Comparison and Affecting Factors of Three tDCS Montages in Motor Recovery of Chronic Stroke Patients: A Resting-State EEG Study

Objective: We aimed at exploring the modulation of tDCS on spontaneous cortical activity through the changing of EEG rhythms to different tDCS montages and the interaction between cortical responses and variability factors of stroke individuals. Methods: 19 stroke subjects underwent 4 tDCS sessions with 3 different tDCS montages (anodal (atDCS), cathodal (ctDCS) and bilateral (bi-tDCS)) and sham stimulation in a single-blind, randomized, controlled crossover design. We acquired resting-state (eyes closing and opening alternately) EEG data before and after tDCS, and calculated the spectral power of each frequency band. Paired-samples T test was applied to examine the difference of spectral power between pre- and post-stimulation of each montage. Three-way repeated measures analysis of variance with lesion hemispheres, stimulation montages and locations were carried out to investigate tDCS effects of different lesion, montages, and channel locations, and the interaction. Further, the effects of tDCS over time were analyzed applying three-way repeated ANOVAs as well with post trials, lesion hemispheres and channel locations separately to each montage. Finally, linear and quadratic regression model were used separately to describe the association between clinical factors of stroke patients and change of spectral power. Results: We found that induced effect of tDCS was limited to the alpha rhythm of opening-state. atDCS increased the alpha power especially alpha1 (8-10 Hz) in local and distant areas of mainly frontal and partial. bi-tDCS affected the alpha power as well, but in a smaller area which mainly focused on alpha2 (10-13 Hz). ctDCS and sham had no effect on alpha rhythm. No signicant difference of alpha band was found over the observed time range after the stimulation over. Results further showed that the quadratic model can better characterize the relationship between clinical factors and the tDCS effects of alpha rhythm than linear model. The changing of alpha especially alpha2 in contralateral hemisphere induced by atDCS was related to time since stroke, and alpha2 in ipsilateral hemisphere induced by bi-tDCS to motor impairment level. Conclusion: Our results provide electrophysiological evidence that different tDCS montages in stroke subjects modulate rhythmic cortical activity of alpha band in different ways, and the effects maintained for at least 30 minutes. The tDCS modulation effect was related to clinical factors, especially the time since stroke and the level of motor impairment. These ndings are of great signicance for the knowledge on modulation effect to stroke patients and for therapeutic application of motor recovery following stroke. stroke patients. We focused on spectral power changing after three kinds of tDCS montages (atDCS, ctDCS and bi-tDCS) and the difference of after-effects among them. We are also interested in the after-effects over time and the relation between changing of alpha and clinical factors of stroke patients. There were four important ndings from our study: (1) the tDCS effect was limited to the alpha rhythm of opening-state. (2) atDCS increased the alpha power especially alpha1 (8–10 Hz) in local and other areas. bi-tDCS affected the alpha power as well, but in a smaller area and mainly focused on alpha2 (10–13 Hz). ctDCS had no effect on alpha rhythm. (3) No signicant difference of alpha band was found over the observed time range after stimulation. (4) The change of alpha especially alpha2 in contralateral hemisphere induced by atDCS was related to time since stroke, and alpha2 in ipsilateral hemisphere induced by bi-tDCS to motor impairment level. Scatter plots and tted curves of representative channels with clinical scale on the abscissa and alpha power’s ratio of post- to pre-stimulation. We applied quadratic tting to these scatters with co-variables of age, gender and lesion hemisphere. The coecient of determination R2 and p-value of Fisher’s F-test were shown in the plots.


