidity and the urgent need for nursing home placement[2].The current pathophysiology underlying the freezing phe-nomenon is not well-understood [3]. Some previous studies,such as functional magnetic resonance imaging (MRI)-basedwork, have attempted to understand the pathophysiology ofFOG episodes and have identified some distinct patternsassociated with freezing [4], [5]. Additionally, several elec-troencephalography (EEG)-based studies have also attemptedto establish the neurophysiological correlate of FOG in PD.Increased beta power was previously observed in the sub-thalamic nucleus (STN) of patients with FOG [6] and morerecent work has shown a temporal relationship with an increasein pathological beta and theta rhythms in the STN recordedduring the performance of a virtual reality gait paradigm [18].Previously the brain dynamics associated with FOGepisodes during turning measured by using ambulatory EEGhas revealed significant changes in the high-beta and thetapower spectral densities across the occipital and parietal areasduring FOG episodes with turning [7]. In addition, EEGdynamics have demonstrated great potential in identifying theonset of freezing in patients with PD [8]. Indeed, EEG featureshave been suggested to be useful in predicting the transitionfrom normal walking to freezing by using a 5-s time windowbefore the episode [7], [9].The advances in EEG have made it an efficient tool forunderstanding not only FOG episodes but also other ordi-nary motor tasks related to movements, including walking.For example, two studies [10], [11] have suggested specificroles for EEG activity within particular frequency bands inthe completion of ongoing motor tasks. Specifically, this a r X i v : . [ phy s i c s . m e d - ph ] F e b OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 2
EEG activity includes beta activity during motor preparation,gamma activity during motor commission and gating [12],[13], and theta activity during the processing of conflict-relatedsignals [14], [15]. Additionally, event-related EEG potentials,particularly the EEG signals occurring before the initiation ofan action, were found to be useful in identifying the intentionto move [16], [17].Since gait dynamics alone cannot accurately predict freezingepisodes or independently distinguish freezing from voluntarystopping, we hypothesise that EEG could prove useful for thereliable detection of FOG. Anticipating that the brain dynamicsduring the transition to freezing can be confidently discernedfrom those during the transition to voluntary stopping, “real-time” EEG may offer a novel therapeutic intervention forthe prediction and alleviation of freezing episodes. Recently,researchers have demonstrated that pathological subthalamicnucleus activity associated with bouts of freezing was dis-cernible from that of voluntary stopping when assessed usinga virtual “gait” paradigm while lying down and navigatingthe virtual environment by using a set of foot pedals [18].This research provides positive support, but little work hasbeen done to identify how brain dynamics during freezing aredistinct from those during voluntary stopping while walkingusing ambulatory EEG.In our study, to discern the neurophysiological differencesbetween freezing and voluntary stopping episodes, we con-ducted the experiment building upon the timed up-and-go(TUG) protocol [19]. The TUG protocol consists of a sequenceof sit-to-stand, walking, turning, and stand-to-sit tasks, each ofwhich can be affected by freezing, especially when performedas a sequence [20]. In our study, we further added a conditionof “voluntary stopping”, where the participants reacted to theverbal instruction to voluntarily “stop” while performing theTUG task. We tracked the EEG dynamics of patients withPD before and during FOG episodes and during voluntarystopping. Contrasting the signature of freezing from that ofvoluntary stopping could significantly help pave the way to-wards more effective therapeutics that accurately predict FOGevents while excluding potential false positives associated withvoluntary stopping.We hope that our findings will provide a potential avenuefor therapeutic prediction and alleviation of freezing episodesin patients with PD and promote exploration of voluntarystopping in the gait cycle by characterising the EEG signatures,as the identification of these movement intentions could bevery useful in motor rehabilitation processes.II. MATERIALS AND METHODS
A. Subjects
In this study, seventeen (17) patients from the Parkinson’sDisease Research Clinic at the Brain and Mind Centre, Uni-versity of Sydney, were identified by using the score for item3 of the self-reported FOG Questionnaire (FOGQ), whichwas further confirmed by specialist review. Consistent withour previous studies [7], [21], all patients satisfied the UKParkinson’s Disease Society Brain Bank (UKPDSBB) criteria,had a Mini-Mental State Examination (MMSE) score ≥
24 and were deemed unlikely to have dementia or major depressionaccording to DSM-IV criteria by the judgement of neurologistSimon J. G. Lewis. This study was approved by The Universityof Sydney Human Research and Ethics Committee, and writteninformed consent was obtained from all the patients.
