Single-channel EEG captures tDCS-induced changes in neural reactivity: a pilot study in disorders of consciousness patients and healthy controls

Abstract

Background:

Understanding and enhancing neural recovery in disorders of consciousness (DoC) remains a critical challenge in clinical care. Clinical scales such as the Coma Recovery Scale–Revised (CRS-R) are essential, but reflect behavioral responsiveness and may not fully capture subtle neurophysiological changes. Transcranial direct current stimulation (tDCS) has shown promise, yet objective biomarkers to assess treatment response are needed.

Objectives:

This proof-of-concept study explores the feasibility and sensitivity of combining a compact single-channel EEG with noninvasive brain stimulation to assess changes in neural reactivity in individuals with DoC.

Design:

Pilot observational study with DoC patients undergoing intervention; healthy controls were assessed to provide a physiological reference benchmark for interpreting mismatch negativity (MMN) responses and assess the specificity of EEG biomarkers relative to DoC patients.

Methods:

Six DoC patients underwent a 2-week open-label anodal tDCS protocol targeting the left dorsolateral prefrontal cortex. Neural activity was tracked using a compact three-electrode EEG alongside traditional 19-electrode EEG. To capture treatment-induced changes, three complementary biomarkers were analyzed: MMN, reflecting pre-attentive auditory discrimination function; Beta oscillatory activity, associated with global attentional processes; and L1, a machine-learning-derived biomarker of cognitive engagement. CRS-R, auditory event-related potentials, and EEG-based machine learning biomarkers were recorded at baseline and immediately post-intervention. Healthy controls underwent the same baseline EEG assessments without receiving stimulation.

Results:

MMN, Beta, and L1 demonstrated task-related modulation in healthy participants across cognitive load levels. In DoC patients, changes were observed following tDCS in MMN, Beta activity, and L1 as well, with patients demonstrating complex heterogeneous response patterns. Two patients showed behavioral improvement (CRS-R), while others exhibited increased EEG reactivity without corresponding behavioral gains.

Conclusion:

These findings highlight the feasibility and potential value of integrating compact EEG systems with neuromodulation protocols to monitor neural reactivity in DoC patients, with the potential to inform and refine tDCS treatment strategies over time.

Trial registration:

NIH Clinical Trials Registry number: NCT04614792 (https://clinicaltrials.gov/study/NCT04614792).

Plain language summary

Detecting brain responses after stimulation in patients with disorders of consciousness using single-channel EEG

This study explored whether a single-channel EEG device can detect changes in brain activity in patients with disorders of consciousness (DoC), such as vegetative state or minimally conscious state. These patients are often unable to communicate, making it difficult to assess brain function using behavior alone. In this pilot study, patients received a 10-day course of noninvasive brain stimulation (tDCS). Brain activity was measured before and after the intervention using EEG. Healthy participants were also tested to provide a reference for typical brain responses. The results showed that EEG could detect changes in brain activity following the stimulation, even when patients’ clinical scores (based on behavior) remained unchanged. In particular, one EEG measure (called L1) was able to distinguish between resting state and active task engagement, suggesting it reflects whether the brain is responding to external stimuli. These findings suggest that EEG may help identify brain activity that is not visible through behavior alone, potentially improving the assessment and monitoring of patients with disorders of consciousness. Larger studies are needed to confirm these results.

Keywords

disorders of consciousness EEG EEG biomarkers mismatch negativity (MMN)neuroplasticity noninvasive brain stimulation tDCS

Introduction

Disorders of consciousness (DoC) are among the most severe outcomes of brain injury, and their diagnosis and treatment remain major clinical challenges. DoC encompass a spectrum of conditions, primarily including the vegetative state/unresponsive wakefulness syndrome (VS/UWS),¹ characterized by preserved wakefulness without signs of awareness, and the minimally conscious state (MCS),² in which patients demonstrate reproducible but fluctuating signs of conscious behavior, such as command following or purposeful responses.³ While the Coma Recovery Scale–Revised (CRS-R) is considered the clinical gold standard,⁴ it reflects behavioral responsiveness and may not fully capture covert or sub-behavioral neurophysiological changes. This gap has motivated efforts to integrate clinical assessment with neurophysiological markers and neuromodulatory approaches.

Transcranial direct current stimulation (tDCS) is a noninvasive neuromodulation technique to modulates cortical excitability by applying weak electrical currents (0.5–2 mA), shifting neuronal resting membrane potentials toward excitation or inhibition within a physiologically safe range.⁵ By facilitating neuroplasticity, tDCS has shown promise in improving consciousness levels in subsets of patients with DoC.

Multiple studies indicate that tDCS is particularly effective in patients diagnosed with a MCS. A meta-analysis of eight trials with 181 patients found significant improvements in CRS-R scores, favoring active stimulation over sham, particularly in MCS patients.⁶ Another meta-analysis of 9 trials with 331 participants confirmed that tDCS improved CRS-R scores without adverse effects, showing efficacy in MCS but not in VS/UWS patients.⁷ Consistent with these findings, a double-blind, sham-controlled crossover study involving 55 patients found that a single 20-min session of anodal tDCS over the left dorsolateral prefrontal cortex (DLPFC) led to transient improvements in consciousness (as measured by CRS-R), with MCS patients showing a significantly greater response than VS/UWS patients.⁸ Repeated stimulation protocols appear to have an even stronger impact, as shown by a study where 10 consecutive daily sessions of anodal tDCS over the left DLPFC significantly improved clinical coma scores immediately after treatment in MCS but not VS/UWS patients.⁹ Notably, one MCS patient who received a second round of tDCS 3 months later exhibited further recovery, suggesting cumulative benefits of repeated stimulation. Another study supporting this notion investigated the effects of home-based tDCS in 16 MCS patients, presenting either active or sham tDCS over the left DLPFC for 20 min daily over 4 weeks, and showed sustained improvements in CRS-R scores for up to 3 months post-treatment. This suggests that repeated tDCS sessions could lead to sustained enhancements in consciousness levels in MCS patients.¹⁰

The DLPFC is integral to a range of cognitive processes, including working memory, attention regulation, and the integration of environmental and verbal stimuli. The left DLPFC is known to be involved in higher-order cognitive functioning; residual neural activity in this region is thought to influence both awareness and event-related EEG responses in DoC patients.¹¹ Event-related potentials (ERPs) are time-locked EEG responses to sensory stimuli that reflect the sequential processing of information in cortical networks. The mismatch negativity (MMN)¹² is one such marker: it reflects pre-attentive auditory discrimination occurring in temporal–frontal cortex networks prior to conscious awareness. MMN is assumed to index detection of changes in auditory stimuli and is commonly used to assess early sensory processing in research, intrinsically related to higher-order cognitive functions.¹³ In DoC patients, MMN amplitude and latency correlate with consciousness level, cognitive and functional decline, and changes in temporal and prefrontal cortex excitability.¹⁴ These associations make MMN a promising marker for monitoring cognitive and awareness changes and assessing the impact of interventions such as tDCS in DoC.