Introduction
Transcranial direct current stimulation (tDCS) -a noninvasive brain stimulation (NIBS) technique modulating the local eld potential in neural tissue and cortical excitability has been widely used in recovery post stroke including motor rehabilitation and evidenced its behavioral and neurophysiological effects by numbers of previous studies [1][2][3][4]. However, despite its increasingly application in experimental and clinical settings, the results remain variable, some studies failed to show a positive response to the stimulation in some kinds of stroke patients [5][6][7][8]. Many researchers have thus tried to clarify the sources of variability affecting tDCS's e ciency. Factors like current amplitude of stimulation, placement of electrodes, polar of stimulation electrodes and so on are proved to be possible suspects [8,9]. However, the precise mechanism of how the above factors affecting the stimulating results remains largely unclear.
Among the above mentioned sources of variability, one of the key considerations in using tDCS to improve motor performance after stroke is how the stimulation modulates cerebral cortex [9,10]. Polarity of the electrodes is proved to be especially critical in individuals with stroke due to the spread of functional reorganization in the poststroke brain [10]. Given the hypothesis that rebalancing interhemispheric interactions and/or restoring excitability in the ipsilesional hemisphere is thought to be bene cial for post-stroke motor recovery [11,12], present studies show three montages of tDCS position modes to regulating the excitability of cerebral cortex in poststroke patients: upregulating excitability of the ipsilateral hemisphere through posing the anode tDCS (atDCS) on it; down regulating excitability of the contralateral hemisphere through posing the cathode tDCS (ctDCS) on it; upregulating the ipsilateral cortex and downregulating the contralateral cortex at the same time [13][14][15]. Some studies have found that the excitability or suppression to the brain is not a "one size ts all" approach to recovery following stroke [1,10,15]. It may be related to stroke states like motor impairment level or stroke period (acute, sub-acute or chronic). But the interaction between regulation montages and the above factors is not clari ed clearly.
Previous studies have proved that particular QEEG indices like the delta/alpha power ratio (DAR) and (delta + theta)/(alpha + beta) power ratio (DTABR) are sensitive to some cerebral pathophysiology following stroke and can inform clinical decision-making, including the e cacy of acute reperfusion therapies and outcome prognostication [16]. Some research has found electrophysiological changes in EEG oscillations of healthy people over rest and task states following tDCS over motor related cortex. Ardolino et al. [17] reported that cathodal stimulation of the motor cortex increased the power of delta and theta frequency bands. Mangia et al. [18] found an increase in alpha and beta power of spontaneous EEG during and after atDCS stimulation over postero-parietal cortex. In some motion imagination experiments, event-related desynchronization (ERD) of mu rhythm increased post anodal stimulation over the left primary motor area and decreased post cathodal stimulation [19]. Results from nger motion task showed an increase of ERD in low alpha and beta bands in sensorimotor related regions after atDCS but not ctDCS [20]. The above all focus on healthy people and we found few studies investigate EEG changes after tDCS in stroke patients.
In this study, we aimed at exploring the modulation effects of three different tDCS montages and the association between tDCS effects and variability of clinical factors in stroke individuals. First, we assessed the induced effects of different tDCS montages on spontaneous cortical activity through the changing of power of EEG rhythms. Then, we also investigated the duration effects thirty minutes since the end of tDCS stimulation. Finally, we tried to nd out whether the clinical status (including time since stroke, location of stroke, level of motor impairment and so on) impact these effects. left-hemisphere lesion) at least 3 months after subcortical cerebral infarction were included in this study. All of the patients were diagnosed with ischemic stroke according to MRI. Basic information of subjects was shown in Table 1. All of participants were informed of all aspects of the experiment including the possibility of minor adverse effects related to tDCS, such as transient sensations of itching, burning and prickling on the scalp and signed an informed consent before the experiment started. The study was approved by the ethics committee of Tianjin Union Medical Centre.

Experimental design
This was a single-blind, randomized, controlled crossover experiment, which consisted of four within-subject experimental sessions: three active conditions (anodal, atDCS; cathodal, ctDCS and bilateral, bi-tDCS) and a sham condition. Sham stimulation was used as control in the experiment to isolate the effects solely due to current stimulation and not due to the placebo and somatosensory effects that could arise from tDCS application. We generated a random table using block random method through Matlab program to determine the implementation order of atDCS, ctDCS, bi-tDCS and sham condition. Patients performed the four sessions as the order shown in the randomized table and were blinded to the condition. The interval of each two of the four conditions was at least 1 week.
Each session contained four blocks: Baseline of Electroencephalogram (EEG), tDCS, Electrodes placing for EEG and EEG of post-stimulation. All of the four blocks were conducted in a quiet and electrically shielded room.
Patients were asked to sit and relax on a comfortable chair during the experiment. In the baseline block, patients were required every 2 minutes to open or close their eyes according to a voice prompt produced by E-prime software. This process lasted for 12 minutes (6 trails with 3 eyes closing state and 3 eyes opening state). EEG signals were collected at the same time (Block 1). After that, we conducted one of the four tDCS montages (atDCS, ctDCS, bi-tDCS or sham) according to the random table for 20 minutes (Block 2). Then, electrodes of EEG acquisition were placed on the brain, which lasted for 10 minutes (Block 3). Finally, patients were asked to open and close their eyes alternately (per 2 minutes) again for 20 minutes (10 trials with 5 eyes closing state and 5 eyes opening state). EEG signals were collected meanwhile (Block 4). Figure 1 shows the experimental design of each session.