B. Experimental design
All patients underwent a structured series of video-recordedTUG tasks while in their practically defined off state (havingwithdrawn from PD medication overnight for a minimum of12 hours). All TUG tasks started in a seated position on achair, from which patients walked along the centre of a largeopen corridor. A target box (0.6 m x 0.6 m) located six metres(6 m) from the chair was marked on the floor with white tape,and turning movements were performed in this box. A TUGtask involved a 180° or 540° turn within the box and a returnto the starting chair. It was performed with counterbalancedturns to the patient’s left or right side. Furthermore, the videoswere independently reviewed, and FOG episodes were scoredby two experienced clinical researchers.Our experimental paradigm consisted of two types of TUGtasks: a standard TUG task to trigger FOG and a TUG taskwith a voluntary stopping component in which “stop” or“walk” verbal instructions were provided by the investigatorsto guide voluntary walking or stopping. For the standard TUGtask (Fig. 1A), patients performed a series of timed TUGtasks on a standardised course to evoke FOG events [20].Each FOG trial consisted of three epochs. Specifically, wedefined a “normal walking” epoch as a 2-s period in whichthe patient with PD walked normally with no cessation. Incontrast, a “transition” epoch was the 2-s period before afreezing episode. Finally, a “FOG” episode was defined asan involuntary stop and was identified as the 2-s period fromwhen a patient experienced an unwanted cessation in theirnormal stride.For the TUG with a voluntary stopping component withverbal instructions to “stop” and “walk” (Fig. 1B), a targetbox (0.6 m x 0.6 m) more than 10 m from the chair wasmarked on the floor with white tape, and turning movementswere performed in this box. To promote voluntary stoppingin the patients, the observer said the word “stop”, and thepatients were required to stop immediately. In the next 5-10s, the observer said the word “walk”, and the subjects wereasked to start walking immediately. These voluntary stoppingtrials also consisted of three epochs. More specifically, wedefined a “normal walking” epoch as a 2-s period in which thepatient was walking normally with no cessation. A “transition”epoch was identified as the interval between giving a “stop”instruction and when the patient physically stopped walking.A “voluntary stopping” episode was identified as the 2-s epochfrom when the patient physically stopped walking.
C. Data recording and analysis
Each patient wore a wearable BioSemi Active-Two systemwith 32 Ag-AgCl electrodes to record the EEG signals. The 32channel electrodes (Fp1, Fp2, F7, F3, Fz, F4, F8, FT7, FC3,FCz, FC4, FT8, T7, C3, Cz, C4, T8, TP7, CP3, CPZ, CP4,
OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 3
Camera RecordingNormal WalkingObserverStartEnd Freezing of Gait Normal Walking Transition TurningObserverStartEnd Normal Walking StopWalk TransitionVoluntary Stop ObserverNormal Walking Turning AB Fig. 1. Experimental protocol of the TUG tasks. (A) The standard TUGtask. (B) The TUG task with a voluntary stopping component with verbalinstructions to “stop” and “walk” for assessing voluntary stopping.
TP8, P7, P3, Pz, P4, P8, O1, Oz, O2, A1, and A2) used theInternational 10–20 system locations to maintain a standardclinical setting, and the A1 and A2 electrodes were used asthe reference channels. Before calibrating the electrodes, thepatient’s skin under the reference electrodes was washed with70% isopropyl alcohol. The EEG signals were recorded at asample rate of 500 Hz with 16-bit quantisation.
D. EEG pre-processing
The EEG data were analysed with the EEGLAB toolbox[23] in MATLAB R2016a. As shown in Fig. 2, the raw EEGsignals were initially pre-processed by a 1-30-Hz bandpassfilter to remove line noise, low-frequency noise and high-frequency components. Removing artefacts, such as eye move-ment and muscle activity, is crucial, as they can adverselyaffect further processing steps. For artefact rejection, eyemovement contaminants in the EEG signals were identifiedby visual inspection and manually removed.Additionally, artefacts were removed by the AutomaticArtifact Removal (AAR) [24], [25] plug-in for EEGLAB,which integrates many state-of-the-art methods for automaticcorrection of ocular and muscular artefacts in the EEG signals.This toolbox uses blind source separation to decompose theEEG data into several spatial components before automaticallyremoving artefact-related components. It then reconstructs theEEG signal by using the non-artefactual components. After-wards, the FOG and voluntary stopping trials were extractedfor further analyses, and a time-warping method was used tomeasure latencies.In this study, a total of 178 FOG episodes and 54 voluntarystopping episodes were analysed. These 178 FOG episodeswere extracted from both TUG tasks, of which 74 FOGepisodes occurred during turning, while 104 occurred duringthe normal stride.