Moreover, frontal oscillatory activity, particularly within the Theta (4–7 Hz) and Beta (13–30 Hz) frequency bands, has shown promise in differentiating levels of consciousness. For example, studies have identified increased Theta power in the frontal regions of DoC patients following prefrontal tDCS treatment.¹⁵ In DoC, this frontal Theta increase is associated with enhanced cognitive processing, clinical recovery, and long-term survival, especially in response to verbal commands.¹⁶ Beta activity, on the other hand, is often linked to attentional control and active cognitive engagement,¹⁷ suggesting that its modulation might also play a role in tDCS-induced changes in DoC patients. EEG-based biomarkers have shown promise as a complementary tool to behavioral assessment in detecting neural processes not always reflected in CRS-R performance. For instance, a study found that a combination of EEG markers, including low-frequency oscillatory power, EEG complexity, and information exchange, served as reliable indicators of consciousness levels and can identify covert cognition in DoC patients who did not exhibit overt behavioral responses during CRS-R assessments.³ Another study demonstrated that combining EEG evaluations with CRS-R scores improved the prediction of clinical improvement.¹⁸ Moreover, EEG-based neural reactivity markers, including MMN amplitudes and Theta power following verbal commands, were more sensitive than standard clinical assessments in predicting 1-year survival in DoC patients.¹⁶ Beyond traditional spectral measures, prior work found that machine learning (ML)-derived EEG feature L1 can distinguish between low- and high-load working memory conditions, outperform Theta power in detecting subtle task-related variations, and differentiate attentional states such as responses to high- versus low-importance stimuli.^19,20 Its inclusion alongside conventional markers like ERP and frequency bands allows for the exploration of advanced analytic features that are rarely applied in DoC neuromodulation studies.

These findings underscore the potential of EEG-based markers in assessing consciousness alongside clinical measures. However, practical challenges associated with full-cap EEG limit its feasibility for routine clinical use. One drawback is its cumbersome setup, which requires extensive preparation time and precise electrode placement, making it impractical for routine bedside assessments. Additionally, full-cap EEG systems are often expensive and require trained personnel for proper use and interpretation, limiting their accessibility in many clinical settings, where rapid and accessible monitoring is essential. These challenges highlight the need for more practical alternatives, such as reduced-channel EEG systems and wearable technologies, which aim to balance ease of use with accurate brain activity capture for reliable neurophysiological assessments.

The primary aim of this pilot study was to evaluate the feasibility and sensitivity of a compact single-channel EEG to detect neurophysiological changes following a 10-day open-label prefrontal tDCS intervention in individuals with DoC. Secondary exploratory aims included examining associations between EEG biomarkers (MMN, spectral power, and L1) and behavioral measures (CRS-R), and validating biomarker sensitivity across graded cognitive load in healthy participants.

Materials and methods

Study design

This prospective pilot observational study evaluated the effects of tDCS in DoC patients, while healthy controls were assessed to provide a physiological reference benchmark for interpreting MMN responses in the patient cohort and to validate the sensitivity of the EEG biomarkers. All participants were recruited and recorded at Herzog Medical Center using the same EEG systems. Patients’ enrollment took place over a 2-year period (2021–2023), and healthy controls were tested during several consecutive weeks in 2023. The study was conducted as a nonrandomized, single-arm feasibility study and was prepared and reported in accordance with the STROBE guidelines (combined checklist for observational studies, see Supplementary Materials).

Participants

DoC patients

Six chronic DoC patients (one female and five males, ages 24–81; see full demographic details in Table 1) clinically classified as either VS/UWS (n = 4) or MCS (n = 2) based on CRS-R scores were included in the study. The patients constituted a pilot sample recruited for feasibility assessment of the intervention protocol. Chronicity was defined as more than 3 months post-injury for nontraumatic etiologies and more than 12 months post-injury for traumatic etiologies, consistent with standard definitions in DoC²¹.

Table 1.

Demographic and clinical characteristics of individuals with DoC included in the study.

Patient	Age (years)	Sex	Etiology	Time since injury (months)	Diagnosis	Baseline CRS-R
P 11	81	M	CVA	17	VS/UWS	9
P 12	25	M	Anoxic	79	MCS	6
P 13	71	M	CVA	10	VS/UWS	9
P 14	24	M	TBI	86	MCS	8
P 15	60	F	Anoxic	90	VS/UWS	7
P 16	72	M	Anoxic	20	VS/UWS	5

CRS-R, Coma Recovery Scale–Revised; CVA, cerebrovascular accident; DoC, disorders of consciousness; F, female; M, male; MCS, minimally conscious state; TBI, traumatic brain injury; UWS, unresponsive wakefulness syndrome; VS, vegetative state.

Inclusion criteria included a diagnosis of anoxic brain injury, traumatic brain injury, cerebrovascular accident, progressive dementia, or encephalopathy, with a clinical classification as either VS/UWS or MCS based on CRS-R; age between 18 and 90 years; and written informed consent provided by a legally authorized representative. Exclusion criteria included intracranial metal implants; implanted devices potentially affected by tDCS (e.g., pacemaker, medication pump, cochlear implant, or implanted brain stimulator); medical conditions compromising stability for participation (including significant ECG abnormalities, cardiac arrhythmia, or uncontrolled hypertension); and participation in other interventional research studies. Informed consent was obtained from all participants’ legal guardians, and the study was approved by the Herzog Medical Center Institutional Review Board in accordance with the Declaration of Helsinki.

