Brain-computer interface training for multimodal functional recovery in patients with brain injury: a case series

Introduction

Brain injuries caused by conditions such as intracranial tumors or hemorrhagic stroke often result in functional impairment across multiple domains, including motor skills, language, and cognition. Traditional rehabilitation approaches typically target a single domain, and thus often overlook the complex interplay between brain networks involved in multimodal recovery (1). One study reported that brain-computer interface (BCI) training, particularly when based on motor imagery, can enhance neuroplasticity and lead to improvements across several functional systems (2). However, the mechanisms underlying these cross-domain effects remain poorly understood.

Emerging evidence has linked BCI-driven improvements in motor function to changes in brain oscillatory activity and resting-state connectivity, particularly in the ipsilateral hemisphere of patients with subacute stroke (3). A functional magnetic resonance imaging (fMRI) study further revealed enhanced connectivity across widespread cortical regions—including the frontal, parietal, and occipital lobes—after BCI interventions. These findings suggest that BCI training may promote not only localized reorganization but also large-scale network remodeling (4,5).

In our study, we combined electroencephalography (EEG) and magnetic resonance imaging (MRI)-based analyses to investigate the multimodal effects of BCI training in patients with subacute brain injury. By integrating motor imagery-based training with multidimensional clinical assessments and advanced connectomics, we aimed to characterize the potential mechanisms of functional recovery across motor, language, and cognitive domains. We also examined how individualized patterns of network reorganization might relate to functional gains, providing novel insights into the personalization of neurorehabilitation strategies. We present this article in accordance with the AME Case Series reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1136/rc).

Methods

A prospective, single-center case series was conducted at the Department of Neurosurgery, Shanxi Provincial People’s Hospital between November 2021 and March 2023. Five patients with subacute brain dysfunction underwent 6 weeks of BCI rehabilitation training (Figure 1). The intervention was delivered by certified professionals with more than 3 years of experience in BCI therapy.

Figure 1 Photograph of brain-computer interaction training.

Baseline characteristics collected included age, sex, handedness, education level, occupation, diagnosis, and duration of illness (Table 1). Each patient participated in 25-minute training sessions approximately five times per week. Follow-up evaluations were conducted 2 months after the final training session and included the following: (I) baseline data (1 week before intervention); (II) process data (collected within 24 hours of each session); (III) endpoint data (48 hours post-intervention); and (IV) follow-up data (via telephone and outpatient visits).

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was approved by the Ethics Committee of Shanxi Provincial People’s Hospital (2022 Province Medical Ethics Review No. 79). Written informed consent was obtained from all participants.

The training facility was equipped with a 16-channel EEG cap (10–20 system), a 3.0-T MRI scanner (Discovery MR750, GE HealthCare, Chicago, IL, USA), and a standardized rehabilitation assessment suite. Data collection was supported by the hospital’s electronic medical record and the picture archiving and communication system (PACS) systems. All data were entered in duplicate to ensure accuracy.

Participants

The study included five adults (four males and one female; mean age 54.4±10.3 years) with diagnoses including basal ganglia hemorrhage and frontoparietal glioma resection. The inclusion criteria were as follows: (I) brain injury confirmed by MRI or computed tomography (CT); (II) motor, language, or cognitive dysfunction; (III) within 1–12 months after the first impairment of brain function; (IV) age ≥18 years; (V) Mini-Mental State Examination (MMSE) score ≥20; and (VI) ability to understand basic instructions and provide informed consent.

Meanwhile, the exclusion criteria were as follows: (I) severe psychiatric or cognitive disorder; (II) significant cardiac, hepatic, renal, or pulmonary disease; (III) limb deformities, wounds, or severe pain; (IV) profound visual or auditory impairment; (V) contraindications to MRI; and (VI) concurrent participation in other clinical trials.

Intervention measures

The intervention measures for the five patients included BCI training, functional scale assessment, and online classification accuracy (CA) evaluation.