Bandpass filter (1 – 30 Hz) Manual Noise Removal by Visual Inspection Automatic Artefact RemovalExtract FOG and Voluntary Stop Epochs
ERSPRaw EEG Pre-Processing
Fig. 2. Diagram of the EEG pre-processing and analysis scheme.
E. EEG ERSP analysis
The FOG and voluntary stopping epochs were extractedfrom the continuous EEG signals, and each epoch containedthe sampled EEG data from -2000 ms to 4000 ms withthe stimulus onset at 0 ms. To investigate brain dynamicsduring the FOG and voluntary stopping episodes and thesubsequent motor responses, each epoch was separately trans-formed into a time-frequency representation by using an event-related spectral perturbation (ERSP) routine [26], which is atime-frequency analysis to transform time-domain signals intothe spectral-temporal domain to characterise the event-relatedfrequency changes by using the short-time Fourier transform(STFT).To compute the ERSP for each trial, spectra prior to theevent onset are considered the baseline spectra for that trial.The mean of the baseline spectra (in dB) is subtracted fromthe spectral power after event onset to visualise the spectral“perturbation” from the baseline. The mean normal walkinglog power spectrum (in dB) of the optimal epochs, which wasused as the reference (baseline) value, was subtracted fromeach estimated spectrum. The ERSP is derived by computingthe power spectrum over a sliding latency window and thencomputing the average across all the trials. Therefore, for ntrials, if F k ( f, t ) is the spectral estimate of trial k at frequency f and time t , then the ERSP is given by ERSP ( f, t ) = 1 n ∗ (cid:88) | F k ( f, t ) | . (1)In EEGLAB, F k ( f, t ) is calculated by using the STFT [27].The relative power at a particular frequency and latency isindicated by the colour of the corresponding pixel in the ERSPimage.In this study, the mean power spectra, including the delta (1-3.5 Hz), theta (4-7.5 Hz), alpha (8-12.5 Hz), and beta (13-30Hz) bands, were vertically stacked from normal walking andthe transition to FOG or voluntary stopping. We consideredfour important electrodes (Fz – supplementary motor area,Cz – primary motor area, P4 – navigational movement area,and O1 – primary visual receiving area), as well as the left-hemisphere (F3, FC3, C3, CP3, and P3) and right-hemisphere(F4, FC4, C4, CP4, and P4) brain areas, to explore event-related band power dynamics that accompanied the transitionof power changes before and after the FOG and voluntarystopping episodes (the x-axis is time, and the y-axis is thepower spectrum). OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 4 F re qu e n c y ( H z ) -2 0 FzCzP4O1 S1 S3
Normal Walking Transition Freezing of Gait F re qu e n c y ( H z ) F re qu e n c y ( H z ) F re qu e n c y ( H z ) Fig. 3. Event-related band power dynamics during FOG episodes whileperforming the TUG task. S1 denotes the timing from normal walking tothe transition period, and S3 denotes the timing from the transition period toFOG. The ERSP changes from normal walking ( < -2 s), transition (-2 s to 0s), to FOG ( > F. Statistical analysis
The ERSPs for normal walking, the transition to FOGand the transition to voluntary stopping were compared usingpaired t-tests. Independent t-tests were performed to comparethe ERSPs between the FOG and voluntary stopping episodes.The significance level was set at α < . . The statisticalanalysis was performed in MATLAB (R2016a).III. R ESULTS
A. Event-related band power dynamics during FOG in theTUG trials
The two-dimensional images in Fig. 3 show the ERSPsof the frontal, central, parietal, and occipital channels forpatients with PD accompanied by FOG. As shown in Fig. 3,the mean EEG power spectra before and after the onset offreezing at the Fz, Cz, P4 and O1 electrodes are presented. Inparticular, the ERSP at the Cz electrode showed a decreasedEEG power spectrum during the transition period relative tonormal walking and an increased EEG power spectrum in thetransition period relative to FOG episodes.Furthermore, we compared the different periods during FOGepisodes, as shown in Fig. 4. Compared to those during normalwalking, the EEG alpha power band and a portion of the beta F re qu e n c y ( H z ) TransitionFreezing of Gait *
FzCz P4O1
Normal WalkingTransition *
20 0 2 dBTime (second) Time (second)0000 F re qu e n c y ( H z ) F re qu e n c y ( H z ) F re qu e n c y ( H z )
20 0 2Time (second) Time (second)
TransitionFreezing of Gait *Normal WalkingTransition * * Significant Differences
Fig. 4. Comparisons of the ERSP differences between the normal walkingand the transition period (left column) and the transition period and the FOGepisode (right column) at four important electrodes (Fz, Cz, P4 and O1). power band in the transition period were significantly elevatedat the Cz and O1 electrodes (p < < B. Event-related band power dynamics between voluntarystopping and FOG episodes during the TUG trials