Healthy controls

Ten healthy participants (4 females (40%) and 6 males (60%), age range 27–71, mean age 39.8 years, SD = 15.62) were enrolled in the study as healthy controls. These participants constituted a pilot control sample recruited to provide a physiological reference benchmark for interpreting MMN responses in the patient cohort to validate the sensitivity of the EEG biomarkers. Inclusion criteria included age 18 years or older, absence of neurological or psychiatric disorders, normal hearing, and provision of written informed consent. Exclusion criteria included use of psychoactive medication, history of epilepsy or significant head injury, metallic cranial implants or other contraindications to EEG recording, or any condition that could interfere with cognitive task performance or EEG data quality. All participants provided informed consent, and the study was approved by the Tel Aviv University Ethics Committee.

Procedure

DoC patients

Patients were enrolled according to the study’s inclusion criteria, and written informed consent was obtained from their legal guardians prior to participation. The procedure involved administering active tDCS over the left DLPFC, following the protocol outlined by Thibaut et al.⁸ Each patient received 20-min tDCS sessions across 10 consecutive weekdays (excluding weekends) over a 2-week period. Pre- and post-tDCS assessments included both clinical and neurophysiological measurements. Clinical evaluation was conducted using the CRS-R.⁴ Neurophysiological data were recorded using a clinical 19-electrode EEG to capture MMN, and a single-channel EEG to record frontal frequency bands, and EEG biomarker L1 activity during standard CRS-R assessments. ERP recordings were conducted at the bedside under standard clinical conditions during periods of observable wakefulness (i.e., eyes open/arousal) as assessed by clinical staff, consistent with CRS-R evaluation procedures. CRS-R improvement was defined as an increase in total CRS-R score from baseline to post-intervention. No diagnostic category shift was required to define improvement. Clinical vital signs, including pulse rate, oxygen saturation, and breathing rate, were continuously monitored throughout the study, along with blood count and biochemistry assessments.

Healthy controls

The healthy participants underwent a cognitive auditory assessment battery and a CRS-R exam while being recorded with the single-channel EEG. Participants remained seated during the assessment, receiving instructions through a loudspeaker. The entire recording session typically lasted 20–30 min, including a 15-min cognitive assessment battery. The battery comprised pre-recorded tasks: a musical detection and n-back tasks, and a resting-state period, as previously described.^22,23 Behavioral measurements were recorded using a clicker, including reaction times and accuracy. Additionally, healthy participants underwent a recording with a 19-electrode EEG system to extract MMN results. The presence of robust MMN responses in healthy participants confirms intact automatic auditory change detection mechanisms and provides a physiological reference benchmark for interpreting MMN responses in the patient cohort. The CRS-R was administered to healthy controls to ensure the comparability of the assessment context and to confirm the expected ceiling performance under standard administration.

tDCS intervention

tDCS was delivered using a battery-driven constant current stimulator (BrainSTIM Transcranial Stimulator, E.M.S. s.r.l., Bologna, Italy) with saline-soaked synthetic sponge electrodes. The anode was placed over the left DLPFC in electrode location F3, and the cathode was positioned on the right supraorbital area above the right eyebrow to ensure consistent electrode placement across participants. The stimulation parameters were set to a current intensity of 2 mA and an electrode size of 35 cm². We used the same tDCS protocol as in Thibaut et al.⁸ Participants received tDCS for 20 min/day over 10 consecutive weekdays, over a period of 2 weeks.

EEG measures and acquisition

To capture complementary aspects of brain function, three classes of neurophysiological measures were analyzed: (1) MMN, reflecting pre-attentive auditory discrimination, acquired using a 19-electrode EEG system; (2) frontal spectral power, particularly in the Beta band, reflecting global cortical activation and attentional state, derived from single-channel EEG recordings; and (3) L1, an ML-derived EEG feature reflecting cognitive engagement and working memory load, also derived from single-channel EEG recordings. These measures were selected to probe distinct but complementary levels of processing, ranging from automatic sensory responses to higher-order cognitive function. Each measure was preprocessed and analyzed using modality-specific pipelines, as detailed in the following sections.

19-Electrode EEG and ERP analysis (MMN task)

EEG data were recorded using a 19-electrode EEG system according to the International 10–20 system, with a reference electrode placed at Fz and ground at Cz. EEG acquisition and stimulus presentation were performed using PSYTASK 2.x software synchronized using WinEEG 2.115.82 software (Mitsar Co. Ltd, St-Petersburg, Russian Federation). Pre-processing pipeline included artifact correction using independent component analysis (ICA), band-pass filtering, high-order band-pass filtering, and automatic artifact rejection.

ERP analysis focused on auditory evoked responses elicited by an oddball paradigm. Auditory stimuli included a standard low-frequency tone (1000 Hz; probability 0.8) and two deviant stimuli: a high-frequency tone (1300 Hz) and a complex tone composed of five brief tones (500, 1000, 1500, 2000, and 2500 Hz), each presented with a probability of 0.1. Stimuli were presented for 100 ms with an interstimulus interval of 850 ms. Following artifact rejection, only clean standard and deviant trials were included in the analysis. The MMN waveform was computed by subtracting the standard ERP waveform from the deviant ERP waveform within a 500-ms post-stimulus window, using a 100-ms pre-stimulus baseline correction. MMN peak negativity was identified within the 50–250 ms time window post-stimulus onset. Amplitude and latency differences pre- and post-intervention were assessed at the midline Fz electrode using paired-sample t-tests. Waveform-level differences within the 50–250 ms window were further evaluated using point-wise permutation testing. Additional analyses examined peak negativity across midline Fz, Cz, and Pz electrodes.

