Effect of EEG Neurofeedback Training in Patients with Moderate – Severe Traumatic Brain Injury: A Clinical and Electrophysiological Outcome Study

Traumatic brain injury (TBI) is a leading cause of death, and its survivors with a disability are considered to be an important global health priority. In view of a diverse range of disability and its impact on TBI survivors, the need for effective rehabilitation modalities is on a high rise. Therefore, the present study was aimed to investigate the efficacy of EEG neurofeedback training (EEG-NFT) in moderate–severe TBI patients on their clinical and electrophysiological outcomes. The study was an experimental longitudinal design with a pre-post comparison. A total of 14 TBI patients in a postinjury period between 3 months to 2 years were recruited. All participants received twenty sessions of EEG-NFT. Baseline and post-NFT comparisons were made on postconcussion symptoms (PCS) and electrophysiological variables. The result indicates a significant reduction in the severity of PCS following EEG-NFT. A consistent pattern of reduced slow waves and fast waves amplitude ratios was also noted at post-NFT, although it was not significant across all the brain regions. The present study suggests EEG-NFT as a contributing factor in improving PCS and normalization of qEEG in TBI patients, which holds an implication for clinical decision-making of EEG-NFT as a viable alternative to be offered to TBI patients.


Introduction
Traumatic brain injury (TBI) disrupts the normal functioning of the brain caused by a bump, blow, or jolt to the head (Marr & Coronado, 2004). It is a major concern worldwide, also referred to as "The Silent Epidemic" (Rusnak, 2013;Vaishnavi, Rao, & Fann, 2009).
The global incidence of TBI is estimated to be 69 million individuals per year (Dewan et al., 2018). In India, it is estimated that annually approximately 1.6 million individuals sustain a TBI (Gururaj, 2002). The prevalence of TBI increased by 8.4% from 1990 to 2016 and accounts for a considerable portion of the global injury burden (GBD 2016 TBI andSCI Collaborators, 2019). The major etiological factors of TBI are road traffic accidents (60%), falls (20-25%), and violence (10%; Gururaj, 2002). From 2003 to 2013, in India road accidents have increased by 5% per year while the population increased by 1.4% per year, suggesting a high prevalence of TBI (Singh, 2017).
The consequences of TBI are not only circumscribed to these overt changes and dysfunctions but also lead to the disruptions and alterations of brain function, including changes in electrophysiological patterns. These alterations have been found to be associated with poor functional outcomes. EEG abnormalities can be focal, multifocal, or widespread depending upon the severity and location of the injury (Brigo & Mecarelli, 2019;Galovic, Schmitz, & Tettenborn, 2018).
Acute disruption of cortical-thalamic networks led to an increase in delta and theta band and a decrease in beta band in TBI (Moeller, Tu, & Bazil, 2011). A consequential higher theta-alpha, theta-beta, and delta-alpha amplitude ratio and minimized EEG coherence were also noted in mTBI (Chen, Tao, & Chen, 2006;Modarres, Kuzma, Kretzmer, Pack, & Lim, 2016;Moeller et al., 2011;Watson et al., 1995). An epileptiform activity has been observed immediately followed by a diffuse slowing of the EEG after head injury (Walker, Kollros, & Case, 1945).
With a diverse range of disability and its impact on TBI survivors, new intervention modalities are being attempted to address the TBI-related issues. One of such modalities is EEG neurofeedback training (EEG-NFT) that uses electrophysiological measures of an individual to self-regulate their psychophysiological state (Ali, Viczko, & Smart, 2020). It is a noninvasive and nonpharmacological intervention based on the principles of operant conditioning.
There are very limited to no studies being attempted of investigating clinical and electrophysiological changes in the moderate-severe TBI following EEG-NFT. Therefore, the present study uses the alpha reinforcement and theta inhibition training with the aim to reduce theta-alpha amplitude ratio to explore the electrophysiological alterations and the subsequent consequences on PCS among patients with moderate-severe TBI.

Methods and Materials Participants
The sample comprised of 19 individuals (15 males and 4 females) diagnosed with TBI with normal or corrected hearing and vision in the age range of 18-50 years (mean age = 32.47 years; SD = 7.52). All participants with TBI had a Glasgow Coma Scale (GCS) score 12 or less with a postinjury period between 3 months to 2 years.
Participants with a diagnosis of mTBI (GCS: 13-15), with extracranial injuries, having a previous history of any comorbid neurological, psychiatric, or neurosurgical conditions, substance dependence, or mental retardation, and those who underwent any form of structured psychological intervention in the last year were excluded.

