Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
Named Entity Recognition in Track Circuits Based on Multi-Granularity Fusion and Multi-Scale Retention Mechanism
Previous Article in Journal
Robust Regression and Redundant Measurement Noise Estimation Adaptive Filtering for Localization in Urban Environments
Previous Article in Special Issue
An Analysis of Traditional Methods and Deep Learning Methods in SSVEP-Based BCI: A Survey
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Electronics
Volume 14
Issue 5
10.3390/electronics14050827
electronics-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Brain-Computer Interface Based Engagement Feedback in Virtual Reality Rehabilitation: Promoting Motor Cortex Activation

by
Hyunmi Lim
,
Bilal Ahmed
and
Jeonghun Ku
*
Department of Biomedical Engineering, College of Engineering, Keimyung University, Daegu 42601, Republic of Korea
*
Author to whom correspondence should be addressed.
Electronics 2025, 14(5), 827; https://doi.org/10.3390/electronics14050827
Submission received: 9 January 2025 / Revised: 13 February 2025 / Accepted: 17 February 2025 / Published: 20 February 2025
(This article belongs to the Special Issue EEG Analysis and Brain–Computer Interface (BCI) Technology)
Browse Figures
Figure 1
<p>Procedure diagram for experiment.</p> "> Figure 2
<p>The experimental setting and conditions. (<b>a</b>) A participant wearing the EEG device and HMD is performing the target tracking task with the controller in their hand. The EEG data are recorded and processed in real-time, and reveal the neurofeedback state according to whether the participant is attentively staring at the flickering target. (<b>b</b>) Three conditions (NoF, PF, and NF) were present: the NoF condition (upper) provided no feedback on tracking, while the PF and NF (lower) conditions included showering and blooming feedback with sound. (<b>c</b>) An example of a target trajectory. The target moves along an invisible circle-shaped track, and transitions into another adjacent cloned circle-shaped track when it changes direction. (<b>d</b>) The EEG electrodes are located on areas where the HMD straps do not pass, so that the EEG signal has minimal interference from body movements.</p> "> Figure 3
<p>Power spectrum density corresponding to the attentive and inattentive states in channels O1 and O2. The SSVEP peaks were observed in both states, but those in the attentive state were stronger.</p> "> Figure 4
<p>The performance outcomes. The tracking error (<b>left</b>) represents the distance between the target and the cursor, and the feedback ratio (<b>right</b>) represents the ratio of the amount of feedback provided during the tracking period. The error bar represents the standard error.</p> "> Figure 5
<p>The EEG outcomes. The SSVEP power (<b>left</b>), engagement index (<b>middle</b>), and Mu suppression on C3 (<b>right</b>) are the strongest in the NF condition, which shows significantly different results from the other two conditions. ° represents the trend toward significance; * and ** represent <span class="html-italic">p</span> &lt; 0.05 and <span class="html-italic">p</span> &lt; 0.01, respectively, after Bonferroni correction; and the error bars represent the standard error.</p> "> Figure 6
<p>The correlation results between the SSVEP (upper) or engagement index (lower) and Mu suppression on C3. The SSVEP showed a negative correlation tendency with Mu suppression in all three conditions, which was particularly significant in the NF condition, while the engagement index did not show any pattern. ° represents the trend toward significance, and * represents <span class="html-italic">p</span> &lt; 0.05.</p> ">
Versions Notes

Abstract

:
Maintaining optimal levels of engagement during rehabilitation training is crucial for inducing neuroplasticity in the motor cortex, which directly influences positive rehabilitation outcomes. In this research article, we propose a virtual reality (VR) rehabilitation system that incorporates a steady-state visual evoked potential (SSVEP) paradigm to provide engagement feedback. The system utilizes a flickering target and cursor to detect the user’s engagement levels during a target-tracking task. Eighteen healthy participants were recruited to experience three experimental conditions: no feedback (NoF), performance feedback (PF), and neurofeedback (NF). Our results reveal significantly greater Mu suppression in the NF condition compared to the other conditions. However, no significant differences were observed in performance metrics, such as tracking error, among the three conditions. The amount of feedback between the PF and NF conditions also showed no substantial difference. These findings suggest the efficacy of our SSVEP-based engagement feedback paradigm in stimulating motor cortex activity during rehabilitation. Consequently, we conclude that neurofeedback, based on the user’s attentional state, proves to be more effective in promoting motor cortex activation and facilitating neuroplastic changes. This research highlights the potential of integrating VR rehabilitation with an engagement feedback system for successful rehabilitation training.

1. Introduction

Patients with impaired motor function resulting from neuronal injury require appropriate rehabilitation. Successful rehabilitation is usually accompanied by neuronal reorganization, called brain plasticity, also referred to as neuroplasticity, defined as the brain’s intrinsic ability to rearrange its neural pathways in response to learning, damage, or rehabilitation interventions. Neurological injuries recover when the brain adapts to new synaptic connections and strengthens existing ones [1]. Appropriate rehabilitation paradigms are an important consideration for effective brain plasticity [2], and offer enriched training environments and stimuli [3].
Constraint-induced movement therapy (CIMT) is a prominent example of neuroplasticity; it is widely used as a rehabilitation technique for stroke patients to induce cortical reorganization, enhance motor control, and restore functional abilities. Forcing patients to only depend on their affected limb for movement in CIMT has shown significant improvement in upper limb function by promoting neural rewiring in the sensorimotor cortex [4,5].
Similarly, the integration of the brain–computer interface (BCI) with rehabilitation enhances neuroplasticity to promote motor recovery by providing real-time feedback on brain activity [6,7]. Successful motor learning that facilitates brain plasticity also depends on the ability to continuously monitor and adjust one’s performance. This process can be supported by a real-time assistance system, such as one that provides visual feedback according to the movements being performed [8]. Such systems may encourage participants to pay more attention to performing their movements, which has been emphasized as an important factor in motor learning [9].
When individuals experience a loss of attention, a reduction in motor cortex activity is observed [10], which interferes with rehabilitation training. The gradual and subtle occurrence of these errors due to attention loss can lead to a series of error events that can result in irreversible false learning, which may reduce the effectiveness of brain plasticity [11]. Therefore, to enhance patients’ rehabilitation outcomes, thereby increasing brain plasticity, sustained attention ability is crucial. Several neuroscience studies have revealed that attention is a crucial regulator of plasticity [12,13], which directly affects how well patients recover from their injuries [14]. Significant efforts have been made to improve patients’ attentional capacity using neurofeedback, which facilitates the ability to regulate patterns of brain signals [15] or brain areas [16] that may affect their rehabilitation outcomes.
Although a performance-based guidance system, which monitors performance/movement errors, is helpful in rehabilitation, people often make errors in performing the movements because of lapses in attention and their ability to self-monitor at any given time [17], which can lead to suboptimal motor learning [18]. However, research suggests that sustained attention is a critical modulator of neuroplasticity, and higher attentional engagement enhances motor learning and rehabilitation results [9].
A clinical study by Ramos et al. demonstrated that BCI-based neurofeedback training enhances motor attention and engagement more than traditional rehabilitation [19]. Another clinical study revealed stronger activation in motor cortical areas and enhanced functional improvement in patients of stroke rehabilitation with higher attentional engagement [20,21] during the movement therapy [22,23].
However, in our general observations, even healthy individuals and patients who have obtained the ability to regulate brain functions cannot perform tasks perfectly, and may experience disruptions in their focus due to distracting factors in their surroundings, which can affect their performance during rehabilitation training [9]. In addition, subtle lapses in attention during training often go unnoticed by patients and their clinicians, which hinders timely intervention and mediation to refocus the patient on the rehabilitation task [24]. It is strongly suggested that an assistive concept that can investigate the patient’s brain state should be used to alert patients whenever there is attentional loss. Attentional lapses delay neuroplasticity and eventually disturb treatment efficiency [25]. On the other hand, conventional feedback methods, despite advancements in rehabilitation feedback technologies, depend on movement/performance-based error correction and cannot take cognitive/attentional engagement changeability, so they are unable to conduct real-time feedback intervention [24,26].
Therefore, a system is needed that can monitor people’s attention during task execution, rather than simply monitor their performance [27,28]. Accordingly, the authors of this research paper developed a neuroplasticity-based rehabilitation model that combines a BCI with real-time monitoring to enhance motor recovery. Our system monitors attentional states in real time using SSVEP-based neurofeedback.
Despite the importance of alerting individuals to regain their attention, few approaches have focused on providing feedback during training. One representative system showed that an alerting attentional state plays a positive role in helping participants to sustainably maintain attention to a task, resulting in performance improvement [29]. While this study showed that only attentional alertness can improve cognitive abilities, no suggested paradigm has been proposed for motor rehabilitation. In terms of motor rehabilitation, several of our previous studies have shown consistent superior motor circuit activity in both healthy participants and patients with stroke during action observation, when the steady-state visual evoked potential (SSVEP) paradigm was embedded. Several brain-computer interface (BCI) studies have employed the SSVEP paradigm to select one command from a set of flickering visual stimuli, each representing a distinct operation in a variety of contexts. These include spelling [30], virtual telephony [31], and robot control [32]. This is because the SSVEP response to the target stimulus is amplified when the participant focuses their attention on the target stimulus among several stimuli, which demonstrates that the SSVEP response is influenced by whether an individual is engaged in focused attention or is distracted [33,34].
Therefore, adopting the SSVEP paradigm for action observation has been proven to be a superior and effective method for monitoring the user’s attentiveness [35], allowing for effective and robust monitoring of the participant’s attentive state and providing feedback to encourage them to maintain attention during action observation training, resulting in superior motor cortex facilitation [34,36,37,38].
In this study, we apply our attentional state monitoring paradigm used in action observation to a BCI–virtual reality (VR) neurofeedback system. This approach aims to encourage patients to maintain their attention on the rehabilitation task by providing feedback regarding patients’ focus on the motor task during the training, and to investigate whether it could induce stronger facilitation of motor cortex activity. To accomplish this goal, we applied an SSVEP paradigm to monitor patients’ attentional state to the task and integrated it into a virtual reality training system for upper limb rehabilitation. This system was validated by showing motor cortex activity patterns in electroencephalogram (EEG) signals from healthy participants.