Single-channel EEG

EEG data processing was performed in three stages: (1) preprocessing, (2) feature extraction (spectral power and L1), and (3) statistical analysis. EEG recordings were obtained using the high dynamic range Neurosteer^® Recorder (hdrEEG, neurosteer Inc. NY, USA). A three-electrode medical-grade patch was applied to the subject’s forehead with dry gel to ensure optimal signal transduction. The noninvasive electrodes were positioned over the prefrontal regions, with the single EEG channel derived from the difference between Fp1 and Fp2 according to the International 10/20 system, and a reference electrode placed at Fpz. Data were continuously digitized at a sampling frequency of 500 Hz. EEG preprocessing was performed using a standardized pipeline, including artifact correction using ICA, band-pass filtering, and automatic artifact rejection.

Power spectrum and frequency bands

The EEG power spectrum was obtained through the fast Fourier transform of the EEG signals within a 4-s window, using a Hamming window. Power spectral density was calculated from the frontal channel (Fp1–Fp2) and transformed to dB(µV²/Hz) using 10·log10 transformation, for Delta (0.5–4 Hz), Theta (4–7 Hz), Alpha (8–15 Hz), Beta (16–31 Hz), and lower Gamma (32–45 Hz) frequency bands.

Signal processing and EEG biomarkers

The EEG signal was decomposed into multiple components using harmonic analysis mathematical models, as previously described.^22,24 In summary, the Neurosteer signal-processing algorithm analyzes EEG data using a time/frequency wavelet-packet analysis. This analysis, previously conducted on a separate dataset of EEG recordings, identified an optimal orthogonal basis decomposition from a large collection of wavelet packet atoms, optimized for that set of recordings using the Best Basis algorithm.²⁵ This basis results generated a new representation of 121 optimized components, each consisting of time-varying fundamental frequencies and their harmonics.

The EEG features presented here are produced by a secondary layer of ML applied to labeled datasets (previously gathered by Neurosteer), to derive several linear combinations. Specifically, the EEG feature L1 was calculated using the linear discriminant analysis (LDA) technique,²⁶ which is designed to identify an optimal linear transformation that maximizes class separability. L1 was computed using previously derived weight matrices trained on independent datasets and applied to the single-channel EEG data in the present study without retraining, to avoid overfitting. L1 has previously been validated as a marker of working memory load, showing clear separation between 1-back and 3-back conditions and sensitivity to subtle cognitive state changes in healthy subjects.²⁷

Auditory battery

The cognitive tasks were presented as part of an auditory battery consisting of pre-recorded tasks: musical detection, musical n-back, and resting state tasks, as outlined in prior studies.^22,23 The auditory battery was presented during the single-channel EEG recordings.

The detection task comprised sequences of melodies performed by a violin, trumpet, and flute. In the easier level (level 1), melodies lasted 3 s and were repeated throughout the block, each appearing 5–6 times and interspersed with 10–18 s of silence. In the more difficult level (level 2), melodies were shortened to 1.5 s and the three instruments were intermixed within a single block. Each trial included 6–8 melodies, with 8–14 s of silence between them, and the target melody appeared 2–3 times. Instructions were delivered at the beginning of each block in a simplified manner, encouraging participants to focus attention on specific auditory features and, when possible, to imagine responding. For example, participants were asked to try to notice when a specific instrument (e.g., flute) was played and, if able, to imagine or attempt lifting a finger in response.

The n-back task included a sequence of melodies played by different instruments. In line with the patients’ clinical condition, instructions were delivered in a simplified manner, encouraging participants to attend to the auditory stimuli and, when possible, to imagine or attempt responding, or lifting a finger. In the 0-back condition, participants were instructed to respond to each melody played (n = 0). This condition consisted of a single 90-s block comprising nine trials, with each melody lasting 1.5 s and followed by 6–11 s of silence. In the 1-back condition, participants were instructed to respond when a melody repeated consecutively (n = 1). This condition included two 90-s blocks, each with 12–14 trials. Melodies lasted 1.5 s and were followed by 4–6 s of silence, with approximately 30%–40% of trials containing target repetitions. In the 2-back condition, participants were instructed to respond when a melody matched the one presented two steps earlier (n = 2). This condition consisted of a single 45-s block with 12 trials. Melodies lasted 1.5 s, followed by 4–6 s of silence, with approximately 25% of trials containing target repetitions.

Resting-state tasks were interspersed within the auditory battery to provide baseline measurements. The participants were instructed to allow their minds to wander for 60 s in each block.

Cognitive load levels were defined based on task difficulty: Rest was categorized as load level 0. Detection (level 1) and 0-back tasks were categorized as load level 1. Detection (level 2) and 1-back tasks were categorized as load level 2. The 2-back task was categorized as load level 3, representing the highest working memory demand.

Statistical analysis

Each biomarker (MMN, spectral power, and L1) was analyzed separately according to its modality and experimental context, as detailed below. For the DoC group, change in behavioral outcome was quantified as the difference in total CRS-R score between post-intervention and baseline. MMN amplitude and latency were summarized at baseline and post-intervention, and within-participant MMN presence (deviant vs standard difference) was assessed using the waveform comparison procedure implemented automatically using WinEEG (Mitsar Co. Ltd) within the pre-specified MMN window (50–250 ms post-stimulus). Given the pilot nature of the study and the limited sample size, analyses were primarily within-subject and descriptive, with emphasis on effect direction and inter-individual variability.

For healthy controls, task-related modulation of EEG biomarkers was tested using within-subject models with task/cognitive load level as the repeated factor. Specifically, the effect of cognitive load level (0–3) and the CRS-R segment on L1 activity was evaluated using a linear mixed-effects model, followed by post hoc pairwise comparisons. Cognitive load levels were defined based on task difficulty as previously described.^22,23 Rest and passive listening were categorized as load level 0. Detection (level 1) and 0-back tasks were categorized as load level 1. Detection (level 2) and 1-back tasks were categorized as load level 2. The 2-back task was categorized as load level 3, representing the highest working memory demand.

For both groups, associations between MMN amplitude and frontal biomarkers (spectral band power and L1) were assessed using Pearson correlation coefficients, reported with corresponding p-values. The significance threshold for all analyses was set at p < 0.05. Post hoc analyses were conducted using the Benjamini–Hochberg correction²⁸ to control for false discovery rates following significant main effects or interactions. All statistical analyses were performed using Python’s Statsmodels library.²⁹

Results

DoC patients

Six chronic DoC patients clinically classified as either VS/UWS or MCS were included in the study (see Table 1 for full demographic details). The patients constituted a pilot sample recruited for feasibility assessment of the intervention protocol. All patients included completed the full 10-session tDCS protocol. No adverse events or dropouts occurred.