The two-dimensional images in Fig. 5 plot the ERSPs ofthe voluntary stopping (Fig. 5A) and FOG episodes duringthe TUG task (Fig. 5B) at the frontal, central, parietal, andoccipital channels for patients with PD accompanied by FOG.In terms of the voluntary stopping episodes during the TUGtask (Fig. 5), a globally increased EEG power spectrum inthe delta, theta, alpha and beta bands was observed in thetransition period relative to normal walking. Furthermore, apartially decreased EEG power spectrum was observed duringvoluntary stopping relative to the transition period, and inparticular, significantly reduced delta and beta power bandswere observed during voluntary stopping at the Cz electrode.We compared the ERSP differences between the FOG andvoluntary stopping episodes while performing the TUG taskduring the normal walking, transition, and freezing/voluntarystopping periods at the four important electrodes (Fz, Cz,P4, and O1), as shown in Fig. 5C. Our results showedsignificantly enhanced delta, theta, alpha and beta ERSPs inthe voluntary stopping episodes relative to the FOG episodesduring the normal walking and transition periods (p < < OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 5
A B C F re qu e n c y ( H z ) F re qu e n c y ( H z ) F re qu e n c y ( H z ) F re qu e n c y ( H z ) S1 D1 D2 D3
FzCzP4O1
S3S1 S2 -2 0
Time (second)Normal Walking Transition Voluntary Stop Normal Walking Transition Freezing of Gait -2 0
Time (second) Time (second)Normal Walking * Transition * Voluntary StopFreezing of Gait * * Significant Differences
22 5-50dB
Fig. 5. Event-related band power dynamics between the voluntary stopping and FOG episodes during the TUG task. (A) The ERSP changes during thevoluntary stopping episodes in the normal walking ( < -2 s), transition (-2 s to 0 s), and voluntary stopping ( > < -2 s), transition (-2 s to 0 s), and FOG ( > To make the number of FOG trials and voluntary stop trialscomparable, we studied the ERSPs of voluntary stopping andFOG trials after removing all the FOG episodes that occurredduring turning. There were 119 FOG trials remaining after allthe FOG trials at turning were removed. Fig. 6 showed theERSPs of voluntary stopping (Fig. 6A) and FOG (Fig. 6B)trials during the TUG task without the FOG during turningat the frontal, central, parietal, and occipital channels forthe PD patients. The comparison between voluntary stoppingtrials and FOG trials without the turning FOG during normalwalking, transition, and freezing of gait or voluntary stopperiods at the Fz, Cz, P4 and O1 electrodes were shown inFig. 6C. The observed results were similar to the ERSPresults in which FOG trials at turning were also included.These results also showed an enhanced delta, theta, alphaand beta ERSP in voluntary stopping trials when comparedto the ERSP in FOG trials during the normal walking andtransition periods (p < < C. Brain hemisphere power dynamics between the voluntarystopping and FOG episodes
Fig. 7 displays the results of the average EEG power spectrain the left and right hemispheres for the voluntary stoppingand FOG episodes during the TUG task. For the voluntarystopping episodes during the TUG task, as shown in Figs. 7A,a partial decrease in the delta power and an increase in thebeta power were observed in the left and right hemispheres inthe transition period relative to normal walking. Furthermore,the ERSP maintained decreased delta and low-beta power and
OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 6
A B C F re qu e n c y ( H z ) F re qu e n c y ( H z ) F re qu e n c y ( H z ) F re qu e n c y ( H z ) S1 D1 D2 D3
FzCzP4O1
S3S1 S2 -2 0
Time (second)Normal Walking Transition Voluntary Stop Normal Walking Transition Freezing of Gait (without turning) -2 0
Time (second) Time (second)Normal Walking * Transition * Voluntary StopFreezing of Gait * (without turning) * Significant Differences
22 5-50dB
Fig. 6. Event-related band power dynamics between the voluntary stopping and FOG trials during the TUG task in which all the FOG episodes at turningwere removed. (A) The ERSP changes during the voluntary stopping episodes in the normal walking ( < -2 s), transition (-2 s to 0 s), and voluntary stopping( > < -2 s), transition (-2 s to 0 s), and FOG ( > increased high-beta power during a voluntary stopping in theleft and right hemispheres. For FOG episodes during the TUGtask, as shown in Fig. 7B, no significant change in the ERSPwas noted except for a slight increase in the alpha powerduring freezing.The differences in the brain hemisphere power dynamicsbetween the voluntary stopping and FOG episodes during theTUG tasks are shown in Fig. 7C. Specifically, our resultsdemonstrated that the EEG power spectrum during the tran-sition was significantly higher in the beta band and slightlylower in the delta band during the voluntary stopping episodesthan during the FOG episodes (p < < < OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 7 F re qu e n c y ( H z ) F re qu e n c y ( H z ) Time (second)Normal Walking Transition Voluntary Stop Normal Walking Transition Freezing of Gait -2 0