2. Materials and Methods

2.1. Participants

We recruited 18 healthy right-handed adults (8 men and 10 women) in this study. The mean age of the participants was 27.33 years (±2.45 years). A standardized tool for assessing laterality, the Edinburgh Handedness Inventory [39], was used to confirm hand dominance. All the participants had no history of neurological disorders or prior experience with BCI systems, as shown in demographic Table 1. Healthy participants were selected as a standard group to investigate the system’s effectiveness under controlled conditions that aligned with prior neurorehabilitation studies. Likewise, this study follows the selection criteria for participants from previous studies [38,40], where healthy participants are investigated first, to check the system’s performance, before applying it to clinical populations. Furthermore, the results for the healthy participants are nearly identical [41,42].

2.2. Ethics Statement

The studies involving humans were approved by the Institutional Review Board of Keimyung University (IRB number: 40525-202106-HR-024-04). The studies were conducted following the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

2.3. Experimental Procedures

When participants arrived at the location of the experiments, they were asked to sit down on a chair in front of a desk with a monitor, and they put on an EEG cap. After checking the EEG signal quality, they put a Head-Mounted Display (HMD) on their head over the EEG cap, and we also ensured that the EEG signal quality was maintained.
First, participants were given a 1 min rest recording, during which they viewed a crosshair in the center of the screen. They were asked to do an example task to familiarize themselves with VR tasks by using VR controllers, while their EEGs were also visually monitored for any interference during the performance. A classifier design session was conducted for 3 min to design the BCI classifier.
When everything was prepared, participants were asked to perform the three tracking tasks. Tracking conditions of no feedback (NoF), performance feedback (PF), and neurofeedback (NF) were presented consecutively for each participant, in a randomized order to minimize order effects and potential bias; each task lasted approximately 5 min. The order of conditions was randomly assigned using a computer-generated randomization sequence, to ensure that each participant was presented with a unique condition order. This simple order randomization approach helped to prevent systematic biases in the task presentation. The tracking task required participants to follow a moving target continuously for 5 min, without interruptions. Since the task was continuous, there was no fixed number of trials. The entire experimental session lasted 20 min (Figure 1).

2.4. BCI-VR Rehabilitation System and Program

The BCI-VR upper limb rehabilitation system consists of an HMD with controllers (Meta Quest 2; Meta Platforms, Inc., Menlo Park, CA, USA) to provide virtual environments in which the positions and orientations of virtual objects in the virtual environment are aligned perfectly with those of controllers in the physical world, as well as Unity3D plugin supports, an eight-channel EEG signal acquisition device (Versatile EEG 8 channels; Bitbrain, Zaragoza, Spain), and a computer for EEG acquisition and analysis. The EEG data were transferred via the Lab Streaming Layer to OpenVibe (Ver. 3.3.1) software for real-time EEG analysis and classification. The classification results were then transferred to the VR application using the Virtual Reality Peripheral Network (VRPN) protocol. After receiving the classification results, the VR application provided feedback in the virtual environment, so that the user could recognize whether they were attentive (Figure 2a,b).
Participants performed a tracking task while wearing the HMD and EEG caps throughout the session. The objective of the tracking task was to follow a ring-shaped moving target and align it with a ring-shaped cursor on a virtual handheld sprinkler held by a virtual avatar’s hand. This was achieved by the participant holding and moving a VR controller with their dominant hand. The target was positioned within a virtual environment, and moved along an invisible circle-shaped track, with an approximate diameter of 16 cm, in a counterclockwise direction. After some time, the target’s trajectory changed, transitioning into a cloned circle-shaped track adjacent to the previous trajectory circle (Figure 2c). The tracking task was specifically designed to ensure accurate EEG recordings, with the direction of movement along a smooth continuous trajectory to minimize movement and prevent abrupt signal artifacts. The target followed the invisible circular path, with the trajectory randomly changing at an applied interval set to be over 1.5 s on average, ranging between 1.2 and 2.5 s. This adaptability stopped participants from expecting movement shifts, confirming that their attentional engagement was natural rather than based on predictive motor adjustments. Furthermore, the designed system helped reduce spontaneous movements, decreasing unnecessary noise in the EEG data.
Participants received visual (blooming and watering) and auditory feedback simultaneously from the system, which lasted 0.5 seconds per instance when they were determined to attend to the task in the NF condition or when they performed well by making small amounts of errors in the PF condition. The target and cursor were flickered at 15 Hz to enable the efficient classification of their engagement states for the task, particularly with the SSVEP component from electrodes over the visual cortex. EEG data were collected. Data analysis involved the use of time–frequency analysis, and a support vector machine (SVM) algorithm was used to classify the participants’ attentiveness.

2.5. Experimental Conditions

For this study, three conditions were used to examine the effectiveness of the engagement feedback system. 1. Conventional Tracking Condition (NoF): This condition provided no feedback based on brain state and performance. 2. Performance-Based Feedback Tracking Condition (PF): Feedback was given when the distance between the cursor and the target fell below a threshold (3.5 cm in this experiment). 3. Neurofeedback Tracking Condition (NF): Feedback was determined by the user’s brain state, indicating their task engagement level. This was evaluated using a trained classifier based on the SVM algorithm, which analyzed the extent of SSVEP evoked when the user focused on the flickering cursor or target.