To capture treatment-induced changes, we analyzed three complementary biomarkers: MMN, reflecting sensory discrimination; Beta oscillatory activity, associated with attentional processes; and L1, an ML-derived biomarker of cognitive engagement. The full results are presented in Tables 2 and 3 and Figures 1 –4.

Table 2.

MMN values of DoC patients before and after tDCS treatment.

Patient	MMN amplitude pre tDCS	MMN amplitude post tDCS	MMN latency pre tDCS	MMN latency post tDCS	MMN p-value pre tDCS	MMN p-value post tDCS	CRS-R score Δ
P 11	−0.81	−2.583	290	240	0.2808	0.0952	1
P 12	−1.095	−0.934	92	148	0.034	0.0197	0
P 13	−0.872	−0.032	102	204	0.18	0.0003	2
P 14	−0.62	−0.764	282	108	0.0548	0.0034	0
P 15	0.2	0.3	166	214	0.45	0.85	0
P 16	−0.3	−0.6	228	248	0.02	0.06	0

The table presents MMN amplitude (in µV), latency (in ms), and p-values for pre- and post-tDCS measurements, along with the difference in CRS-R scores (Δ). Negative MMN amplitudes reflect the polarity of the deviant-standard difference waveform. p-Values reflect statistical significance of the deviant-standard tone difference within the MMN time window.

CRS-R, Coma Recovery Scale–Revised; DoC, disorders of consciousness; MMN, mismatch negativity; tDCS, transcranial direct current stimulation.

Table 3.

Mean frontal EEG activity across standard frequency bands and L1 feature for each patient before and after the 10-day tDCS intervention.

Patient	Pre/post tDCS	Delta	Theta	Alpha	Beta	Gamma	L1
P 11	Post	2.90	−6.09	−10.89	−9.83	−9.50	65.59
P 11	Pre	1.88	−9.63	−15.31	−15.89	−15.70	59.70
P 12	Post	−1.64	−10.75	−19.30	−22.85	−24.32	54.82
P 12	Pre	−9.63	−18.16	−24.16	−26.91	−27.89	44.88
P 13	Post	2.22	−8.18	−16.69	−19.94	−20.55	55.61
P 13	Pre	1.55	−10.06	−21.10	−26.89	−28.07	56.26
P 14	Post	2.88	−7.37	−15.12	−20.65	−24.30	59.09
P 14	Pre	0.01	−9.73	−10.63	−9.48	−8.81	48.91
P 15	Post	−7.07	−16.10	−20.32	−21.96	−22.77	43.89
P 15	Pre	4.25	−10.18	−16.21	−17.41	−17.57	48.07
P 16	Post	7.41	−6.49	−14.55	−18.37	−18.77	51.71
P 16	Pre	5.83	−7.82	−17.03	−22.71	−22.73	62.32

Values reflect averages computed from single-channel EEG recordings.

tDCS, transcranial direct current stimulation.

Figure 1.

MMN results for each DoC patient. (a) MMN amplitude (µV). (b) MMN latency (ms), comparing pre-tDCS (blue) and post-tDCS (orange) measurements. (c) Topographic maps illustrating MMN responses for each patient, with the upper row representing pre-tDCS and the lower row post-tDCS results.

Figure 2.

Individual frontal PSD profiles of single-channel EEG in DoC patients before and after tDCS. Each panel represents a single patient, showing baseline (Pre, solid line) and post-intervention (Post, dashed line) PSD across the 1–50 Hz frequency range. The plots illustrate changes in spectral profiles observed following the tDCS intervention.

Figure 3.

L1 activity for each patient before (blue) and after (red) tDCS intervention. Bars represent mean L1 activity with error bars indicating standard deviation. The results show inter-individual variability, with some patients demonstrating increased L1 activity following tDCS, while others exhibit a decrease.

Figure 4.

Pearson correlations presenting the relationship between EEG activity and MMN amplitude during the post-tDCS CRS-R assessment in DoC patients. Negative relationship between MMN amplitude and Beta activity (right) and L1 activity (left).

CRS-R results

Out of the six patients, two (P11 and P13, both diagnosed as VS/UWS) demonstrated improvements in CRS-R scores following the 10-day tDCS intervention (see Table 2). The remaining patients showed no change in behavioral assessment scores. Given the pilot nature of the study, analyses focused on total CRS-R scores, and detailed subscale-level behavioral changes and diagnostic transitions were not systematically assessed.

MMN results

At baseline, two patients (P12 and P16) exhibited a significant MMN response, indicated by a significant peak-negativity difference between odd and standard tones (i.e., MMN amplitudes; see Table 2, p < 0.05). Following the tDCS intervention, three patients (P12, P13, and P14) showed a significant odd–standard difference suggesting the presence of a significant MMN response. This result was observed in both patients classified as MCS, as well as in one patient classified as VS/UWS. Interestingly, one patient (P14, diagnosed as MCS) exhibited a difference in MMN amplitude and a marked decrease in MMN latency following the intervention, without a corresponding improvement in CRS-R scores.

EEG features

The distribution of EEG activity across frequencies (1–50 Hz), measured before and after tDCS for each patient, is presented in Figure 2, and the corresponding L1 activity is summarized in Figure 3. In several patients, EEG activity increased following the intervention compared to baseline. Table 3 summarizes mean values of EEG features before and after tDCS. Patients P11 and P12 exhibited increases in all frequency bands and L1 activity. P13 and P16 also demonstrated increases in all spectral bands, while showing decreases in L1 activity. In contrast, P15 showed decreases across all frequency bands and L1 activity after the intervention. P14 displayed a more mixed response, with increases in Delta, Theta, and L1 activity, while Alpha, Beta, and Gamma activity decreased following tDCS. These findings highlight the inter-individual variability in EEG responses to tDCS, with some patients showing consistent increases across all bands, while others demonstrated more complex patterns of change, particularly in higher frequency bands.