EEG recording and processing
Recording Resting-state EEG with eyes closed and opened was recorded in a quiet and electromagnetic shielding room.
Participants were instructed to stay awake and avoid movement during the acquisition. EEG was recorded using were also recorded to remove ocular artifacts. All of the electrode impedances were kept below 10 kΩ. The EEG signal was ampli ed with a band pass of 0.1-70 Hz and sampled at 1000 Hz. The forehead was set as ground and linked earlobes were set as reference electrodes.

Spectral power analysis
Firstly, EEG data was desampled into 250 Hz and ltered by a 0.25-45 Hz bandpass lter. Then, Independent component analysis (ICA) was used to remove eyes artifacts. 100 seconds (10-110 seconds) of EEG signals of eye closing state or eye opening state of each trial (120 seconds) were selected for the following spectrum analysis separately. We conducted all the above process through EEGLAB tool box of Matlab software.
Each channel of each trail dataset (100 seconds) applied the following method to calculate the spectral power.

Statistical analysis
To explore the effects of stimulation, paired-samples T test was applied to examine the difference of spectral

Association between tDCS and clinical facotrs
The regression analysis showed that the changing of alpha power (ratio of post-stimulation to pre-stimulation)

Discussion
This study aimed at investigating the tDCS effects of different montages on cortical activity of chronic ischemic stroke patients. We focused on spectral power changing after three kinds of tDCS montages (atDCS, ctDCS and bi-tDCS) and the difference of after-effects among them. We are also interested in the after-effects over time and Alpha band of EEG has been proved by several studies to be a brain rhythm involved in several cerebral functions, ranging from sensory-motor processing to memory formation [22,23]. Ischemic stroke showed an attenuation of normative, faster activity, particularly in the alpha band (8-12 Hz) [24,25]. Alpha band of stroke patients was found to be locally reduced in brain regions critical to observed motor or cognitive behavioral de cits. The decrease of alpha band synchrony was found related to cognitive and motor de cits in post-stroke patients [26].
Some study showed that motor recovery could be predicted by increased alpha-band functional connectivity in motor-related areas [27]. So we supposed that the increasing of alpha band may be bene cial to stroke recovery.
Previous studies investigating the changing of cortical activity after tDCS through rs-EEG power spectrum analysis were mainly focused on healthy people and showed response difference among stimulation montages.
Some research showed increase of alpha band after atDCS in healthy people, but not after ctDCS, which is similar to our result in stroke subjects. Notturno et al. [20] found a higher low alpha band power post-than pre-atDCS over motor related regions, but not for ctDCS. Spitoni et al [28] explored the tDCS effect over the right posterior parietal area in healthy subjects and found that the effect was limited to the alpha band, and atDCS signi cantly affected the alpha band whereas the ctDCS did not elicit any modi cations. This is consistent with our nding in stroke patients. However, the impact was shown in eye-closing but not in eye-opening state, which is different with this study. We hypothesized that this might be related to sleepiness with eye-closing state in some patients. Besides that, the difference of area of stimulation target may be another impact factor. For bilateral tDCS, studies are mainly focused on its rehabilitation e cacy on stroke patients [29][30][31][32]. A reduction of inter-hemispheric imbalance was found after a long-term effect of tDCS associated with physical therapy according to the analysis of motor evoked potential (MEP) [30]. We found no reports about EEG power spectrum following the bi-tDCS montage. In our study, both atDCS and bi-tDCS modulated alpha band, but atDCS preferred low-band alpha and bi-tDCS preferred high-band alpha. Studies have shown that different alpha components correspond to different cognitive processes. Low-band alpha rhythm was supposed to be related to anticipatory attentional processes and high alpha would indicate task-speci c visuo-motor processes according to some task-related ERD/ERS study [33]. We speculated that these different changes on alpha rhythm induced by the two montages may imply that they work in different ways.
There are also some inconsistent reports with our results. Besides alpha band, some study showed power changing in other frequency band after tDCS [17,34,35]. In some study rs-EEG power spectrum analysis showed no difference comparing baseline with post stimulation in any of the tDCS conditions (one-hemispheric tDCS or bi-lateral tDCS) over dorsolateral prefrontal cortex in healthy subjects [36,37]. Confounding results may be due to the difference in stimulation target, current density, time of duration and participants.