2.6. EEG Acquisition and Analysis

EEG data were collected from a semi-dry, 8-channel EEG device (Versatile EEG 8 channels; Bitbrain, Zaragoza, Spain) with a 256 Hz sampling rate. This EEG device uses a saline solution to acquire EEG signals. The electrodes were placed with a standard 10–20 scheme over F3, F4, C3, C4, P4, P4, O1, and O2. AF3 was used as the ground channel, and A2 was used as the reference. The eight channels were selected to circumvent the potential for interference caused by co-motion with the HMD strap, considering that the strap’s movement could result in significant motion artifacts. Furthermore, the leads of the electrodes were also connected to an amplifier over the HMD, to prevent contact with the strap (Figure 2d). A signal quality check was performed before experimenting to ensure that the impedance values of all the channels meet the manufacturer’s criteria, as confirmed by the manufacturer’s software.
For the real-time neurofeedback application, we used an analysis pipeline on the OpenVibe platform (Ver. 3.3.1). The participants experienced a 3 min classifier calibration session to train the BCI system, in which users were asked to look at the target object (the virtual handheld sprinkler held by the virtual avatar). During the session, the ring on the virtual sprinkler was flickered at 15 Hz for 10 s, followed by 10 s of rest. This cycle was repeated 10 times, generating a sufficient training dataset to distinguish between attentive (SSVEP-present) and non-attentive (SSVEP-absent) states from the O1 and O2 electrodes, ensuring reliable real-time engagement detection. The following calibration parameters were used:
  • Intervals of 10 s fixation + 10 s rest (repeated 10 times): this provides a balanced dataset of engaged and non-engaged states, ensuring that the classifier can accurately differentiate attentional shifts.
  • A 3-minute total duration: this minimum duration is optimal for avoiding participant fatigue while gathering a sufficient training dataset [43].
  • A 15 Hz flickering frequency: This was selected due to its reliably eliciting a strong and stable SSVEP response [30].
After acquiring the training EEG dataset containing data corresponding to flickered and non-flickered stimulation, a common spatial pattern (CSP) filtering method was used to optimize the dataset, which can find a linear transformation of the data that makes the signal conditions more distinct [44], and is designed to maximize the difference between the two conditions, so can be used to discriminate the attentive state from the non-attentive state. The CSP method is especially effective in EEG feature extraction and optimally enhances class separability by extending the difference between two conditions (attentive and non-attentive states) [45]. However, independent component analysis is not optimized for extending the variance difference between two conditions, and principal component analysis does not explicitly enhance class separability.
On the other hand, the first 1 s of data from the stimulus onset was removed to compensate for any delay in attention, so that the labeled data were each 9 seconds long, and these were then passed through a CSP filter trainer. The CSP filter consisted of an 8 × 4 matrix (number of input channels × number of spatial filters), while the CSP filter transformed signals from 8 channels into four filtered signals, thus generating four SSVEP features. The CSP method ensures high classification accuracy with preprocessing steps, and maintains real-time preprocessing, making the CSP method an optimal choice for neurofeedback applications.
After the CSP filter design was completed, the EEG data analysis pipeline for classifier design was added, so that the pipeline included bandpass filtering (14.5–15.5 Hz) to extract the SSVEP component, CSP filtering, and then time-based epoching of a 1 s duration. Finally, the classifier was built using a linear kernel support vector machine (SVM) algorithm, with the data passed through the pipeline.
Once the classifier was designed, it was applied to the real-time signal analysis pipeline for the neurofeedback system. It had the same preprocessing pipeline as that for the classifier design, but the classifier was added at the end of the pipeline, and then the classification result would be sent to the neurofeedback system, using the VRPN protocol, every 0.1 s.
For the after-experiment analysis to investigate the differences in motor cortex activities in each condition, the collected EEG was bandpass filtered to extract a spectrum component in the 4–30 Hz frequency range. Furthermore, artifact subspace reconstruction (ASR) [46] and independent component analysis methods [47] were applied. Specifically, ASR was implemented to remove unwanted EEG artifacts (physiological noise components) while sustaining brain-related neural activity. ASR was applied to the data using the following parameters [48,49]:
  • Window length—0.5 s: balances artifact detection efficiency while avoiding unnecessary data.
  • Window overlap—66%: ensures smooth transitions between close EEG segments, minimizing signal discontinuities.
  • Method—Euclidean distance: ensures robustness in detecting large deviations in EEG signals.
  • Standard deviation cutoff—5: removes significant artifacts while preserving needed brain activity.
This ASR configuration was chosen to effectively identify and remove artifacts, while preserving brain-related signals and enabling accurate analysis of participants’ reflected cognitive states without contamination from noisy data.
The power spectrum density for the 1 min rest period and each condition was calculated using Welch’s method, with a 2 s epoch window and 1 s overlap. From the power spectrum, the averaged powers of Mu, beta, theta, and SSVEP were calculated.

2.7. Dependent Variables

2.7.1. Behavioral Measurement

  • Performance
Performance is defined as the distance between the target and the cursor, in cm. In the PF condition, this parameter was used for giving feedback. In addition, as a dependent variable, it is used to estimate the participants’ behavioral performance regarding how accurately they performed the tracking tasks for each condition.

2.7.2. EEG Measurements

  • SSVEP Power
In this study, SSVEP power was used to detect whether participants engaged with the task by staring at the target stimuli. Therefore, stronger SSVEP power represents greater staring at the stimuli flickering in the tracking task, which means that more attention was given to performing the task.
  • Mu suppression index
Mu suppression index quantifies motor cortex activation during the performance of tasks. It is defined as the logarithmic ratio of the Mu power of C3 in the task to that in rest. The index was calculated with the equation 10 log 10 ⁡ ( t a s k _ m u / r e s t _ m u ) . The Mu suppression index was used to compare motor cortex activation among different conditions and to examine correlations with attention parameters, aiming to identify which attention parameters are more associated with motor cortex activities.
  • Engagement Index
The engagement index is one of the EEG parameters used to quantify task engagement, particularly its deterioration [26], and the state of the mental workload [27]. It was extracted by averaging the index from F3 and F4 using the equation b e t a / ( a l p h a + t h e t a ) [28]. This engagement index was compared among different conditions and assessed for correlation with the Mu suppression index.

2.8. Statistical Analysis

The requisite sample size was determined using G*Power software (version 3.1). This was for the analysis of Mu suppression as the primary endpoint, using repeated measures analysis of variance (RMANOVA). The hypothesis was that there would be differences among three conditions: NoF, PF, and NF. The effect size was set at a moderate level of 0.25, with a correlation of 0.7 among repeated measures. After the analysis, the effect size of the ANOVA obtained in this study was 0.67. This effect size is substantially larger than the initially assumed medium effect size of 0.25, and even exceeds the threshold for a large effect size, which is 0.4.
A significance level of 0.05 and a power of 80% were assumed. The calculated sample size suggests that 17 participants were required per group. However, considering the 5% dropout rate, a total of 18 participants per group was recommended. This sample size was deemed sufficient to detect statistically significant results and ensure the reliability and validity of the study.
Before conducting the analysis, normality testing was performed using both the Kolmogorov–Smirnov and Shapiro–Wilk tests to assess the assumption of normality. Most of the data satisfied the assumption of normality, supporting the use of parametric analysis methods. For comparison among the three conditions, rmANOVA was conducted for each outcome variable (averaged SSVEP power of O1 and O2, Mu suppression on C3 and C4, and engagement index), and a post hoc analysis, considering Bonferroni correction, was conducted if a significant main effect was observed. Moreover, Pearson’s correlation analysis was conducted to determine the relationship between the amount of Mu suppression and the EEG parameters, such as the intensity of SSVEP and engagement index. All statistical tests were two-tailed, with p < 0.05 considered statistically significant and 0.05 ≤ p < 0.15 considered a trend towards significance, to increase the sensitivity to detect potential selection bias, and they were conducted with the SPSS Version 11.0 statistic software package.

3. Results

3.1. SSVEP Responses

The SSVEP responses were observed while participants performed the rehabilitation tasks. The amplitude of the responses differed, depending on whether they were attentive or non-attentive to the task (Figure 3). The SSVEP power amplitudes were measured in µV2/Hz(dB). The mean attentive SSVEP was 3.84 ± 1.76 in O1 and 3.90 ± 2.05 in O2. The non-attentive SSVEP was −0.50 ± 1.32 and −1.12 ± 1.50, respectively. A paired t-test revealed that the attentive SSVEP was significantly higher than the non-attentive SSVEP in both O1 and O2 (O1: t (17) = 11.822, p < 0.001; O2: t = 8.565, p < 0.001).

3.2. Performance and Feedback

In the comparison of performance, the tracking errors observed in each condition were not different (Figure 4). The error distances between the target and cursor were 2.26 ± 0.13, 2.27 ± 0.14, and 2.28 ± 0.15 for the no feedback (NoF), performance feedback (PF), and neurofeedback (NF) conditions, respectively. Moreover, the amount of feedback received was not different between the PF and NF conditions, equaling 82.6% ± 3.1 and 80.8% ± 2.8.

3.3. SSVEP and Engagement

However, the intensity of SSVEP and engagement index showed significant main effects (F (2,34) = 5.474, p = 0.009 for the SSVEP, and F (2,34) = 5.821, p = 0.007 for the engagement index). Post hoc analysis revealed differences among conditions, in which the NoF condition showed the lowest and was statistically different from two other conditions (t (17) = 2.191, p = 0.128 for NoF versus PF and t (17) = 3.751, p = 0.005 for NoF versus NF in SSVEP; t (2.733) = 2.733, p = 0.043 for NoF versus PF and t (17) = 2.799, p = 0.037 for NoF versus NF in Engagement Index). No significant differences were found between the PF and NF condition in both SSVEP and engagement index (Figure 5).