Lower Beta and L1 activity during the CRS-R clinical assessment was significantly associated with higher MMN amplitude post-tDCS (r = −0.83, p = 0.039 and r = −0.85, p = 0.03 for Beta and L1, respectively; see Figure 4). No other significant correlations were found between MMN values and EEG features.

Healthy controls

Ten healthy participants were enrolled in the study as healthy controls.

CRS-R results

As expected, all healthy participants attained the maximal CRS-R score, indicating consistently normal responses across the healthy cohort.

MMN results

The MMN results for control participants are presented in Figure 5. MMN amplitude results show clear MMN responses in most participants. Eight out of the 10 participants demonstrated statistically significant MMN responses (p < 0.05), indicating reliable detection of deviant auditory stimuli. The MMN latency values in the healthy population generally fall within the 100–250 ms range.

Figure 5.

MMN results in healthy controls. (a) MMN amplitude. (b) MMN latency across tasks. (c) Topographic EEG maps illustrating MMN mean responses in healthy participants.

EEG features

Among all EEG features, L1 exhibited a significant overall effect of cognitive load across conditions, characterized by lower values during rest compared to task conditions (β = −4.55, SE = 1.29, z = −3.53, p < 0.001), with no significant differences observed across increasing cognitive load levels (see Figure 6). Post hoc analysis with Benjamini–Hochberg²⁸ correction indicated higher L1 activity during high-load tasks (cognitive load level 3) compared to resting state (p = 0.026, Cohen’s D = 1.618, CI = (2.221, 6.971)). No other task-related main effects were observed for the remaining frequency bands.

Figure 6.

L1 activity across cognitive load conditions (cognitive load levels 1–3, rest) and during the CRS-R test. Boxplots illustrate the distribution of L1 activity across conditions, showing lower activity at rest compared to task conditions.

Consistent with the patient group, Beta and L1 activity showed a negative correlation with MMN amplitude (see Figure 7). Specifically, higher Beta activity was associated with lower MMN amplitude during tasks with the medium cognitive load level (cognitive load 2: r = −0.67, p = 0.048). Similarly, L1 activity was negatively correlated with MMN amplitude across multiple conditions: tasks with medium cognitive load (cognitive load 2; r = −0.70, p = 0.036); tasks with high cognitive load (cognitive load 3: r = −0.68, p = 0.043), and during the CRS-R clinical test (r = −0.68, p = 0.044). No other significant correlations were found.

Figure 7.

Pearson correlations presenting the relationship between EEG activity and MMN amplitude in healthy controls. (a) Beta activity during medium cognitive load tasks as a function of MMN amplitude. (b) L1 activity during medium cognitive load tasks. (c) L1 activity during high cognitive load tasks. (d) L1 activity during the CRS-R clinical test.

Discussion

This pilot study examined the feasibility and sensitivity of a compact single-channel EEG to detect neurophysiological reactivity changes following a 10-day tDCS intervention in DoC. Feasibility was evaluated based on protocol adherence (completion of the 10-session tDCS intervention), tolerability (absence of adverse events), and the successful acquisition of analyzable EEG data across sessions. The electrophysiological measures examined, including MMN, frontal oscillatory activity, and the ML feature L1, captured changes following the intervention, even when CRS-R remained unchanged. These preliminary findings are consistent with prior work on tDCS responsiveness in DoC and suggest that EEG biomarkers may provide a sensitive window into residual cognition beyond behavioral assessment alone.^8,30 In particular, MMN and large-scale EEG activity have been shown to reflect covert neural processing and to relate to clinical outcomes in DoC patients.

Following tDCS treatment, MMN responses became more pronounced in three patients post-intervention. In two cases, changes were observed without parallel behavioral improvement. This divergence highlights the potential for MMN measures to reflect neurophysiological changes not immediately captured by clinical assessments. This aligns with prior research suggesting that ERP measures like MMN are sensitive to subtle cognitive changes in DoC patients.¹⁶ These findings may reflect modulation of prefrontal networks involved in auditory discrimination,^3,13 consistent with prior evidence linking frontal engagement to sensory processing and cognitive function. Given its association with automatic auditory change detection, MMN may provide complementary information regarding neural responsiveness in DoC. Prior tDCS studies report greater behavioral responsiveness in MCS compared to VS/UWS patients.^8–10 In the present sample, increases in CRS-R scores were observed in two VS/UWS patients, whereas no behavioral improvement was detected in MCS patients. In contrast, electrophysiological changes, including increased MMN significance, were observed in both MCS patients and in one VS/UWS patient, consistent with reports of neurophysiological reactivity in behaviorally nonresponsive patients.^16,30 Given the small sample size, these findings should be interpreted with caution and do not allow for conclusions regarding differential responsiveness between diagnostic groups.

Significant associations between MMN amplitude and frontal Beta and L1 in patients mirror healthy-control patterns, whereby higher Beta and L1 correlated with lower MMN amplitudes. This relationship may reflect the dynamic interplay between attention, cognitive load, and sensory processing. Higher Beta and L1 alongside reduced MMN may indicate load-dependent competition between sustained engagement and pre-attentive deviance detection, consistent with prior studies demonstrating that MMN amplitude can be modulated under varying cognitive demands in healthy individuals.³¹ At the same time, previous work suggests that increased cognitive engagement may enhance neural sensitivity to auditory deviance under certain task conditions, further indicating that MMN modulation depends on the balance between attentional allocation and task demands.³² These findings are particularly relevant given the established association of Beta activity with active cognitive processes, such as sustaining attention and working memory,³³ and thus might provide a complementary perspective on the cognitive state of DoC patients.