BCI training

The intervention involved a noninvasive motor imagery-based BCI system with real-time EEG acquisition. Patients were instructed to imagine specific hand movements, which were translated into control signals for an exoskeleton device that assisted in hand grasping and release. Each session included approximately 60 motor imagery tasks, displayed via alternating left-right hand videos and auditory cues. Training was structured into 20-task sets with 1-minute rest intervals between sets. Head and body movements were minimized during sessions. All programs follow the Chinese clinical trial guidelines related to medical devices.

Functional scale assessment

Two doctors with more than 3 years of clinical experience conducted the following functional scale assessments for the participants before and after the entire training stage: the Fugl-Meyer Assessment Scale (FMA) for evaluating motor function of the upper and lower limbs, the Modified Ashworth Scale (MAS) for grading the muscle tone of the target muscle groups, the Western Aphasia Battery (WAB) for assessing the type of aphasia and the severity of speech disorders, and the MMSE for reflecting the degree of cognitive dysfunction. These assessments provided a comprehensive evaluation of the participants before and after the training, including of motor function, muscle tone, aphasia type, severity of speech disorders, and cognitive function.

Online CA

This study measured training performance based on the CA of the patients during task execution. The first stage involved offline data collection without real-time feedback. After completion, 10-fold cross-validation was used to obtain the offline accuracy at the time. All data from the first stage were used as the training set to obtain the optimal classifier for predicting the EEG data in the second stage. Subsequently, the CA during the second stage of motor imagery tasks was obtained. Higher CA values in the patients indicated better motor imagery ability during the training period.

EEG signal processing

EEG data were acquired on training days 1 and 28 with a 16-channel cap (10–20 system). Signals during 4-second motor imagery periods were extracted from the central motor cortex, and power spectral density (PSD) was calculated for the 8 to 30-Hz frequency band. Topographic maps were generated to analyze energy distribution in the C3 and C4 electrodes and to identify event-related desynchronization (ERD), a biomarker of cortical activation during motor imagery.

Image data processing

MRI and fMRI scans were performed before and after the intervention with a 3-T Discovery MR750 scanner (GE HealthCare). High-resolution T1-weighted images, diffusion tensor imaging (DTI), and resting-state fMRI were acquired. The scanning protocol included (I) T1-weighted images with a three-dimensional brain volume (BRAVO) sequence [repetition time (TR) 8.5 ms, echo time (TE) 3.2 ms, and voxel size 1 mm × 1 mm × 1 mm]; (II) DTI (TR 15,000 ms, TE 74.8 ms, voxel size 2 mm × 2 mm × 2 mm); and (III) resting state fMRI (TR 3,000 ms, TE 30 ms, voxel size 3 mm × 3 mm × 3 mm, and 180 volumes per session).

For patients showing strong motor imagery performance and clinical improvement, advanced connectomics analysis was performed according to the Human Connectome Project Multimodal Parcellation 1.0 (HCP-MMP 1.0) (6). Functional connectivity matrices were generated across 379 cortical and subcortical regions via comparative analysis with healthy controls.

Results

Self-reported status and clinical scale outcomes

Among the five patients, four (patients 1, 2, 3, and 4) had aphasia at baseline. Following the BCI training period, improvements in WAB scores were observed in patients 1, 3, and 4. Patient 1, diagnosed with thalamic aphasia, showed marked improvements in spontaneous speech, speech volume, and comprehension, with retained repetition ability. Patient 3, who had anomic aphasia, demonstrated increased spontaneous speech and greater accuracy in naming and comprehension tasks. Patient 4, initially presenting with telegraphic speech and impaired understanding, showed moderate gains in expressive language and memory. In contrast, patient 2, who exhibited global aphasia, did not improve in verbal output or comprehension and remained unable to complete standard assessments.