Procedure
After obtaining ethical clearance from the Institute Ethics Committee, a written informed consent form was sought from each participant who met inclusion criteria. Sociodemographic and clinical details were obtained, and baseline assessments were conducted using the Rivermead Postconcussion Symptoms Questionnaire (RPQ) and a resting-state eyes-opened EEG recording.
Following the baseline, all the participants received 20 sessions of EEG-NFT (those who completed 80% of sessions were also considered as completers). To examine the posttraining effect, the same baseline assessments were readministered immediately after the completion of EEG-NFT.

Rivermead
Postconcussion Symptoms Questionnaire. It was used to assess the severity of the symptoms reporting in the postinjury period. It consists of 16 items assessing the most commonly reported PCS. The scores ranged from 0-4 where 0 indicates the symptoms were not experienced, 1 as the symptom was no more a problem, 2 as a mild problem, 3 as a moderate problem, and 4 as a severe problem. The participants were asked to rate the degree to which they experienced the symptoms. The total score represented the overall severity of PCS.
EEG recording. The EEG was conducted in a dimly lit, sound-attenuated room while the patient was seated comfortably. The recording was performed using SynAmps amplifiers (Compumedics Neuroscan, Charlotte, NC) with 32 Ag/AgCl, passive electrodes, fitted in the lycra stretch cap. Sampling frequency was kept at 1 kHz with a notch filter at 50 Hz. For eye movement, horizontal and vertical electrooculograms (EOG) were used bipolarly. One electrode on each mastoid was used as a reference. Electrodes impedance was ascertained less than 10 kΩ.
Intervention. The participants received 20 sessions of EEG-NFT conducted three times a week spanning the whole intervention program over a period of 2 months. It was carried out in a quiet, dimly lit room using a dedicated NFT system (Atlantis system, BrainMaster Technologies, Inc., Bedford, OH). Each participant received alpha-theta training (reinforcing alpha and inhibiting theta) activity with the aim of reducing the theta-alpha amplitude ratio. The active sites were fixed at O1 and O2 locations as per the 10-20 International system, each reference electrode on mastoid, and the ground electrode on the forehead. An abrasive gel was used to clean and prepare the scalp/skin followed by mounting the electrode using a conductive paste. Before the procedure, the goal and nature of the task were explained thoroughly to the participant. The display screen was selected as per the participants' choice. The participants were instructed to relax and focus on the screen. The reward was given through visual feedback (i.e., an increase in the score), which is displayed on the screen.
Each NFT session lasted for 40-min duration.
The training was done under the supervision of a trained clinical neuropsychologist (as per the norms of the rehabilitation council of India).
Finite impulse response (FIR) bandpass filter from 0.1 to 30 Hz with a zero-phase shift at 12 dB/octave was applied to retain all relevant frequencies. For eye movement and other artifacts corrections, EEG data were marked manually, and spatial filter transformation was performed through principal component analysis (PCA) using singular value decomposition (SVD). Spectral analysis was performed on artifact-free data using 1024 data points. The signals from all the electrode positions underwent the fast Fourier transformation (FFT) on 500 ms epochs with a Hanning window of 1024 Hz.
Further statistical analyses were carried out on SPSS v20.0. To check the normality for all values of interest Shapiro-Wilk test was performed (Shapiro & Wilk, 1965). The data group that was normally distributed a paired t-test was performed, while for the data that violated the normality assumption, a similar nonparametric Wilcoxon signed-rank test was used. A statistical significance threshold was set at p < .05.

Results
From the 19 participants with TBI who were recruited for EEG-NFT, two participants dropped out (did not turn up for sessions after baseline assessment or did not complete up to 80% of the sessions). From the remaining 17 participants, three patients could not complete baseline and/or post-NFT assessments.

Rivermead Postconcussion Symptoms Questionnaire (RPQ)
The RPQ-total score which forms the severity of TBI symptoms significantly reduced (p = .018) in post-NFT compared to the baseline. The effect size within-subjects also showed a medium effect (0.725) on RPQ-T scores (Table 1; Figure 1).

EEG Neurofeedback Training (EEG-NFT)
For the EEG-NFT sessions, a ratio of an average amplitude of theta and alpha frequency bands was calculated at O1 and O2 locations in the first and last session. The result indicates that the thetaalpha ratio has reduced at both O1 (p = .665) and O2 (p = .011) locations, although this was not statistically significant at O1 (Table 2; Figure 2).