Time (second) Time (second)Normal Walking * Transition * Voluntary StopFreezing of Gait * * Significant Differences
22 5-50dB
Left HemisphereRight Hemisphere
S1 S2 S1 S3 D1 D2 D3
A B C
Fig. 7. Brain hemisphere power dynamics during voluntary stopping and FOG episodes in the TUG task. (A) The ERSP changes during the voluntary stoppingepisodes in the normal walking ( < -2 s), transition (-2 s to 0 s), and voluntary stopping ( > < -2 s), transition (-2 s to 0 s), and FOG ( > increase in power in the central brain area before the onset offreezing. Another key observation was that the ERSP changesfor voluntary stopping episodes showed an elevated high-betapower at the Fz, Cz, P4, and O1 electrodes. These changeswere verified in each voluntary stopping episode, and everyepisode showed a significantly increased high-beta power atthese four electrodes. In addition, decreases in the EEG powerwere observed in the frontal, central, parietal and occipitalareas during the voluntary stopping compared with the FOGepisodes. A. FOG episodes related to EEG power dynamics in thecentral area
Tracking and identifying FOG episodes by characterisingtheir EEG signatures may provide a potential method fortherapeutic prediction and alleviation of freezing episodes inpatients with PD. Our study presented a novel finding that thetransition from normal walking to freezing was associated witha significant increase in the EEG power in the central brainarea. The results also showed an increase in delta and thetaactivity during the transition to freezing in the central area.The enhanced EEG power may reflect the brain’s inhibitoryprocesses activated by freezing episodes, mainly in the centralarea.Furthermore, clinical studies have shown that freezing be-haviour in PD is associated with a paroxysmal increase in thetaoscillations (5 - 7 Hz), known as “trembling in place” [28].The changes observed in the EEG power spectrum in the theta band during freezing may be due to mechanical oscillationstransmitted to the scalp electrodes. It is worth noting thatthe increase in the theta power was unlikely due to globalmotor interference, as we observed a significant increase inthe theta power only in the central area. Perhaps a moreacceptable explanation is that a dysfunctional neuronal circuitin subcortical brain structures drives the creation of thetaoscillations in patients with PD [29]. A similar mechanismwas recently found to underlie the oscillations observed duringFOG [30].The results also showed that the beta activity in the centralarea was enhanced during this transition period. The changein the beta power might be associated with the facilitation ofpostural activities, including a tonic holding contraction andinhibition of voluntary movement [31]. Moreover, beta sup-pression is critical for the facilitation of continuous movementsequences [32], suggesting that high-beta activity may inter-face with anticipatory postural adjustments in preparation forstepping, which can facilitate excessive postural contraction ofthe lower limbs associated with FOG episodes [33].
B. Acute prediction and detection of FOG episodes
In previous studies, the transition period prior to FOG wasdefined as the time window of 5 to 1 s before the occurrenceof freezing [7], [21]. However, this study analysed a shortertransition period of 2 s before freezing, which might have moreclinical utility, as gait parameters usually change prior to freez-ing within a short time window. Compared to those of previous
OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 8 studies, the EEG power in this study showed similar changesat the Cz electrode but different changes at the O1 electrode inthe transition period from normal walking, suggesting that theEEG in the occipital area may be susceptible to the preparationof freezing. In this study, considering a 2-s transition period toa FOG event, we observed an increase in the beta power in thecentral region of the brain, as with the 5-s transition period,compared to normal walking. However, we also observedan increase in the alpha power in the central region, whilean increase in the theta power was observed during the 5-s transition period compared to the normal walking period.When the 5-s transition period was compared to the normalwalking period, increases in the alpha and beta power wereobserved in the parietal area; however, only a partial increasein the beta power was observed when the 2-s transition periodwas compared to the normal walking period. While an increasein the beta power in the occipital area was observed only whenthe 5-s transition period was compared to normal walking, weobserved increases in both the alpha and beta power at theoccipital region when comparing the 2-s transition period tonormal walking.