Regarding motor impairment, patients 2, 4, and 5 presented with post-stroke unilateral hemiparesis. After the intervention, (I) patient 2 exhibited significant improvements in upper limb strength and lower limb mobility, regaining independent ambulation with wall support; (II) patient 4 experienced improved joint mobility and standing balance, with gains in proximal upper limb strength and partial voluntary ankle control; and (III) patient 5 showed increased spontaneous activity and improved coordination, enabling independent standing and enhanced hand grip.

Cognitive function, as measured by MMSE, improved in patients 1, 3, 4, and 5. At baseline, patients 1 and 4 had moderate impairment, while patients 3 and 5 had mild impairment. Patient 2 could not be evaluated due to global aphasia. The average MMSE score increased from 17.5 to 21.8 across the cohort. The full data are provided in Tables 2,3.

Following the 5-week BCI intervention, all five participants demonstrated measurable gains in motor imagery CA, with the individual improvement rate ranging from 3.7% to 28.6%. The full data are summarized in Table 4. Compliance with the training protocol was high, with 135 out of 140 sessions completed (96.4%). Only one patient (patient 2) reported a mild, transient headache, which resolved without intervention.

EEG changes: PSD and topographic mapping

Patients 3, 4, and 5 initially displayed hemispheric asymmetry in PSD patterns between the C3 and C4 electrodes during left-right hand motor imagery. As training progressed, PSD curves became smoother, with moderate energy reductions and convergence toward central regions.

Patients 1 and 2 showed less distinction between hemispheres at baseline, possibly due to compensatory mechanisms. However, by day 28, patient 2 demonstrated a clear decrease in energy during right-hand imagery, revealing a post-intervention hemispheric shift.

Topographic EEG maps revealed ERD patterns in the affected motor cortex in all patients, particularly in patients 2 and 5, suggesting increased cortical recruitment over the course of training (Figures 2-6).

Figure 2 Intervention adjustment flowchart (including learning curve) and dynamic changes in the C3/C4 energy difference (including PSD from electrodes C3 and C4 and topographic maps of the affected brain during the 1st and 28th training sessions) of patient 1. The value in color bar of topographic map represents the energy subtracted from the baseline during the action, measured in µV2/Hz. C3/C4, standard electrode positions based on the international 10–20 electroencephalography system, corresponding to the left (C3) and right (C4) sensorimotor cortices; PSD, power spectral density.

Figure 3 Intervention adjustment flowchart (including learning curve) and dynamic changes in C3/C4 energy difference (including PSD from electrodes C3 and C4 and topographic maps of the affected brain during the 1st and 28th training sessions) of patient 2. The value in color bar of topographic map represents the energy subtracted from the baseline during the action, measured in µV2/Hz. C3/C4, standard electrode positions based on the international 10–20 electroencephalography system, corresponding to the left (C3) and right (C4) sensorimotor cortices; PSD, power spectral density.

Figure 4 Intervention adjustment flowchart (including learning curve) and dynamic changes in C3/C4 energy difference (including PSD from electrodes C3 and C4 and topographic maps of the affected brain during the 1st and 28th training sessions) of patient 3. The value in color bar of topographic map represents the energy subtracted from the baseline during the action, measured in µV2/Hz. C3/C4, standard electrode positions in the international 10–20 electroencephalography system, corresponding to the left (C3) and right (C4) sensorimotor cortices; PSD, power spectral density.

Figure 5 Intervention adjustment flowchart (including learning curve) and dynamic changes in C3/C4 energy difference (including PSD from electrodes C3 and C4 and topographic maps of the affected brain during the 1st and 28th training sessions) of patient 4. The value in color bar of topographic map represents the energy subtracted from the baseline during the action, measured in µV2/Hz. C3/C4, standard electrode positions in the international 10–20 electroencephalography system, corresponding to the left (C3) and right (C4) sensorimotor cortices; PSD, power spectral density.

Figure 6 Intervention adjustment flowchart (including learning curve) and dynamic changes in C3/C4 energy difference (including PSD from electrodes C3 and C4 and topographic maps of the affected brain during the 1st and 28th training sessions) of patient 5. The value in color bar of topographic map represents the energy subtracted from the baseline during the action, measured in µV2/Hz. C3/C4, standard electrode positions in the international 10–20 electroencephalography system, corresponding to the left (C3) and right (C4) sensorimotor cortices; PSD, power spectral density.