3.4. Mu Suppression

The Mu suppression on C3 was the strongest in the NF condition, followed by the PF and NoF conditions (main effect: F (2,34) = 7.532, p = 0.002). Significant differences in the post hoc analysis were observed between NF and NoF (t (17) = 3.188, p = 0.016) and between NF and PF (t (17) = 2.780, p = 0.039) (Figure 5). NF is primarily different from conventional PF systems in that it furnishes real-time participatory feedback based on the user’s attentional state. PF directs the user based on movement accuracy but does not actively measure brain state engagement [50,51]. Conversely, NF uses SSVEP signals to assess attention and cognitive engagement, allowing users to actively participate in the rehabilitation process [52,53]. The direct link between attention and motor cortex activation explains why NF has stronger Mu inhibition than PF [54].
Studies demonstrate that attention-dependent motor learning strengthens neuroplasticity in stroke rehabilitation [55]. The attentional load directly modulates the neuroplasticity [56,57]. This supports our findings that NF is superior to PF for motor learning. Our results are coherent with previous research, proving that neurofeedback can enhance engagement, improve sensorimotor integration, and increase motor cortical excitability, indicating compelling neuroplastic adaptations.

3.5. Mu-SSVEP Correlation

The correlation analysis showed a significant correlation between Mu suppression and SSVEP in the NF condition (r (18) = −0.481, p = 0.043), while the other conditions did not reveal any significant correlation, but showed a trend towards significance. In addition, engagement did not significantly correlate with Mu suppression in any condition (Figure 6).

4. Discussion

Maintaining a patient’s focus and engagement in the task during rehabilitation is crucial for achieving positive outcomes. In this study, we introduced a novel VR rehabilitation system that incorporates BCI-based neurofeedback for upper limb rehabilitation. Our system aims to induce stronger motor cortex activation by encouraging patients not to lose their engagement while performing training through providing real-time feedback to them when it is determined that they are actively engaged in tasks, allowing patients to be aware of their own attentiveness. The findings of our study highlight the efficacy of this innovative system and its impact on training outcomes in terms of motor cortex activities.
Our approach utilized an SSVEP-based concentration classification paradigm [35], which has previously demonstrated its robustness in classifying distraction states during a tracking task; the study showed that SSVEP-BCI-based attentiveness state classification paradigm based on exogenous components could reveal more robust results, which vary less among individuals compared to those based on endogenous parameters such as spectrum components or time characteristics of the EEG.
In addition, our correlation results showing the significance, or trend towards significance, of the correlation between SSVEP and Mu suppression, and the lack of correlation between the engagement index and Mu suppression, in the three conditions indicate that longer staring at the target and cursor in the motor rehabilitation task evoked stronger Mu suppression, which means that it induced more motor cortex activity. This supports the idea that the SSVEP paradigm is more appropriate for motor rehabilitation tasks.
We integrated this paradigm for detecting the attentive state of participants into the VR-BCI neurofeedback rehabilitation system, which provides a tracking exercise where flickering cursors and target stimuli are provided. The flickering cursor and target allow SSVEP signals to be evoked in a user’s brain, and they can be observed to be stronger when a participant is attentively focused on the flickering target or cursor.
Both the SSVEP signal and the engagement index during the three conditions (neurofeedback, performance feedback, and no feedback) exhibited significantly different amplitudes, with the highest amplitude observed in the NF condition, followed by the PF condition, and the lowest in the NoF condition. These differences in SSVEP amplitude and engagement index likely reflect variations in attentional engagement with the task [33,34,42].
The noteworthy finding that SSVEP amplitude was significantly high in the NF and PF conditions suggests that the system that provides feedback effectively drew users’ attention to the task, particularly to the target, throughout their performance, encouraging them to keep their engagement high. The role of feedback in drawing the user’s engagement is also supported by the engagement index. This superior engagement likely led to stronger suppression of the Mu rhythm on C3, indicating enhanced activation of the motor cortex. This is a coincidence, with many studies insisting that attention is strongly associated with motor network activities, which could directly influence motor learning by facilitating brain plasticity. In motor execution or learning, many neuroscience studies show that paying attention [58,59,60,61] or being distracted [62,63] could lead to enhanced or reduced motor cortex activity.
In terms of the role of feedback, we could consider the direction of attention, which is one of the factors in motor learning that emphasizes the role of feedback. The external focus of attention, which refers to focusing on the outcomes or effects of movements made during exercise, which was realized in both the PF and NF conditions, is more beneficial for motor learning than the internal focus of attention [9,64,65,66]. Particularly, an external focus on the task in comparison with an internal focus is associated with higher activity in the motor network in learning new motor skills [67].
At this point, however, we need to focus on the fact that Mu suppression in the PF condition did not differ from the condition without feedback, while Mu suppression in the PF condition was significantly weaker than that of the NF condition. This is interesting when we consider the similar amount of feedback given and the comparable number of errors made in both the PF and NF conditions. Although behavioral similarity may predict that the feedback system based on performance also works effectively, the highest Mu suppression in the NF condition indicates that our proposed paradigm, which uses the SSVEP response to measure the user’s attention toward the task, has something to facilitate motor-related brain activation and drive neuroplastic changes. In the NF condition, patients may need to keep staring at the cursor and target for much longer than in the PF condition, where they may monitor their movement outcomes to reduce errors. This means that letting patients focus on their brain state based on their SSVEP strength has great potential to facilitate motor cortex activities.
Reinforcement and error-based learning based on trial-and-error correction are conventional motor learning approaches. Individuals refine their movements by correction; however, cognitive engagement, which is critical for long-term neuroplastic adaptations, does not appear [68]. On the other hand, NF continuously assesses attentional engagement in real time [69], and attentional engagement enhances synaptic potentiation, leading to compelling motor cortex plasticity and regained motor learning outcomes [7]. Our results demonstrate that the NF condition was significantly stronger than PF condition in Mu suppression, indicating greater activation of motor-related neural networks.
The stronger Mu suppression observed in the NF condition compared to the PF condition, despite similar feedback provision, implies that the NF condition possesses specific factors that induce superior motor cortex activation. This means that, although the presence of feedback could induce strong engagement reflected by the engagement index and SSVEP, playing an essential role in promoting engagement in motor learning [8,70,71], if the feedback is provided based on performance, it may not be enough to facilitate the corresponding motor circuits, and it might emphasize the superiority of a system based on the user’s attentive state rather than based on their brain.
The application of stimulation based on the brain state in rehabilitation has received attention, and several studies have emphasized that the manipulation of stimulation according to the brain state, like intention or engagement, is more effective and necessary in maximizing rehabilitation outcomes. Training with stimulation like functional electrical stimulation (FES), physical movements induced by orthoses, and visual feedback, or their combinations, that are provided according to participants’ brain states, such as when their motor imagination activates their brain, have shown to provide superior functional improvements in motor skills and induce motor circuity activities in BCI studies with stroke patients [19,72,73]. Even though it has noticeable outcomes, this kind of paradigm has shortcomings, in that patients should activate target areas of their brain, such as the motor cortex. However, patients who have deficits in that area may have difficulty activating the area. Therefore, the paradigm we propose here provides an advantage in that the system encourages patients to keep staring at the target stimuli while performing their motor tasks, which subsequently results in superior motor cortex activities by letting them pay attention to the task.
Moreover, stimulation in an attentive state could lead to a stronger association between brain function and stimulation, which facilitates brain plasticity. Clinically, coordinating feedback with the patient’s attentiveness and intention could be an important factor in properly facilitating the brain network and accelerating neural plasticity [74,75]. Several studies have shown that pairing PES with the movement intention of participants by detection based on EEG causes greater cortical plasticity [74,76,77]. Additionally, while performing action observation training alone or combined with FES, motor cortex facilitation was superiorly enhanced when feedback was given according to participants’ engagement using BCI technology [36,38,42].
In addition, the concept of the neurofeedback system, which provides feedback according to the brain state of a user while performing a task, is different from the conventional neurofeedback system, where individuals are asked to shape their EEG signal patterns. The traditional neurofeedback system focuses on enhancing attentional ability by inducing a tonic change in brain state rather than directly improving motor network plasticity and learning. Moreover, even a patient whose attention is enhanced after neurofeedback training may also lose engagement during rehabilitation. Therefore, in this case, the system proposed in this study, which could alert the individual whenever they lose their engagement during their rehabilitation training, would be helpful.
The findings of our study also highlight the importance of brain state-based feedback strategies in rehabilitation. While performance-based feedback plays a role in encouraging patients to exert more effort, our results suggest that it may not be sufficient to drive motor-related brain activation and induce neuroplastic changes. By tailoring feedback based on the user’s attentive state, rehabilitation programs can be better adapted to individual needs and optimize the engagement of motor circuits.
Despite its promising implications, our study had several limitations. Firstly, the sample size in our study was relatively small, which may limit the generalizability of the findings. However, a statistical power analysis was conducted using G*Power software to ensure that the sample size was appropriate. The effect size (η2 = 0.25) was estimated based on previous neurofeedback studies. With α = 0.05, power = 0.80, and three conditions (NoF, PF, NF), the necessary sample size was 17 participants per group. To account for potential dropouts (5%), we recruited 18 participants per group, ensuring statistical robustness. A post hoc analysis revealed an actual effect size of 0.67, which exceeds the threshold for a large effect (0.4), further confirming the study’s sufficient statistical power to detect meaningful differences among conditions.
Future studies with larger and more diverse participant groups are needed to validate our results. Additionally, the study focused specifically on upper limb rehabilitation, and the findings may not extend to other rehabilitation areas or different patient populations. Another limitation is the reliance on SSVEP-based concentration classification as the primary measure of attentive engagement. While SSVEP has been validated as a reliable measure in previous studies, incorporating multiple measures of attention and engagement, such as eye-tracking or subjective self-reports, would provide a more comprehensive understanding of the patient’s involvement in the task. Furthermore, the duration of the study was relatively short, and the long-term effects of the VR-BCI neurofeedback system on functional recovery and motor learning remain unknown. Future research should investigate the sustained benefits of the system over extended periods and explore the potential for the transferability of motor skills to real-world functional tasks.