C. EEG power dynamics between involuntary and voluntarystopping
Currently, few studies [7], [9] have highlighted the dis-tinctions of different motor inhibition processes, and thus,investigating brain dynamics under conditions of involuntarystopping (freezing) and voluntary stopping is insightful. Ourresults showed that the voluntary stopping period had a signifi-cantly decreased EEG power spectrum compared to that in thefreezing period, and a distinctly different pattern in the deltaand low-beta power was particularly observed in the centralarea. These suppressions may reflect motion preparation andexecution [34] as well as EEG oscillatory patterns, reflectingthe process of inhibitory control in the brain [35], [36], whichcontrols the cessation of foot movement when patients hear“stop” cues during TUG trials.In fact, the voluntary stopping task in our TUG task withthe voluntary stopping component is similar to the stop-signalparadigm, wherein a participant engaged in a task stops theirprimary task when presented with a signal to stop [11]. Aswith the stop-signal paradigm, voluntary stopping during theTUG task may also involve initiation and inhibition. Thetransition period in the voluntary stopping episode is similarto response inhibition, where the subject is required to stopwalking after initiating movements quickly. In terms of thetransition period, our results showed increased EEG powerin the voluntary stopping episodes compared to the FOGepisodes. This strengthening of the EEG power indicatesan effective preparation stage for response inhibition [29],whereas a low EEG power spectrum during freezing mayinvolve incomplete or ineffective inhibitory control over somemovement options. V. C
ONCLUSION
This study investigated the brain dynamics of FOG andvoluntary stopping episodes using EEG data collected from
TABLE IEEG
DYNAMICS DURING THE TRANSITION TO
FOGEEG Pattern Frequency Observation at Brain CorticesCentral Parietal OccipitalPower spectral δ N/A N/A N/Adensity (PSD) θ ↑ N/A N/Awith a 5-s α N/A ↓ N/Atransition [7] a β ↑ ↑ ↑ ERSP with a δ N/A N/A N/A2-s θ N/A N/A N/Atransition (Our α ↑ N/A ↑ findings) b β ↑ ↑ ↑ a The transition period, defined as 5 s before a FOG episode.The extracted period of normal walking lasts 5 s. b The transition period, defined as 2 s before a FOG episode.The extracted period of normal walking lasts 2 s. ↑ ↓
The up and down arrows represent increased and decreasedEEG power in the transition period comparedwith the normal walking period, respectively. patients with PD accompanied by FOG. Our findings high-lighted that FOG episodes were associated with abnormal EEGdynamics and that voluntary stopping could be discriminatedfrom FOG episodes. Comparing the transition to freezingperiod to the freezing period itself, our findings show thatfreezing episodes are associated with significantly increasedtheta and alpha band power within the central and occipitalareas. Furthermore, the EEG power significantly decreasedduring the voluntary stopping period compared with the FOGperiod. Our results provide novel insights into the rapidtransition dynamics underlying the phenomenon of FOG andmay provide a potential means for the therapeutic predictionand alleviation of freezing episodes in susceptible patients.These findings are very useful for the development of futuretechnologies that predict FOG episodes, as they suggest thatvoluntary stopping will not activate false positives, allowingfor the accurate detection of freezing.R
EFERENCES[1] Giladi, N., R. Kao and S. Fahn, “Freezing phenomenon in patients withparkinsonian syndromes,” in
Movement disorders: official journal of theMovement Disorder Society, vol. 12, no. 3, 1997, pp. 302–305.[2] Aarsland, D., J. P. Larsen, E. Tandberg and K. Laake, “Predictors ofnursing home placement in Parkinson’s disease: a population-based,prospective study,” in
Journal of the American Geriatrics Society, vol.48, no. 8, 2000, pp. 938–942.[3] Weiss, D., A. Schoellmann, M. D. Fox, N. I. Bohnen, S. A. Factor, A.Nieuwboer, M. Hallett and S. J. G. Lewis, “Freezing of gait: understand-ing the complexity of an enigmatic phenomenon,” in
Brain, vol. 143, no.1, 2020, pp. 14–30.[4] Shine, J. M., P. B. Ward, S. L. Naismith, M. Pearson and S. J. G. Lewis,“Utilising functional MRI (fMRI) to explore the freezing phenomenon inParkinson’s disease,” in
Journal of Clinical Neuroscience, vol. 18, no. 6,2011, pp. 807–810.[5] Lenka, A., R. M. Naduthota, M. Jha, R. Panda, A. Prajapati, K. Jhun-jhunwala, J. Saini, R. Yadav, R. D. Bharath and P. K. Pal, “Freezingof gait in Parkinson’s disease is associated with altered functional brainconnectivity,” in
Parkinsonism and Related Disorders, vol. 24, 2016, pp.100–106.[6] Toledo, J. B., J. Lopez-Azcarate, D. Garcia-Garcia, J. Guridi, M. Valencia,J. Artieda, J. Obeso, M. Alegre and M. Rodriguez-Oroz, “High betaactivity in the subthalamic nucleus and freezing of gait in Parkinson’sdisease,” in
Neurobiology of Disease, vol. 64, 2014, pp. 60–65.[7] Handojoseno, A. M. A., J. M. Shine, T. N. Nguyen, Y. Tran, S. J. G.Lewis and H. T. Nguyen, “Analysis and Prediction of the Freezing of GaitUsing EEG Brain Dynamics,” in
Ieee Transactions on Neural Systems andRehabilitation Engineering, vol. 23, no. 5, 2015, pp. 887–896.