Furthermore, L1 activity demonstrated a main effect of task in healthy controls, with significantly higher activity observed during increased cognitive load (cognitive load level 3) compared to rest, further supporting its relevance as a marker of cognitive engagement.^20,27 Previous work showed greater sensitivity of L1 compared to Theta in distinguishing subtle variations in working memory load, including differentiation between 1-back and 2-back conditions (Maimon, Molcho and Intrator, 2021) and between high- and low-importance task interruptions in healthy individuals.²⁰ L1 was derived using LDA,²⁶ designed to identify an optimal linear transformation that maximizes class separability. LDA has been effectively utilized in DoC research to distinguish between patient outcomes. For instance, a study applied LDA to EEG features to predict the emergence from prolonged DoC, demonstrating its utility in classifying patient prognoses.³⁴ Another study utilized LDA to EEG data to assess the effects of tDCS on patients with DoC, aiming to identify EEG markers that correlate with improvements in consciousness levels.³⁵ The use of LDA enables extraction of task-sensitive features, supporting its applicability in clinically heterogeneous populations. Notably, L1 was extracted from datasets different from the ones analyzed in the current study, to avoid overfitting the data. Consequently, the weight matrices previously determined were applied to transform the data acquired in the present study. L1 may, therefore, offer an objective indicator of residual cognitive function and help identify subtle signs of awareness.

A key observation was that EEG features provided insights that behavioral assessments might overlook.^16,30 This is particularly relevant for tracking cognitive fluctuations in minimally responsive patients, where EEG measures such as MMN, L1, and Beta could serve as complementary tools for evaluating treatment effects. Moreover, EEG-based neural reactivity has been shown to predict long-term survival in DoC patients, reinforcing its clinical significance.³⁶

Importantly, some of these findings were obtained using a single-channel EEG system integrated into routine clinical practice, supporting its feasibility for bedside monitoring.³⁷ In the present study, this approach enabled the detection of neurophysiological changes, even in the absence of behavioral improvement. This suggests that simplified EEG configurations may be sufficient to capture clinically relevant neural dynamics in DoC. Given its minimal setup and accessibility, single-channel EEG may provide a practical tool for repeated monitoring of treatment-related changes, particularly in settings where full-cap EEG is not feasible. This approach could open the door for adaptive neuromodulation strategies. A recent study proposed a closed-loop framework in which tDCS is triggered based on EEG-derived vigilance states in MCS patients, ensuring that stimulation is applied during periods of optimal neural receptivity.³⁸ This real-time adaptive model underscores the potential of EEG biomarkers to personalize neuromodulation in DoC. Future work should directly compare single-channel and multi-channel recordings to establish their relative diagnostic and prognostic value and test single-channel EEG within closed-loop protocols to advance precision interventions. While showing promising initial results, additional research is needed to address the limitations encountered in this study. The small sample size should be taken into account when considering the results, as it may limit statistical power and increase the risk of both negatives and false positives. Interpretation of the results is also limited by the known variability in CRS-R measurements, and etiological heterogeneity of the patients which may also contribute to variability in electrophysiological responsiveness. An additional limitation is the diagnostic imbalance of the sample, with the majority of participants being males classified as VS/UWS, which limits subgroup interpretation. This concern is especially relevant in EEG research where effect sizes may vary substantially between individuals.³⁹ Long-term clinical outcomes were not evaluated, limiting conclusions regarding the durability of the observed changes. Future research with larger sample sizes is warranted to improve statistical robustness and further validate the diagnostic and prognostic value of the EEG features, particularly their ability to monitor long-term treatment effects in DoC patients. The absence of a sham-tDCS control condition limits the generalizability and should be addressed in future controlled trials. Additionally, while the single-channel EEG approach offers a practical and scalable alternative to full-cap EEG, further research is needed to validate the biomarkers in DoC populations and diverse clinical settings to fully establish their clinical utility. Future studies should incorporate larger patient cohorts and extended follow-up periods. Direct comparisons with standard EEG will be needed to establish monitoring and prognostic accuracy. Altogether, these findings support the development of scalable EEG biomarkers as a foundation for precision neuromodulation strategies in DoC.

Conclusion

This pilot work shows that a portable single-channel EEG can capture functional neural changes elicited by prefrontal tDCS in individuals with DoC. The integration of MMN responses, frontal oscillatory activity, and the ML feature L1 provides converging markers of residual cognitive processing that extend beyond behavioral scales. These results point toward the value of compact EEG tools for routine bedside monitoring, while larger controlled studies are needed to establish reliability, optimize stimulation protocols, and determine their role in rehabilitation strategies.

Supplemental Material

sj-docx-1-tan-10.1177_17562864261456536 – Supplemental material for Single-channel EEG captures tDCS-induced changes in neural reactivity: a pilot study in disorders of consciousness patients and healthy controls

Supplemental material, sj-docx-1-tan-10.1177_17562864261456536 for Single-channel EEG captures tDCS-induced changes in neural reactivity: a pilot study in disorders of consciousness patients and healthy controls by Lior Molcho, Neta B. Maimon, Nathan Intrator, Jeremy Barron, Efraim Jaul, Roxane S. Hoyer, Steven Laureys and Oded Meiron in Therapeutic Advances in Neurological Disorders

Footnotes

Acknowledgements

The authors express their sincere gratitude to the study participants and the supportive staff for their contributions to this research.

Declarations

ORCID iD

Lior Molcho

Supplemental material

Supplemental material for this article is available online.

References

Laureys

Celesia

Cohadon

, et al. Unresponsive wakefulness syndrome: a new name for the vegetative state or Apallic syndrome. BMC Med 2010; 8: 68.

Giacino

Ashwal

Childs

, et al. The minimally conscious state: definition and diagnostic criteria. Neurology 2002; 58(3): 349–353.

Sitt

King

El Karoui

, et al. Large scale screening of neural signatures of consciousness in patients in a vegetative or minimally conscious state. Brain 2014; 137(Pt 8): 2258–2270.

Giacino

Kalmar

Whyte

The JFK Coma Recovery Scale-Revised: measurement characteristics and diagnostic utility. Arch Phys Med Rehabil 2004; 85(12): 2020–2029.

Nitsche

Cohen

Wassermann

, et al. Transcranial direct current stimulation: state of the art 2008. Brain Stimul 2008; 1(3): 206–223.

Zheng

Xia

, et al. Behavioral effects in disorders of consciousness following transcranial direct current stimulation: a systematic review and individual patient data meta-analysis of randomized clinical trials. Front Neurol 2022; 13: 940361.

Zhao

Jia

, et al. Effect of transcranial direct current stimulation for patients with disorders of consciousness: a systematic review and meta-analysis. Front Neurosci 2023; 16: 1081278.