For the stimulus target area of tDCS, recent studies showed that brain stimulation leads not only to local changes of cerebral activity under the stimulated region, but also to distant changes in inter-connected brain regions throughout the brain [34,38,39], which is consistent with our results. Besides the local target area, we found that alpha power of some distant areas including frontal and parietal showed an increase following atDCS and bi-tDCS. Besides that, the in uence of atDCS was more widespread associated with bi-tDCS.
For the duration effect of tDCS, some study reported increased alpha power during and after atDCS which persisted for 12 minutes without attenuation [18]. Spitoni et al [28] reported that the strongest change of alpha power occurred in the rst 2 min after the atDCS ended, and the effect diminished systematically and was effective for approximately 8 min. We missed the rst 10 min EEG information immediately after stimulation because of placing electrodes. So only the 10-30 min EEG signals after stimulation were analyzed in the present study. Although alpha power didn't change in the 10-30 min after stimulation, they keep higher level than that of pre-stimulation, indicating that the effect maintained for at least 30 minutes with no signi cant attenuation.
For the clinical factors affecting modulation results, previous studies have found that tDCS stimulation e cacy may vary with time after stroke, nature and location of stroke and level of motor impairment [9,29,40]. We found that the change of alpha especially alpha2 in contralateral hemisphere induced by atDCS was related to time since stroke, and alpha2 in ipsilateral hemisphere induced by bi-tDCS to motor impairment level. Regression analyses con rmed that individuals' response of alpha power change to atDCS could be predicted from their time after stroke. Stroke subjects with 3 to 6 months and longer than 20 months since stroke showed higher alpha power increase than other subjects, indicating higher response to the montage of atDCS. For bi-tDCS, alpha band power increased the most in moderately impaired subjects with respect to mild and severe impairment, implying that subjects with moderate motor impairment were more susceptive to this kind of montage. Plasticity processes were variable with different phases or degree of stroke. tDCS effects may interact with these processes. Studies have suggested that patterns of neural recovery may differ for individuals based on the severity of their stroke [41][42]. The quadratic regression model was better suited to model the variation trend than linear regression model both for atDCS and bi-tDCS, indicating a complicated relation among clinical factors and EEG parameters.
The result may help explain the variable rehabilitation e cacy to individuals with different clinical stroke features using tDCS.

Conclusions
Our results provide electrophysiological evidence that different tDCS montages in stroke subjects modulate rhythmic cortical activity of alpha band in different ways, and the effects maintained for at least 30 minutes. The tDCS modulation effect was related to clinical factors, especially the time since stroke and the level of motor impairment. These ndings are of great signi cance for the knowledge on modulation effect to stroke patients and for therapeutic application of motor recovery following stroke.
There are some limitations in our study. Present study only found the changing of alpha power induced by tDCS, longitudinal analysis was need to verify the correlation of motor recovery and changing of alpha power. Besides that, although we did not nd any difference in spectral power between pre-and post-stimulation on ctDCS montage, it does not mean ctDCS has no effect on cortical activity or stroke patients. We plan to try other methods like network connectivity or non-linear dynamic analysis to explore the performance of cortical electrical activity after ctDCS and other montages in our next work.
Declarations the statistical analysis; JD sponsored this study and supervised the nal manuscript. All authors read and approved the nal manuscript. Availability of data and material The datasets used and analyzed during the current study are available from the corresponding author on request.
Ethics approval and consent to participate All of participants were informed of all aspects of the experiment including the possibility of minor adverse effects related to tDCS, such as transient sensations of itching, burning and prickling on the scalp and signed an informed consent before the experiment started. The study was approved by the ethics committee of Tianjin Union Medical Centre.

Consent for publication
Not applicable

Declarations of interest
The authors declare that they have no competing interests.
52. Klimesch W. EEG alpha and theta oscillations re ect cognitive and memory performance: a review and analysis.