OURNAL OF L A TEX CLASS FILES, VOL. XX, NO. XX, 2021 9 [8] Handojoseno, AM Ardi, James M. Shine, Tuan N. Nguyen, Yvonne Tran,Simon JG Lewis, and Hung T. Nguyen, “The detection of Freezing ofGait in Parkinson’s disease patients using EEG signals based on Waveletdecomposition,” presented at the
Behavioural brain research, vol. 137, no. 1-2,2002, pp. 3–25.[11] Alegre, M., J. Lopez-Azcarate, I. Obeso, L. Wilkinson, M. C. Rodriguez-Oroz, M. Valencia, D. Garcia-Garcia, J. Guridi, J. Artieda and M.Jahanshahi, “The subthalamic nucleus is involved in successful inhibitionin the stop-signal task: a local field potential study in Parkinson’s disease,”in
Experimental neurology, vol. 239, 2013, pp. 1–12.[12] Kuhn, A. A., S. A. Brandt, A. Kupsch, T. Trottenberg, J. Brocke, K.Irlbacher, G. H. Schneider and B. U. Meyer, “Comparison of motor effectsfollowing subcortical electrical stimulation through electrodes in theglobus pallidus internus and cortical transcranial magnetic stimulation,”in
Exp Brain Res, vol. 155, no. 1, 2004, pp. 48–55.[13] Androulidakis, A. G., A. A. Kuhn, C. C. Chen, P. Blomstedt, F. Kempf,A. Kupsch, G. H. Schneider, L. Doyle, P. Dowsey-Limousin, M. I. Harizand P. Brown, “Dopaminergic therapy promotes lateralised motor activityin the subthalamic area in Parkinson’s disease,” in
Brain, vol. 130, no. 2,2007, pp. 457–468.[14] Fumagalli, M., G. Giannicola, M. Rosa, S. Marceglia, C. Lucchiari,S. Mrakic-Sposta, D. Servello, C. Pacchetti, M. Porta, M. Sassi, R.Zangaglia, A. Franzini, A. Albanese, L. Romito, S. Piacentini, S. Zago,G. Pravettoni, S. Barbieri and A. Priori, “Conflict-dependent dynamic ofsubthalamic nucleus oscillations during moral decisions,” in
Soc Neurosci, vol. 6, no. 3, 2011, pp. 243–256.[15] Cavanagh, J. F., C. M. Figueroa, M. X. Cohen and M. J. Frank, “Frontaltheta reflects uncertainty and unexpectedness during exploration andexploitation,” in
Cereb Cortex, vol. 22, no. 11, 2012, pp. 2575–2586.[16] Ib´a˜nez, J., J. I. Serrano, M. D. del Castillo, J. A. Gallego and E. Rocon,“Online detector of movement intention based on EEG—Application intremor patients,” in
Biomedical Signal Processing and Control, vol. 8,no. 6, 2013, pp. 822–829.[17] Planelles, D., E. Hortal, A. Costa, E. Ianez and J. M. Azorin, “Firststeps in the development of an EEG-based system to detect intention ofgait initiation,” presented at the
Systems Conference (SysCon),
Brain, vol. 142,no. 12, 2019, pp. 3906–3916.[19] Morris, S., M. E. Morris and R. Iansek, “Reliability of measurementsobtained with the Timed ”Up and Go” Test in people with Parkinsondisease,” in
Physical Therapy, vol. 81, no. 2, 2001, pp. 810–818.[20] Zampieri, C., A. Salarian, P. Carlson-Kuhta, K. Aminian, J. G. Nuttand F. B. Horak,“The instrumented timed up and go test: potentialoutcome measure for disease modifying therapies in Parkinson’s disease,”in
Journal of Neurology Neurosurgery and Psychiatry, vol. 81, no. 2,2010, pp. 171–176.[21] Shine, J. M., S. T. Moore, S. J. Bolitho, T. R. Morris, V. Dilda, S.L. Naismith and S. J. G. Lewis,“Assessing the utility of Freezing ofGait Questionnaires in Parkinson’s Disease,” in