5. Conclusions and Future Work

In conclusion, our study demonstrates the effectiveness of a VR-BCI neurofeedback rehabilitation system for upper limb rehabilitation, emphasizing the importance of maintaining patient focus and engagement. The findings support that feedback based on the user’s attentive state is crucial for promoting motor cortex activation and facilitating neuroplastic changes. This suggests that VR rehabilitation combined with an engagement feedback system could be successfully applied in clinics. However, hardware requirements (a BCI-compatible EEG headset with a VR setup), the education of clinicians, patient appropriateness (best-suited for patients with residual movement ability), and customization for clinical use (stroke survivors or individuals with neurodegenerative conditions) should be considered when integrating this system into clinical rehabilitation.
Future research should consider the potential habituation effects that may arise from repeated flickering stimuli, possibly affecting our paradigm’s reliability. Although we did not account for this effect, there was no decrease in the detection rate over time, as revealed by the classifier, and our results consistently showed that the NF system performed well in terms of SSVEP. Nevertheless, the habituation effect remains a factor to be addressed in the design of NF systems in further studies.
Additionally, further research should focus on optimizing the feedback paradigm to maximize its effectiveness, with advancements in portable EEG devices for at-home-based rehabilitation, and machine learning-based adaptive AI algorithms for personalized feedback based on patient progress.

Author Contributions

Conceptualization, H.L., B.A. and J.K.; methodology, H.L. and J.K.; software, H.L.; validation, H.L. and J.K.; formal analysis, H.L. and B.A.; investigation, H.L. and J.K.; resources, H.L. and J.K.; data curation, H.L.; writing—original draft preparation, H.L.; writing—review and editing, H.L., B.A. and J.K.; visualization, H.L.; supervision, J.K.; project administration, J.K.; funding acquisition, J.K. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Research Foundation of Korea (NRF) under a grant funded by the Korean government (MSIT) (No. NRF-2022R1A2B5B01001443, RS-2024-00397674).

Data Availability Statement

The datasets used and analyzed in this study are available from the corresponding author upon reasonable request.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
VRVirtual reality
SSVEPSteady-state visual evoked potential
BCIBrain-computer interface
NoFNo feedback
PFPerformance feedback
NFNeurofeedback