Connectomics and functional brain network changes

Detailed network-level analysis was conducted for patient 4, who showed significant functional improvement and robust motor imagery ability. Preintervention imaging revealed damage to the left basal ganglia, affecting adjacent sensorimotor and language networks.

Post-training, functional connectivity was markedly enhanced in the following areas: (I) the primary sensorimotor cortex (L-1 and L-3b), involved in tactile processing; (II) the premotor cortex (L-22dd and L-22dv), responsible for coordinating trunk and limb movements; (III) the supplementary motor area (L-6ma and L-6mp), which is engaged by visual cues; (IV) and the language-related frontal lobe region (L-55), which is critical in speech generation and processing.

The connectivity matrix showed new links and increased coupling between the motor and language areas. Changes were also noted in the posterior attention network, indicating broader network plasticity (Figures 7-11).

Figure 7 Patient 4: (A) axial and (B) coronal T1 weighted MRI showed a fissured T1WI low signal shadow in the left basal ganglia region. MRI, magnetic resonance imaging; T1WI, T1-weighted imaging.

Figure 8 Pre- and post-evolution of pixels in the corresponding cortex of the sensorimotor and language networks in patient 4. Compared to the left column before training, the right column exhibited a significant increase in density or range (within the red box) of certain cortical pixel particles corresponding to functional connections in the patient’s brain network after training.

Figure 9 Pre- and post-evolution of the functional connectivity matrix of the language (blue) and sensorimotor (orange) networks in patient 4.

Figure 10 Pre- and post-evolution of pixels in the corresponding cortex of the dorsal attention network in patient 4. Compared to the left column before training, the right column exhibited a significant increase in density or range (within the red box) of specific cortical pixel particles corresponding to functional connections in the patient’s brain network after training.

Figure 11 Pre- and post-evolution of the functional connectivity matrix of the dorsal attention (orange) and sensorimotor (blue) network in patient 4.

For the other four patients, consistent patterns of disrupted connectivity in the perilesional motor and language networks were noted preintervention. Following training, functional reorganization was evident, with increased connectivity toward normalization patterns, particularly in the sensorimotor network. Standard deviation matrices with comparison to healthy controls showed gains in interregional connectivity, particularly in patient 4.

Discussion

The integration of artificial intelligence (AI) into tractography holds significant promise for advancing both neuroscientific research and clinical applications. Machine learning (ML) algorithms, such as gated recurrent units (GRUs), convolutional neural networks (CNNs), and random forests, have demonstrated the ability to overcome limitations of traditional tractography methods by reducing false positives, improving spatial extent estimation of white-matter (WM) bundles, and enhancing robustness to noise and anatomical variability. These advancements are particularly valuable in preoperative planning, where precise identification of critical WM pathways (e.g., corticospinal tract and arcuate fasciculus) is essential to minimizing postoperative neurological deficits. For instance, ML-driven tractography could enable surgeons to visualize individualized connectomes with higher accuracy, thereby avoiding damage to sensitive pathways during tumor resections or epilepsy surgery (7).

The effectiveness of motor imagery-based BCI training in limb rehabilitation for patients with brain dysfunction has been previously demonstrated (8-10). This technique reinforces neural circuits by simulating voluntary movement through motor imagery, which promotes plastic changes in the sensorimotor cortex. Since motor imagery involves cognitive processes without actual movement execution, motor imagery-based BCI training places higher demands on cognitive ability while requiring minimal residual motor function.