Parkinsonism and RelatedDisorders, vol. 18, no. 1, 2012, pp. 25–29.[22] Ehgoetz Martens, K. A., F. Pieruccini-Faria and Q. J. Almeida,“Couldsensory mechanisms be a core factor that underlies freezing of gait inParkinson’s disease?,” in
PLoS One, vol. 8, no. 5, 2013.[23] Delorme, A. and S. Makeig,“EEGLAB: an open source toolbox foranalysis of single-trial EEG dynamics including independent componentanalysis,” in
Journal of neuroscience methods, vol. 134, no. 1, 2004, pp.9–21.[24] G´omez-Herrero, Germ´an, Wim De Clercq, Haroon Anwar, Olga Kara,Karen Egiazarian, Sabine Van Huffel, and Wim Van Paesschen, “Auto-matic removal of ocular artifacts in the EEG without an EOG referencechannel,” presented at the
Tampere University ofTechnology,
NeuroImage, vol. 22, no. 3, 2004, pp.1035–1045.[27] Shin, Hanjun, Sunwon Lee, Himchan Kim, Jaewoo Kang, and Koohy-oung Lee,“Extracting signals from noisy single-channel EEG stream forubiquitous healthcare applications,” in
Journal of Internet Technology, vol. 13, no. 1, 2012, pp. 85–94.[28] MSchaafsma, J., Y. Balash, T. Gurevich, A. Bartels, J. M. Hausdorff andN. Giladi,“Characterisation of freezing of gait subtypes and the responseof each to levodopa in Parkinson’s disease,” in
European Journal ofNeurology, vol. 10, no. 4, 2003, pp. 391–398.[29] Uhlhaas, P. J. and W. Singer,“Abnormal neural oscillations and syn-chrony in schizophrenia,” in
Nature reviews neuroscience, vol. 11, no. 2,2010, pp. 100.[30] Shine, J., A. Handojoseno, T. Nguyen, Y. Tran, S. Naismith, H. Nguyenand S. Lewis,“Abnormal patterns of theta frequency oscillations duringthe temporal evolution of freezing of gait in Parkinson’s disease,” in
Clinical Neurophysiology, vol. 125, no. 3, 2014, pp. 569–576.[31] Gilbertson, T., E. Lalo, L. Doyle, V. Di Lazzaro, B. Cioni and P.Brown,“Existing motor state is favored at the expense of new move-ment during 13-35 Hz oscillatory synchrony in the human corticospinalsystem,” in
J Neurosci, vol. 25, no. 34, 2005, pp. 7771–7779.[32] Joundi, R. A., J. S. Brittain, A. L. Green, T. Z. Aziz, P. Brown and N.Jenkinson,“Persistent suppression of subthalamic beta-band activity dur-ing rhythmic finger tapping in Parkinson’s disease,” in
Clin Neurophysiol, vol. 124, no. 3, 2013, pp. 565–573.[33] JSingh, A., A. Plate, S. Kammermeier, J. H. Mehrkens, J. Ilmbergerand K. B¨otzel,“Freezing of gait-related oscillatory activity in the humansubthalamic nucleus,” in
Basal Ganglia, vol. 3, no. 1, 2013, pp. 25–32.[34] Harmony, T., T. Fern´andez, J. Silva, J. Bernal, L. D´ıaz-Comas, A.Reyes, E. Marosi, M. Rodr´ıguez and M. Rodr´ıguez,“EEG delta activity:an indicator of attention to internal processing during performance ofmental tasks,” in
International journal of psychophysiology, vol. 24, no.1–2, 1996, pp. 161–171.[35] Klimesch, W,“EEG alpha and theta oscillations reflect cognitive andmemory performance: a review and analysis,” in
Brain research reviews, vol. 29, no. 2–3, 1999, pp. 169–195.[36] Putman, P., J. van Peer, I. Maimari and S. van der Werff,“EEG theta/betaratio in relation to fear-modulated response-inhibition, attentional control,and affective traits,” in