References

  1. Marzola, P.; Melzer, T.; Pavesi, E.; Gil-Mohapel, J.; Brocardo, P.S. Exploring the Role of Neuroplasticity in Development, Aging, and Neurodegeneration. Brain Sci. 2023, 13, 1610. [Google Scholar] [CrossRef]
  2. Keci, A.; Tani, K.; Xhema, J. Role of Rehabilitation in Neural Plasticity. Open Access Maced. J. Med. Sci. 2019, 7, 1540–1547. [Google Scholar] [CrossRef] [PubMed]
  3. Baroncelli, L.; Braschi, C.; Spolidoro, M.; Begenisic, T.; Sale, A.; Maffei, L. Nurturing Brain Plasticity: Impact of Environmental Enrichment. Cell Death Differ. 2010, 17, 1092–1103. [Google Scholar] [CrossRef] [PubMed]
  4. Tedla, J.S.; Gular, K.; Reddy, R.S.; de Sá Ferreira, A.; Rodrigues, E.C.; Kakaraparthi, V.N.; Gyer, G.; Sangadala, D.R.; Qasheesh, M.; Kovela, R.K.; et al. Effectiveness of Constraint-Induced Movement Therapy (CIMT) on Balance and Functional Mobility in the Stroke Population: A Systematic Review and Meta-Analysis. Healthcare 2022, 10, 495. [Google Scholar] [CrossRef] [PubMed]
  5. Shamweel, H.; Gupta, N. Constraint-Induced Movement Therapy through Telerehabilitation for Upper Extremity Function in Stroke. J. Neurorestoratology 2024, 12, 100108. [Google Scholar] [CrossRef]
  6. Zotey, V.; Andhale, A.; Shegekar, T.; Juganavar, A. Adaptive Neuroplasticity in Brain Injury Recovery: Strategies and Insights. Cureus 2023, 15, e45873. [Google Scholar] [CrossRef] [PubMed]
  7. Ahmed, B.; Khan, S.; Lim, H.; Ku, J. Challenges and Opportunities of Gamified BCI and BMI on Disabled People Learning: A Systematic Review. Electronics 2025, 14, 491. [Google Scholar] [CrossRef]
  8. Sanford, S.; Liu, M.; Selvaggi, T.; Nataraj, R. Effects of Visual Feedback Complexity on the Performance of a Movement Task for Rehabilitation. J. Mot. Behav. 2021, 53, 243–257. [Google Scholar] [CrossRef] [PubMed]
  9. Song, J.-H. The Role of Attention in Motor Control and Learning. Curr. Opin. Psychol. 2019, 29, 261–265. [Google Scholar] [CrossRef] [PubMed]
  10. Milnik, A.; Nowak, I.; Müller, N.G. Attention-Dependent Modulation of Neural Activity in Primary Sensorimotor Cortex. Brain Behav. 2013, 3, 54–66. [Google Scholar] [CrossRef] [PubMed]
  11. Kamke, M.R.; Hall, M.G.; Lye, H.F.; Sale, M.V.; Fenlon, L.R.; Carroll, T.J.; Riek, S.; Mattingley, J.B. Visual Attentional Load Influences Plasticity in the Human Motor Cortex. J. Neurosci. 2012, 32, 7001–7008. [Google Scholar] [CrossRef]
  12. Stefan, K.; Wycislo, M.; Classen, J. Modulation of Associative Human Motor Cortical Plasticity by Attention. J. Neurophysiol. 2004, 92, 66–72. [Google Scholar] [CrossRef] [PubMed]
  13. Kamke, M.R.; Ryan, A.E.; Sale, M.V.; Campbell, M.E.J.; Riek, S.; Carroll, T.J.; Mattingley, J.B. Visual Spatial Attention Has Opposite Effects on Bidirectional Plasticity in the Human Motor Cortex. J. Neurosci. 2014, 34, 1475–1480. [Google Scholar] [CrossRef]
  14. Maier, M.; Ballester, B.R.; Verschure, P.F.M.J. Principles of Neurorehabilitation After Stroke Based on Motor Learning and Brain Plasticity Mechanisms. Front. Syst. Neurosci. 2019, 13, 74. [Google Scholar] [CrossRef]
  15. Ros, T.; Munneke, M.A.M.; Ruge, D.; Gruzelier, J.H.; Rothwell, J.C. Endogenous Control of Waking Brain Rhythms Induces Neuroplasticity in Humans. Eur. J. Neurosci. 2010, 31, 770–778. [Google Scholar] [CrossRef]
  16. Sitaram, R.; Veit, R.; Stevens, B.; Caria, A.; Gerloff, C.; Birbaumer, N.; Hummel, F. Acquired Control of Ventral Premotor Cortex Activity by Feedback Training: An Exploratory Real-Time FMRI and TMS Study. Neurorehabil. Neural Repair 2012, 26, 256–265. [Google Scholar] [CrossRef] [PubMed]
  17. Cramer, S.C.; Sur, M.; Dobkin, B.H.; O’Brien, C.; Sanger, T.D.; Trojanowski, J.Q.; Rumsey, J.M.; Hicks, R.; Cameron, J.; Chen, D.; et al. Harnessing Neuroplasticity for Clinical Applications. Brain 2011, 134, 1591–1609. [Google Scholar] [CrossRef] [PubMed]
  18. Gu, C.; Lin, W.; He, X.; Zhang, L.; Zhang, M. IMU-Based Motion Capture System for Rehabilitation Applications: A Systematic Review. Biomim. Intell. Robot. 2023, 3, 100097. [Google Scholar] [CrossRef]
  19. Ramos-Murguialday, A.; Broetz, D.; Rea, M.; Läer, L.; Yilmaz, O.; Brasil, F.L.; Liberati, G.; Curado, M.R.; Garcia-Cossio, E.; Vyziotis, A.; et al. Brain-Machine Interface in Chronic Stroke Rehabilitation: A Controlled Study. Ann. Neurol. 2013, 74, 100–108. [Google Scholar] [CrossRef]
  20. Liu, Q.; Yang, X.; Chen, Z.; Zhang, W. Using Synchronized Eye Movements to Assess Attentional Engagement. Psychol. Res. 2023, 87, 2039–2047. [Google Scholar] [CrossRef] [PubMed]
  21. Grosse-Wentrup, M.; Schölkopf, B.; Hill, J. Causal Influence of Gamma Oscillations on the Sensorimotor Rhythm. Neuroimage 2011, 56, 837–842. [Google Scholar] [CrossRef] [PubMed]
  22. Johnson, L.; Burridge, J.H.; Demain, S.H. Internal and External Focus of Attention During Gait Re-Education: An Observational Study of Physical Therapist Practice in Stroke Rehabilitation. Phys. Ther. 2013, 93, 957–966. [Google Scholar] [CrossRef]
  23. Phuphanich, M.E.; Droessler, J.; Altman, L.; Eapen, B.C. Movement-Based Therapies in Rehabilitation. Phys. Med. Rehabil. Clin. N. Am. 2020, 31, 577–591. [Google Scholar] [CrossRef]
  24. Levordashka, A.; Stanton Fraser, D.; Gilchrist, I.D. Measuring Real-Time Cognitive Engagement in Remote Audiences. Sci. Rep. 2023, 13, 10516. [Google Scholar] [CrossRef] [PubMed]
  25. Forget, M.; Pertel, N. Le Enhancing Neuroplasticity and Promoting Brain Health at Work: The Role of Learning and Memory in Workplace Performance. In Learning and Memory; Heinbockel, T., Ed.; IntechOpen: Rijeka, Croatia, 2024. [Google Scholar]
  26. Rajashekar, D.; Boyer, A.; Larkin-Kaiser, K.A.; Dukelow, S.P. Technological Advances in Stroke Rehabilitation: Robotics and Virtual Reality. Phys. Med. Rehabil. Clin. N. Am. 2024, 35, 383–398. [Google Scholar] [CrossRef]
  27. Lee, H.Y.; Hyun, S.E.; Oh, B.-M. Rehabilitation for Impaired Attention in the Acute and Post-Acute Phase After Traumatic Brain Injury: A Narrative Review. Korean J. Neurotrauma 2023, 19, 20–31. [Google Scholar] [CrossRef] [PubMed]
  28. Yordanova, J.; Kolev, V.; Nicolardi, V.; Simione, L.; Mauro, F.; Garberi, P.; Raffone, A.; Malinowski, P. Attentional and Cognitive Monitoring Brain Networks in Long-Term Meditators Depend on Meditation States and Expertise. Sci. Rep. 2021, 11, 4909. [Google Scholar] [CrossRef]
  29. deBettencourt, M.T.; Cohen, J.D.; Lee, R.F.; Norman, K.A.; Turk-Browne, N.B. Closed-Loop Training of Attention with Real-Time Brain Imaging. Nat. Neurosci. 2015, 18, 470–475. [Google Scholar] [CrossRef]
  30. Li, M.; He, D.; Li, C.; Qi, S. Brain-Computer Interface Speller Based on Steady-State Visual Evoked Potential: A Review Focusing on the Stimulus Paradigm and Performance. Brain Sci. 2021, 11, 450. [Google Scholar] [CrossRef] [PubMed]
  31. Cheng, M.; Gao, X.; Gao, S.; Xu, D. Design and Implementation of a Brain-Computer Interface with High Transfer Rates. IEEE Trans. Biomed. Eng. 2002, 49, 1181–1186. [Google Scholar] [CrossRef] [PubMed]
  32. Han, X.; Lin, K.; Gao, S.; Gao, X. A Novel System of SSVEP-Based Human-Robot Coordination. J. Neural Eng. 2019, 16, 16006. [Google Scholar] [CrossRef]
  33. İşcan, Z.; Nikulin, V.V. Steady State Visual Evoked Potential (SSVEP) Based Brain-Computer Interface (BCI) Performance under Different Perturbations. PLoS ONE 2018, 13, e0191673. [Google Scholar] [CrossRef]
  34. Ordikhani-Seyedlar, M.; Sorensen, H.B.D.; Kjaer, T.W.; Siebner, H.R.; Puthusserypady, S. SSVEP-Modulation by Covert and Overt Attention: Novel Features for BCI in Attention Neuro-Rehabilitation. In Proceedings of the 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, USA, 26–30 August 2014; Volume 2014, pp. 5462–5465. [Google Scholar] [CrossRef]
  35. Lim, H.; Kim, S.; Ku, J. Distraction Classification During Target Tracking Tasks Involving Target and Cursor Flickering Using EEGNet. IEEE Trans. Neural Syst. Rehabil. Eng. 2022, 30, 1113–1119. [Google Scholar] [CrossRef] [PubMed]