In this case series, BCI-based hand training led to significant improvements in upper limb motor function, with additional gains observed in lower limb movement, speech, and cognitive function. To our knowledge, this is the first study to examine multimodal neural functional remodeling induced by BCI training in subacute brain injury. The mechanisms underlying this remodeling include (I) the functional synergy between the cingulate motor area and frontal language regions, suggesting cross-network coordination; (II) changes in PSD in the 8- to 30-Hz band at the C3/C4 electrodes, which may serve as biomarkers of motor recovery; and (III) increased nodal degree within the sensorimotor network correlating with improvements in both language and cognitive function. These findings underscore the integrative nature of brain plasticity in rehabilitation.

The use of individualized connectomics analysis, incorporating 379 brain regions, allowed for precise mapping of neural changes and overcame the limitations of traditional region-of-interest methods. This approach also enabled synchronized tracking of EEG time-frequency characteristics and multi-network reorganization.

In the patients with aphasia and cognitive impairment—specifically patients 1, 3, and 4—BCI hand training was associated with improved spontaneous speech, speech fluency, comprehension, and naming. These improvements were supported by increased activation in the frontal language areas, particularly the L-55 region, and by strengthened connectivity across the language network. The use of repeated auditory prompts, speech feedback, and visual stimulation during training might have contributed to increased input-output efficiency within the language system.

Other research has shown that BCI-based rehabilitation can modulate auditory and attentional pathways and improve language processing through interactive neurofeedback (11,12). The observed gains in MMSE scores and language coherence suggest that BCI training may also enhance cognitive domains such as attention and executive function. In this study, three patients demonstrated improved task compliance and sustained attention during training, likely facilitated by human-machine interaction and immediate feedback, which are key mechanisms for strengthening neural coupling.

Furthermore, the improvement in language function may reflect not only cortical changes but also reorganization of subcortical networks. For example, patient 1, who had thalamic aphasia, showed a 50% improvement in MMSE score and marked functional recovery following the intervention. These results align with studies implicating the thalamus and related WM tracts in subcortical aphasia (13,14)

In patients with hemiplegia—specifically patients 2, 4, and 5—BCI training aimed at upper limb rehabilitation also contributed to partial recovery of lower limb function. This outcome may be explained by the activation of attentional and sensorimotor networks during focused motor imagery tasks. For instance, patient 4 showed activation of the dorsal attention network and improvements in motor imagery accuracy, as reflected by higher CA during training. Repetitive motor imagery with visual and tactile feedback likely enhanced functional coupling between the regions responsible for sensory integration and voluntary motor control.

In our study, the increased EEG activity in the sensorimotor cortex and more organized cortical topographies after training were consistent with previous reports (15,16). However, interindividual differences were notable. Patient 5, who had more severe damage to the corticospinal tract and reduced spontaneous joint activity at baseline, exhibited less improvement despite participating fully in the intervention. This highlights the importance of structural integrity and baseline function in determining response to BCI rehabilitation.

Our findings suggest that BCI training may facilitate recovery through various mechanisms, including functional compensation via the contralesional hemisphere, intra-hemispheric plasticity, and recruitment of alternative networks (17-20). In three patients, we observed decreases in motor area PSD in the affected hemisphere after training, suggesting improved motor imagery ability and cortical engagement. Additionally, connectomics analysis demonstrated enhanced interregional connectivity across sensorimotor, language, and attentional networks, indicating movement toward a more normalized network organization following intervention

Several limitations to this study should be noted. The most significant constraint is our small sample size (N=5) and absence of a control group, which precludes definitive conclusions regarding the efficacy of BCI in brain injury rehabilitation. Although we observed encouraging trends in functional recovery and neural plasticity, these findings must be considered exploratory and hypothesis-generating rather than confirmatory. In addition, detailed connectomics analysis was applied to only one patient due to variable imaging quality and data availability. No statistical power analysis was conducted, and all observed changes should be interpreted with caution.

Future research should aim to validate these findings in larger, controlled trials and to determine the optimal training protocols tailored to individual connectomics profiles. Integrating BCI with multimodal sensory feedback and identifying the ideal therapeutic window for intervention could further enhance recovery. As BCI technology evolves, its potential for facilitating personalized, multimodal neurorehabilitation strategies warrants continued investigation.