EEG Analysis
For each electrode, EEG amplitude values were averaged across the participants. Further, these electrodes were grouped into five different brain regions to examine the regional differences in EEG amplitude. An average score of the individual electrode in that region formed the score for each region (Figure 3). . 32-electrodes were grouped into five brain regions (frontal, central, temporal, parietal, and occipital) as per 10-20 system.
A consistent pattern of a reduced delta-alpha, thetaalpha, and theta-beta ratios ratio was observed across all the brain regions in post-NFT compared to the baseline. Although this was statistically significant only in the temporal (p = .041) and central (p = .038) regions for delta-alpha and in the occipital (p = .033) for theta-alpha (Table 3; Figure 4).

Table 3
Average EEG amplitude of delta-alpha, theta-alpha, and theta-beta ratio at baseline and post-NFT (n = 14). Note. * Significance at 0.05 level.

Discussion
The current study investigated the efficacy of EEG-NFT in patients with moderate-severe TBI on their clinical and electrophysiological outcomes. Participants were assessed at baseline and post-NFT using Rivermead Postconcussion Symptoms Questionaire total (RPQ-T) score and EEG amplitude.

Effectiveness of EEG-NFT on Clinical Outcome
The findings from the study indicate a significant reduction in the severity of PCS on the RPQ-T score. These findings are in line with previous studies showing that EEG-NFT leads to a significant decrease in PCS (Rajeswaran, Bennett, Thomas, & Rajakumari, 2013;Reddy et al., 2013). A study by Reddy et al. suggested a negative correlation of RPQ with QOL and neuropsychological functioning (Reddy, Rajeswaran, Devi, & Kandavel, 2017). Therefore, the reduction of PCS on RPQ-T might contribute to improving QOL and cognitive functioning in patients with TBI, which is corroborated by earlier studies (Bennett, Sampath, Christopher, Thennarasu, & Rajeswaran, 2017;Hoffman, Stockdale, & van Egren, 1996;Munivenkatappa et al., 2014;Reddy, Rajeswaran, Bhagavatula, & Kandavel, 2014).

Effectiveness of EEG-NFT on the Electrophysiological Outcome
EEG amplitude ratio is potentially an important indicator of cognitive ability (Trammell, MacRae, Davis, Bergstedt, & Anderson, 2017) and constitutes a more reliable index to monitor electrophysiological alterations over time in TBI (Álvarez et al., 2008). The qEEG data reported herein suggest a consistent pattern of reduced slow waves and fast waves (SW/FW) amplitude ratios at post-NFT. Although significant changes were observed only for deltaalpha in the temporal and central regions and for theta-alpha in the occipital region.
A positive association of cognitive deterioration has been found with an increased SW/FW ratio in patients with moderate-severe TBI (Álvarez et al., 2008). A study by Leon-Carrion et al. indicates a negative correlation between delta-alpha ratio and functional outcome in patients with head injury (Leon-Carrion, Martin-Rodriguez, Damas-Lopez, Barroso y Martin, & Dominguez-Morales, 2009). An increased theta-beta ratio has been related to higher impulsive behavior (van Dongen-Boomsma et al., 2010) and lower response inhibition (Putman, van Peer, Maimari, & van der Werff, 2010). Therefore, a reduction in the SW/FW amplitude ratio might be related to better cognitive functioning (Álvarez et al., 2008) and could be attributed to significantly reduced PCS observed in our study.
These qEEG changes can be suggested by modulation in thalamo-cortical networks that refines the intrinsic neural network, led to the normalization of qEEG pattern in TBI following EEG-NFT (Munivenkatappa et al., 2014).
The findings from the EEG-NFT sessions indicate that qEEG changes were not due to chance, as there were progressive changes in qEEG across NFT sessions. It is also worthwhile noticing that electrophysiological changes in the present study were marked 3 months to 2 years of postinjury, suggesting that these changes were not concomitant by the time.
To conclude, the findings suggest EEG-NFT as a contributing factor in improving postconcussion symptoms and normalization of qEEG in patients with moderate-severe TBI. The present study also holds an implication for clinical decision-making of EEG-NFT as a viable alternative to be offered to patients with moderate-severe TBI. The limitations of the present study are the small sample size, limited variables, and lack of control group. Accounting together these limitations affect the generalizability of the study. Therefore, future research would require structural, functional, biochemical, and cognitive correlates on a larger cohort following the intervention.