  36. Kim, M.G.; Lim, H.; Lee, H.S.; Han, I.J.; Ku, J.; Kang, Y.J. Brain-Computer Interface-Based Action Observation Combined with Peripheral Electrical Stimulation Enhances Corticospinal Excitability in Healthy Subjects and Stroke Patients. J. Neural Eng. 2022, 19, 036039. [Google Scholar] [CrossRef] [PubMed]
  37. Lim, H.; Ku, J. Flickering Exercise Video Produces Mirror Neuron System (MNS) Activation and Steady State Visually Evoked Potentials (SSVEPs). Biomed. Eng. Lett. 2017, 7, 281–286. [Google Scholar] [CrossRef]
  38. Choi, H.; Lim, H.; Kim, J.W.; Kang, Y.J.; Ku, J. Brain Computer Interface-Based Action Observation Game Enhances Mu Suppression in Patients with Stroke. Electronics 2019, 8, 1466. [Google Scholar] [CrossRef]
  39. Oldfield, R.C. The Assessment and Analysis of Handedness: The Edinburgh Inventory. Neuropsychologia 1971, 9, 97–113. [Google Scholar] [CrossRef] [PubMed]
  40. Drigas, A.; Sideraki, A. Brain Neuroplasticity Leveraging Virtual Reality and Brain–Computer Interface Technologies. Sensors 2024, 24, 5725. [Google Scholar] [CrossRef] [PubMed]
  41. Jin, W.; Zhu, X.; Qian, L.; Wu, C.; Yang, F.; Zhan, D.; Kang, Z.; Luo, K.; Meng, D.; Xu, G. Electroencephalogram-Based Adaptive Closed-Loop Brain-Computer Interface in Neurorehabilitation: A Review. Front. Comput. Neurosci. 2024, 18, 1431815. [Google Scholar] [CrossRef]
  42. Lim, H.; Ku, J. A Brain-Computer Interface-Based Action Observation Game That Enhances Mu Suppression. IEEE Trans. Neural Syst. Rehabil. Eng. 2018, 26, 2290–2296. [Google Scholar] [CrossRef]
  43. Azadi Moghadam, M.; Maleki, A. Fatigue Factors and Fatigue Indices in SSVEP-Based Brain-Computer Interfaces: A Systematic Review and Meta-Analysis. Front. Hum. Neurosci. 2023, 17, 1248474. [Google Scholar] [CrossRef]
  44. Lotte, F.; Guan, C. Regularizing Common Spatial Patterns to Improve BCI Designs: Unified Theory and New Algorithms. IEEE Trans. Biomed. Eng. 2011, 58, 355–362. [Google Scholar] [CrossRef] [PubMed]
  45. Falcon-Caro, A.; Shirani, S.; Ferreira, J.F.; Bird, J.J.; Sanei, S. Formulation of Common Spatial Patterns for Multi-Task Hyperscanning BCI. IEEE Trans. Biomed. Eng. 2024, 71, 1950–1957. [Google Scholar] [CrossRef] [PubMed]
  46. Chang, C.-Y.; Hsu, S.-H.; Pion-Tonachini, L.; Jung, T.-P. Evaluation of Artifact Subspace Reconstruction for Automatic Artifact Components Removal in Multi-Channel EEG Recordings. IEEE Trans. Biomed. Eng. 2020, 67, 1114–1121. [Google Scholar] [CrossRef] [PubMed]
  47. Hyvärinen, A.; Oja, E. Independent Component Analysis: Algorithms and Applications. Neural Netw. 2000, 13, 411–430. [Google Scholar] [CrossRef] [PubMed]
  48. Cataldo, A.; Criscuolo, S.; De Benedetto, E.; Masciullo, A.; Pesola, M.; Schiavoni, R.; Invitto, S. A Method for Optimizing the Artifact Subspace Reconstruction Performance in Low-Density EEG. IEEE Sens. J. 2022, 22, 21257–21265. [Google Scholar] [CrossRef]
  49. Ronca, V.; Capotorto, R.; Di Flumeri, G.; Giorgi, A.; Vozzi, A.; Germano, D.; Di Virgilio, V.; Borghini, G.; Cartocci, G.; Rossi, D.; et al. Optimizing EEG Signal Integrity: A Comprehensive Guide to Ocular Artifact Correction. Bioengineering 2024, 11, 1018. [Google Scholar] [CrossRef] [PubMed]
  50. Tosti, B.; Corrado, S.; Mancone, S.; Di Libero, T.; Rodio, A.; Andrade, A.; Diotaiuti, P. Integrated Use of Biofeedback and Neurofeedback Techniques in Treating Pathological Conditions and Improving Performance: A Narrative Review. Front. Neurosci. 2024, 18, 1358481. [Google Scholar] [CrossRef]
  51. Tsai, P.-C.; Akpan, A.; Tang, K.-T.; Lakany, H. Brain-Computer Interfaces for Cognitive Enhancement in Older People—Challenges and Applications: A Systematic Review. BMC Geriatr. 2025, 25, 36. [Google Scholar] [CrossRef] [PubMed]
  52. Ordikhani-Seyedlar, M.; Lebedev, M.A.; Sorensen, H.B.D.; Puthusserypady, S. Neurofeedback Therapy for Enhancing Visual Attention: State-of-the-Art and Challenges. Front. Neurosci. 2016, 10, 352. [Google Scholar] [CrossRef]
  53. Sotoodeh, M.S.; Chien, S.H.-L.; Hadjikhani, N. Visual Attention Modulates Mu Suppression during Biological Motion Perception in Autistic Individuals. Eur. J. Neurosci. 2024, 60, 6668–6685. [Google Scholar] [CrossRef]
  54. Hobson, H.M.; Bishop, D.V.M. The Interpretation of Mu Suppression as an Index of Mirror Neuron Activity: Past, Present and Future. R. Soc. Open Sci. 2017, 4, 160662. [Google Scholar] [CrossRef] [PubMed]
  55. Aderinto, N.; AbdulBasit, M.O.; Olatunji, G.; Adejumo, T. Exploring the Transformative Influence of Neuroplasticity on Stroke Rehabilitation: A Narrative Review of Current Evidence. Ann. Med. Surg. 2023, 85, 4425–4432. [Google Scholar] [CrossRef] [PubMed]
  56. Corrado, S.; Tosti, B.; Mancone, S.; Di Libero, T.; Rodio, A.; Andrade, A.; Diotaiuti, P. Improving Mental Skills in Precision Sports by Using Neurofeedback Training: A Narrative Review. Sports 2024, 12, 70. [Google Scholar] [CrossRef] [PubMed]
  57. Cheng, M.-Y.; Yu, C.-L.; An, X.; Wang, L.; Tsai, C.-L.; Qi, F.; Wang, K.-P. Evaluating EEG Neurofeedback in Sport Psychology: A Systematic Review of RCT Studies for Insights into Mechanisms and Performance Improvement. Front. Psychol. 2024, 15, 1331997. [Google Scholar] [CrossRef] [PubMed]
  58. Jueptner, M.; Stephan, K.M.; Frith, C.D.; Brooks, D.J.; Frackowiak, R.S.; Passingham, R.E. Anatomy of Motor Learning. I. Frontal Cortex and Attention to Action. J. Neurophysiol. 1997, 77, 1313–1324. [Google Scholar] [CrossRef] [PubMed]
  59. Rushworth, M.F.; Krams, M.; Passingham, R.E. The Attentional Role of the Left Parietal Cortex: The Distinct Lateralization and Localization of Motor Attention in the Human Brain. J. Cogn. Neurosci. 2001, 13, 698–710. [Google Scholar] [CrossRef]
  60. Rowe, J.; Friston, K.; Frackowiak, R.; Passingham, R. Attention to Action: Specific Modulation of Corticocortical Interactions in Humans. Neuroimage 2002, 17, 988–998. [Google Scholar] [CrossRef]
  61. Binkofski, F.; Fink, G.R.; Geyer, S.; Buccino, G.; Gruber, O.; Shah, N.J.; Taylor, J.G.; Seitz, R.J.; Zilles, K.; Freund, H.-J. Neural Activity in Human Primary Motor Cortex Areas 4a and 4p Is Modulated Differentially by Attention to Action. J. Neurophysiol. 2002, 88, 514–519. [Google Scholar] [CrossRef] [PubMed]
  62. Johansen-Berg, H.; Matthews, P.M. Attention to Movement Modulates Activity in Sensori-Motor Areas, Including Primary Motor Cortex. Exp. Brain Res. 2002, 142, 13–24. [Google Scholar] [CrossRef] [PubMed]
  63. Rodríguez, M.; Muñiz, R.; González, B.; Sabaté, M. Hand Movement Distribution in the Motor Cortex: The Influence of a Concurrent Task and Motor Imagery. Neuroimage 2004, 22, 1480–1491. [Google Scholar] [CrossRef]
  64. Hunt, C.; Paez, A.; Folmar, E. The Impact of Attentional Focus on the Treatment of Musculoskeletal and Movement Disorders. Int. J. Sports Phys. Ther. 2017, 12, 901–907. [Google Scholar] [CrossRef] [PubMed]
  65. Wulf, G.; Prinz, W. Directing Attention to Movement Effects Enhances Learning: A Review. Psychon. Bull. Rev. 2001, 8, 648–660. [Google Scholar] [CrossRef]
  66. Wulf, G.; Shea, C.; Lewthwaite, R. Motor Skill Learning and Performance: A Review of Influential Factors. Med. Educ. 2010, 44, 75–84. [Google Scholar] [CrossRef]
  67. Zentgraf, K.; Lorey, B.; Bischoff, M.; Zimmermann, K.; Stark, R.; Munzert, J. Neural Correlates of Attentional Focusing during Finger Movements: A FMRI Study. J. Mot. Behav. 2009, 41, 535–541. [Google Scholar] [CrossRef] [PubMed]
  68. De Brouwer, A.J.; Areshenkoff, C.N.; Rashid, M.R.; Flanagan, J.R.; Poppenk, J.; Gallivan, J.P. Human Variation in Error-Based and Reinforcement Motor Learning Is Associated with Entorhinal Volume. Cereb. Cortex 2022, 32, 3423–3440. [Google Scholar] [CrossRef] [PubMed]
  69. Seidler, R.D.; Kwak, Y.; Fling, B.W.; Bernard, J.A. Neurocognitive Mechanisms of Error-Based Motor Learning. Adv. Exp. Med. Biol. 2013, 782, 39–60. [Google Scholar] [CrossRef] [PubMed]
  70. Sigrist, R.; Rauter, G.; Riener, R.; Wolf, P. Augmented Visual, Auditory, Haptic, and Multimodal Feedback in Motor Learning: A Review. Psychon. Bull. Rev. 2013, 20, 21–53. [Google Scholar] [CrossRef] [PubMed]