Conclusions

This case series study confirmed that for patients with brain dysfunction, motor imagery BCI training may effectively promote overall recovery in limb movement, language, and cognitive function. With the continuous development of BCI technology and the improvement of disease sources, conducting multicenter prospective cohort studies, establishing blank control groups to determine the specific effects of BCI, employing multimodal feedback to increase patient engagement (19,20), and determining the optimal intervention timing during rehabilitation (21) can be expected to produce notable insights in the field of clinical rehabilitation.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the AME Case Series reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1136/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1136/dss

Funding: This work was supported by Shanxi Province Higher Education Science and Technology Innovation Project (No. 2022L184).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1136/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was approved by the Ethics Committee of Shanxi Provincial People’s Hospital (2022 Province Medical Ethics Review No. 79), and written informed consent was obtained from all participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Wu Q, Ge Y, Ma D, Pang X, Cao Y, Zhang X, Pan Y, Zhang T, Dou W. Analysis of Prognostic Risk Factors Determining Poor Functional Recovery After Comprehensive Rehabilitation Including Motor-Imagery Brain-Computer Interface Training in Stroke Patients: A Prospective Study. Front Neurol 2021;12:661816. [Crossref] [PubMed]
2. Jiang Y, Yin J, Zhao B, Zhang Y, Peng T, Zhuang W, Wang S, Huang S, Zhong M, Zhang Y, Tang G, Shen B, Ou H, Zheng Y, Lin Q. Motor Imagery Brain-Computer Interface in Rehabilitation of Upper Limb Motor Dysfunction After Stroke. J Vis Exp 2023; [Crossref]
3. Pichiorri F, Morone G, Petti M, Toppi J, Pisotta I, Molinari M, Paolucci S, Inghilleri M, Astolfi L, Cincotti F, Mattia D. Brain-computer interface boosts motor imagery practice during stroke recovery. Ann Neurol 2015;77:851-65. [Crossref] [PubMed]
4. Nunes JD, Vourvopoulos A, Blanco-Mora DA, Jorge C, Fernandes JC, Bermudez I, Badia S, Figueiredo P. Brain activation by a VR-based motor imagery and observation task: An fMRI study. PLoS One 2023;18:e0291528. [Crossref] [PubMed]
5. Li Q, Feng J, Guo J, Wang Z, Li P, Liu H, Fan Z. Effects of the multisensory rehabilitation product for home-based hand training after stroke on cortical activation by using NIRS methods. Neurosci Lett 2020;717:134682. [Crossref] [PubMed]
6. Elam JS, Glasser MF, Harms MP, Sotiropoulos SN, Andersson JLR, Burgess GC, Curtiss SW, Oostenveld R, Larson-Prior LJ, Schoffelen JM, Hodge MR, Cler EA, Marcus DM, Barch DM, Yacoub E, Smith SM, Ugurbil K, Van Essen DC. The Human Connectome Project: A retrospective. Neuroimage 2021;244:118543. [Crossref] [PubMed]
7. Poulin P, Jörgens D, Jodoin PM, Descoteaux M. Tractography and machine learning: Current state and open challenges. Magn Reson Imaging 2019;64:37-48. [Crossref] [PubMed]
8. Nojima I, Sugata H, Takeuchi H, Mima T. Brain-Computer Interface Training Based on Brain Activity Can Induce Motor Recovery in Patients With Stroke: A Meta-Analysis. Neurorehabil Neural Repair 2022;36:83-96. [Crossref] [PubMed]
9. Nan W, Dias APB, Rosa AC. Neurofeedback Training for Cognitive and Motor Function Rehabilitation in Chronic Stroke: Two Case Reports. Front Neurol 2019;10:800. [Crossref] [PubMed]
10. Yuan Z, Peng Y, Wang L, Song S, Chen S, Yang L, Liu H, Wang H, Shi G, Han C, Cammon JA, Zhang Y, Qiao J, Wang G. Effect of BCI-Controlled Pedaling Training System With Multiple Modalities of Feedback on Motor and Cognitive Function Rehabilitation of Early Subacute Stroke Patients. IEEE Trans Neural Syst Rehabil Eng 2021;29:2569-77. [Crossref] [PubMed]
11. Baykara E, Ruf CA, Fioravanti C, Käthner I, Simon N, Kleih SC, Kübler A, Halder S. Effects of training and motivation on auditory P300 brain-computer interface performance. Clin Neurophysiol 2016;127:379-87. [Crossref] [PubMed]
12. Kleih SC, Gottschalt L, Teichlein E, Weilbach FX. Toward a P300 Based Brain-Computer Interface for Aphasia Rehabilitation after Stroke: Presentation of Theoretical Considerations and a Pilot Feasibility Study. Front Hum Neurosci 2016;10:547. [Crossref] [PubMed]
13. Barria P, Pino A, Tovar N, Gomez-Vargas D, Baleta K, Díaz CAR, Múnera M, Cifuentes CA. BCI-Based Control for Ankle Exoskeleton T-FLEX: Comparison of Visual and Haptic Stimuli with Stroke Survivors. Sensors (Basel) 2021;21:6431. [Crossref] [PubMed]
14. Latifi S, Mitchell S, Habibey R, Hosseini F, Donzis E, Estrada-Sánchez AM, Nejad HR, Levine M, Golshani P, Carmichael ST. Neuronal Network Topology Indicates Distinct Recovery Processes after Stroke. Cereb Cortex 2020;30:6363-75. [Crossref] [PubMed]
15. Guggisberg AG, Nicolo P, Cohen LG, Schnider A, Buch ER. Longitudinal Structural and Functional Differences Between Proportional and Poor Motor Recovery After Stroke. Neurorehabil Neural Repair 2017;31:1029-41. [Crossref] [PubMed]
16. Fridman EA, Beattie BJ, Broft A, Laureys S, Schiff ND. Regional cerebral metabolic patterns demonstrate the role of anterior forebrain mesocircuit dysfunction in the severely injured brain. Proc Natl Acad Sci U S A 2014;111:6473-8. [Crossref] [PubMed]
17. Hu Y, Gao T, Bai Y. An Exploratory Study on the Rehabilitation of Upper Limb Function After Stroke Using Brain-Computer Interface Combined with Multimodal Perceptual Feedback Training. Chinese Journal of Rehabilitation Medicine 2022;37:1457-62.
18. Miao Y, Chen S, Zhang X, Jin J, Xu R, Daly I, Jia J, Wang X, Cichocki A, Jung TP. BCI-Based Rehabilitation on the Stroke in Sequela Stage. Neural Plast 2020;2020:8882764. [Crossref] [PubMed]
19. Ma W, Liu W, Dong Y, Zhang J, Hao L, Xia T, Wang X, Han C. Predicting the prognosis of patients with renal cell carcinoma based on the systemic immune inflammation index and prognostic nutritional index. Sci Rep 2024;14:25045. [Crossref] [PubMed]
20. Kim MS, Park H, Kwon I, An KO, Kim H, Park G, Hyung W, Im CH, Shin JH. Efficacy of brain-computer interface training with motor imagery-contingent feedback in improving upper limb function and neuroplasticity among persons with chronic stroke: a double-blinded, parallel-group, randomized controlled trial. J Neuroeng Rehabil 2025;22:1. [Crossref] [PubMed]
21. Krueger J, Krauth R, Reichert C, Perdikis S, Vogt S, Huchtemann T, Dürschmid S, Sickert A, Lamprecht J, Huremovic A, Görtler M, Nasuto SJ, Tsai IC, Knight RT, Hinrichs H, Heinze HJ, Lindquist S, Sailer M, Millán JDR, Sweeney-Reed CM. Hebbian plasticity induced by temporally coincident BCI enhances post-stroke motor recovery. Sci Rep 2024;14:18700. [Crossref] [PubMed]

(English Language Editor: J. Gray)