Thibaut

Bruno

M-A

Ledoux

, et al. TDCS in patients with disorders of consciousness: Sham-controlled randomized double-blind study. Neurology 2014; 82(13): 1112–1118.

Angelakis

Liouta

Andreadis

, et al. Transcranial direct current stimulation effects in disorders of consciousness. Arch Phys Med Rehabil 2014; 95(2): 283–289.

10.

Zhang

Song

, et al. Transcranial direct current stimulation in patients with prolonged disorders of consciousness: combined behavioral and event-related potential evidence. Front Neurol 2017; 8; 620.

11.

Brunoni

Vanderhasselt

MA.

Working memory improvement with non-invasive brain stimulation of the dorsolateral prefrontal cortex: a systematic review and meta-analysis. Brain Cogn 2014; 86: 1–9.

12.

Näätänen

The mismatch negativity: a powerful tool for cognitive neuroscience. Ear Hear 1995; 16: 6–18.

13.

Garrido

Kilner

Stephan

, et al. The mismatch negativity: a review of underlying mechanisms. Clin Neurophysiol 2009; 120(3): 453–463.

14.

Näätänen

Sussman

Salisbury

, et al. Mismatch negativity (MMN) as an index of cognitive dysfunction. Brain Topogr 2014; 27(4): 451–466.

15.

Carrière

Mortaheb

Raimondo

, et al. Neurophysiological correlates of a single session of prefrontal tdcs in patients with prolonged disorders of consciousness: a pilot double-blind randomized controlled study. Brain Sci 2020; 10(7): 469.

16.

Meiron

Barron

David

, et al. Neural reactivity parameters of awareness predetermine one-year survival in patients with disorders of consciousness. Brain Inj 2021; 35(4): 453–459.

17.

Miller

Lundqvist

Bastos

AM.

Working memory 2.0. Neuron 2018; 100(2): 463–475.

18.

Scarpino

Lolli

Hakiki

, et al. EEG and Coma Recovery Scale-Revised prediction of neurological outcome in disorder of consciousness patients. Acta Neurol Scand 2020; 142(3): 221–228.

19.

Maimon

Molcho

Intrator

, et al. Novel single-channel EEG features correlate with working memory load. In: 2021—2021 IEEE 3rd global conference on life sciences and technologies (LifeTech), Nara, Japan: IEEE, 2021, pp. 471.

20.

Bolton

Te’Eni

Maimon

, et al. Detecting interruption events using EEG. In: 2021 IEEE 3rd global conference on life sciences and technologies (LifeTech), Nara, Japan: IEEE, 2021, pp. 33–34.

21.

Giacino

Fins

Laureys

, et al. Disorders of consciousness after acquired brain injury: the state of the science. Nat Rev Neurol 2014; 10(2): 99–114.

22.

Molcho

Maimon

Regev-Plotnik

, et al. Single-channel EEG features reveal an association with cognitive decline in seniors performing auditory cognitive assessment. Front Aging Neurosci 2022; 14: 773692.

23.

Molcho

Maimon

Hezi

, et al. Evaluation of Parkinson’s disease early diagnosis using single-channel EEG features and auditory cognitive assessment. Front Neurol 2023; 14: 1273458.

24.

Maimon

Bez

Drobot

, et al. Continuous monitoring of mental load during virtual simulator training for laparoscopic surgery reflects laparoscopic dexterity. A comparative study using a novel wireless device. Front Neurosci 2022; 15: 694010.

25.

Coifman

Wickerhauser

MV.

Entropy-based algorithms for best basis selection. IEEE Trans Inf Theory 1992; 38: 713–718.

26.

Hastie

Buja

Tibshirani

Penalized discriminant analysis. Ann Statist 1995; 23(1): 73–102.

27.

Maimon

Molcho

Intrator

, et al. Single-channel EEG features during n-back task correlate with working memory load. arXiv preprint. arXiv:2008.04987.

28.

Benjamini

Yoav

Controlling the false discovery rate—a practical and powerful approach to multiple testing. J R Stat Soc Series B Methodol 1995; 57(1): 289–300.

29.

Seabold

Perktold

Statsmodels: econometric and statistical modeling with python. In: In Proceedings of the 9th python in science conference, (SciPy 2010), Austin, TX, 2010, p. 61.

30.

Chennu

Annen

Wannez

, et al. Brain networks predict metabolism, diagnosis and prognosis at the bedside in disorders of consciousness. Brain 2017; 140(8): 2120–2132.

31.

Fitzgerald

Todd

Making sense of mismatch negativity. Front Psychiatry 2020; 11: 468.

32.

Näätänen

Alho

Mismatch negativity—the measure for central sound representation accuracy. Audiol Neurootol 1997; 2(5): 341–353.

33.

Gola

Magnuski

Szumska

, et al. EEG beta band activity is related to attention and attentional deficits in the visual performance of elderly subjects. Int J Psychophysiol 2013; 89(3): 334–341.

34.

Siegert

Narayanan

Turner-Stokes

Prediction of emergence from prolonged disorders of consciousness from measures within the UK rehabilitation outcomes collaborative database: a multicentre analysis using machine learning. Disabil Rehabil 2023; 45(18): 2906–2914.

35.

Zhang

Han

, et al. Multidimensional assessment of electroencephalography in the neuromodulation of disorders of consciousness. Front Neurosci 2022; 16: 903703.

36.

Edlow

Claassen

Schiff

, et al. Recovery from disorders of consciousness: mechanisms, prognosis and emerging therapies. Nat Rev Neurol 2021; 17(3): 135–156.

37.

Gosseries

Schnakers

Ledoux

, et al. Automated EEG entropy measurements in coma, vegetative state/unresponsive wakefulness syndrome and minimally conscious state. Funct Neurol 2011; 26(1): 25–30.

38.

Martens

Ibáñez-Soria

Barra

, et al. A novel closed-loop EEG-tDCS approach to promote responsiveness of patients in minimally conscious state: a study protocol. Behav Brain Res 2021; 409: 113311.

39.

Button

Ioannidis

JPA

Mokrysz

, et al. Power failure: why small sample size undermines the reliability of neuroscience. Nat Rev Neurosci 2013; 14: 365–376.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.03 MB

0.00 MB