  71. Ford, K.R.; DiCesare, C.A.; Myer, G.D.; Hewett, T.E. Real-Time Biofeedback to Target Risk of Anterior Cruciate Ligament Injury: A Technical Report for Injury Prevention and Rehabilitation. J. Sport Rehabil. 2015, 24, 2013-0138. [Google Scholar] [CrossRef]
  72. Biasiucci, A.; Leeb, R.; Iturrate, I.; Perdikis, S.; Al-Khodairy, A.; Corbet, T.; Schnider, A.; Schmidlin, T.; Zhang, H.; Bassolino, M.; et al. Brain-Actuated Functional Electrical Stimulation Elicits Lasting Arm Motor Recovery after Stroke. Nat. Commun. 2018, 9, 2421. [Google Scholar] [CrossRef]
  73. Sebastián-Romagosa, M.; Cho, W.; Ortner, R.; Murovec, N.; Von Oertzen, T.; Kamada, K.; Allison, B.Z.; Guger, C. Brain-Computer Interface Treatment for Motor Rehabilitation of Upper Extremity of Stroke Patients—A Feasibility Study. Front. Neurosci. 2020, 14, 591435. [Google Scholar] [CrossRef] [PubMed]
  74. Niazi, I.K.; Mrachacz-Kersting, N.; Jiang, N.; Dremstrup, K.; Farina, D. Peripheral Electrical Stimulation Triggered by Self-Paced Detection of Motor Intention Enhances Motor Evoked Potentials. IIEEE Trans. Neural Syst. Rehabil. Eng. 2012, 20, 595–604. [Google Scholar] [CrossRef]
  75. Jochumsen, M.; Navid, M.S.; Rashid, U.; Haavik, H.; Niazi, I.K. EMG- Versus EEG-Triggered Electrical Stimulation for Inducing Corticospinal Plasticity. IIEEE Trans. Neural Syst. Rehabil. Eng. 2019, 27, 1901–1908. [Google Scholar] [CrossRef]
  76. Ibanez, J.; Serrano, J.I.; del Castillo, M.D.; Monge, E.; Molina, F.; Rivas, F.M.; Alguacil, I.; Miangolarra, J.C.; Pons, J.L. Upper-Limb Muscular Electrical Stimulation Driven by EEG-Based Detections of the Intentions to Move: A Proposed Intervention for Patients with Stroke. In Proceedings of the 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, USA, 26–30 August 2014; Volume 2014, pp. 1646–1649. [Google Scholar] [CrossRef]
  77. McGie, S.C.; Zariffa, J.; Popovic, M.R.; Nagai, M.K. Short-Term Neuroplastic Effects of Brain-Controlled and Muscle-Controlled Electrical Stimulation. Neuromodulation 2015, 18, 233–240. [Google Scholar] [CrossRef] [PubMed]
Electronics 14 00827 g001
Figure 1. Procedure diagram for experiment.
Figure 1. Procedure diagram for experiment.
Electronics 14 00827 g001
Electronics 14 00827 g002
Figure 2. The experimental setting and conditions. (a) A participant wearing the EEG device and HMD is performing the target tracking task with the controller in their hand. The EEG data are recorded and processed in real-time, and reveal the neurofeedback state according to whether the participant is attentively staring at the flickering target. (b) Three conditions (NoF, PF, and NF) were present: the NoF condition (upper) provided no feedback on tracking, while the PF and NF (lower) conditions included showering and blooming feedback with sound. (c) An example of a target trajectory. The target moves along an invisible circle-shaped track, and transitions into another adjacent cloned circle-shaped track when it changes direction. (d) The EEG electrodes are located on areas where the HMD straps do not pass, so that the EEG signal has minimal interference from body movements.
Figure 2. The experimental setting and conditions. (a) A participant wearing the EEG device and HMD is performing the target tracking task with the controller in their hand. The EEG data are recorded and processed in real-time, and reveal the neurofeedback state according to whether the participant is attentively staring at the flickering target. (b) Three conditions (NoF, PF, and NF) were present: the NoF condition (upper) provided no feedback on tracking, while the PF and NF (lower) conditions included showering and blooming feedback with sound. (c) An example of a target trajectory. The target moves along an invisible circle-shaped track, and transitions into another adjacent cloned circle-shaped track when it changes direction. (d) The EEG electrodes are located on areas where the HMD straps do not pass, so that the EEG signal has minimal interference from body movements.
Electronics 14 00827 g002
Electronics 14 00827 g003
Figure 3. Power spectrum density corresponding to the attentive and inattentive states in channels O1 and O2. The SSVEP peaks were observed in both states, but those in the attentive state were stronger.
Figure 3. Power spectrum density corresponding to the attentive and inattentive states in channels O1 and O2. The SSVEP peaks were observed in both states, but those in the attentive state were stronger.
Electronics 14 00827 g003
Electronics 14 00827 g004
Figure 4. The performance outcomes. The tracking error (left) represents the distance between the target and the cursor, and the feedback ratio (right) represents the ratio of the amount of feedback provided during the tracking period. The error bar represents the standard error.
Figure 4. The performance outcomes. The tracking error (left) represents the distance between the target and the cursor, and the feedback ratio (right) represents the ratio of the amount of feedback provided during the tracking period. The error bar represents the standard error.
Electronics 14 00827 g004
Electronics 14 00827 g005
Figure 5. The EEG outcomes. The SSVEP power (left), engagement index (middle), and Mu suppression on C3 (right) are the strongest in the NF condition, which shows significantly different results from the other two conditions. ° represents the trend toward significance; * and ** represent p < 0.05 and p < 0.01, respectively, after Bonferroni correction; and the error bars represent the standard error.
Figure 5. The EEG outcomes. The SSVEP power (left), engagement index (middle), and Mu suppression on C3 (right) are the strongest in the NF condition, which shows significantly different results from the other two conditions. ° represents the trend toward significance; * and ** represent p < 0.05 and p < 0.01, respectively, after Bonferroni correction; and the error bars represent the standard error.
Electronics 14 00827 g005
Electronics 14 00827 g006
Figure 6. The correlation results between the SSVEP (upper) or engagement index (lower) and Mu suppression on C3. The SSVEP showed a negative correlation tendency with Mu suppression in all three conditions, which was particularly significant in the NF condition, while the engagement index did not show any pattern. ° represents the trend toward significance, and * represents p < 0.05.
Figure 6. The correlation results between the SSVEP (upper) or engagement index (lower) and Mu suppression on C3. The SSVEP showed a negative correlation tendency with Mu suppression in all three conditions, which was particularly significant in the NF condition, while the engagement index did not show any pattern. ° represents the trend toward significance, and * represents p < 0.05.
Electronics 14 00827 g006
Table 1. Table of participants’ demographics.
Table 1. Table of participants’ demographics.
CharacteristicValue
Sample Size18
Gender8 Male, 10 Female
Age (Mean ± SD)27.33 ± 2.45
Hand DominanceAll Right-Handed (Confirmed by Edinburgh Handedness Inventory)
Neurological ConditionsNone
Prior BCI ExperienceNone
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Lim, H.; Ahmed, B.; Ku, J. Brain-Computer Interface Based Engagement Feedback in Virtual Reality Rehabilitation: Promoting Motor Cortex Activation. Electronics 2025, 14, 827. https://doi.org/10.3390/electronics14050827

AMA Style

Lim H, Ahmed B, Ku J. Brain-Computer Interface Based Engagement Feedback in Virtual Reality Rehabilitation: Promoting Motor Cortex Activation. Electronics. 2025; 14(5):827. https://doi.org/10.3390/electronics14050827

Chicago/Turabian Style

Lim, Hyunmi, Bilal Ahmed, and Jeonghun Ku. 2025. "Brain-Computer Interface Based Engagement Feedback in Virtual Reality Rehabilitation: Promoting Motor Cortex Activation" Electronics 14, no. 5: 827. https://doi.org/10.3390/electronics14050827

APA Style

Lim, H., Ahmed, B., & Ku, J. (2025). Brain-Computer Interface Based Engagement Feedback in Virtual Reality Rehabilitation: Promoting Motor Cortex Activation. Electronics, 14(5), 827. https://doi.org/10.3390/electronics14050827

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop