A Bibliometric Analysis of Neuromodulation and Brain–Computer Interfaces for Pain Management (2000–2025): Hotspots, Emerging Trends, and Future Perspectives

Xiang Liu , Jiao Duan , Xiangge Liu , Fuwei Wang , Xiongjian Huang , Peiyu Tang , Jiabao Feng , Yan Gong , Qinghua Wu , Kunwei Li , Zurong Hu

Journal of Perioperative Innovation ›› : 1 -20.

PDF (10013KB)
Journal of Perioperative Innovation ›› :1 -20. DOI: 10.2738/JPI.2026.0005
Article
A Bibliometric Analysis of Neuromodulation and Brain–Computer Interfaces for Pain Management (2000–2025): Hotspots, Emerging Trends, and Future Perspectives
Author information +
History +
PDF (10013KB)

Abstract

Background: Brain–computer interfaces (BCIs) and neuromodulation technologies are increasingly being explored for pain management, yet the global research landscape, conceptual boundaries, and evolving technological trends remain incompletely characterized. This study aimed to systematically map the development, research hotspots, and emerging directions of this broader field from 2000 to 2025 and to distinguish emerging BCI- and closed-loop-related technologies from the broader literature.

Methods: A total of 1,269 English-language articles and reviews published between 2000 and October 2025 were retrieved from the Web of Science Core Collection. VOSviewer and CiteSpace were used to analyze publication trends, international and institutional collaboration, disciplinary structure, keyword co-occurrence and clustering, citation bursts, and highly cited references. Publication-growth models were compared using R2, adjusted R2, and the Akaike information criterion, and citation analyses were supplemented with publication-year-adjusted citations-per-year metrics. A focused technology-based sensitivity analysis was performed to distinguish strict BCI/BMI, closed-loop or adaptive neuromodulation, and BCI-enabling pain-decoding publications.

Results: Publication output showed sustained growth over the study period. The United States contributed the largest number of publications and showed the strongest international collaborative connectivity. Neurosciences and Clinical Neurology dominated the field, while Biomedical Engineering showed strong interdisciplinary centrality. The intellectual landscape remained strongly grounded in conventional neuromodulation approaches, including spinal cord stimulation and deep brain stimulation, while adaptive closed-loop systems, neural decoding, and BCI-related technologies emerged more prominently in recent years. In the focused sensitivity analysis, 308 candidate publications were assessed; 29 were classified as strict BCI/BMI, 42 as closed-loop or adaptive neuromodulation, and 9 as BCI-enabling pain decoding. Together, these 80 publications represented 6.3% of the complete corpus.

Conclusions: The literature is best characterized as a broad pain-related neuromodulation and neural-interface landscape rather than as a predominantly BCI-based field. Conventional neuromodulation remains the principal scientific and clinical foundation, while strict BCI, adaptive closed-loop systems, and brain-based pain decoding constitute smaller but increasingly visible technological streams. Bibliometric prominence reflects scholarly attention and technological development rather than established therapeutic superiority.

Keywords

brain-computer interface / pain management / neuromodulation / closed-loop system / neural decoding / bibliometric analysis

Highlight

• Pain-related neuromodulation and neural-interface research expanded substantially between 2000 and 2025.

• Conventional neuromodulation remains the major intellectual foundation, while adaptive control and neural decoding have emerged as more recent research streams.

• A focused sensitivity analysis identified only 29 strict BCI/BMI publications within the 1,269-record corpus.

• Biomedical engineering acts as a critical interdisciplinary bridge, translating basic neuroscience mechanisms into clinical interventions.

Cite this article

Download citation ▾
Xiang Liu, Jiao Duan, Xiangge Liu, Fuwei Wang, Xiongjian Huang, Peiyu Tang, Jiabao Feng, Yan Gong, Qinghua Wu, Kunwei Li, Zurong Hu. A Bibliometric Analysis of Neuromodulation and Brain–Computer Interfaces for Pain Management (2000–2025): Hotspots, Emerging Trends, and Future Perspectives. Journal of Perioperative Innovation 1-20 DOI:10.2738/JPI.2026.0005

登录浏览全文

4963

注册一个新账户 忘记密码

1 Introduction

Chronic pain—persistent for more than three months—afflicts up to 20% of the global adult population and stands as one of the leading causes of disability worldwide[1,2]. The Global Burden of Disease (GBD) study, for instance, projected that low back pain alone—a primary driver of chronic disability—will affect over 800 million people by 2050[3]. Beyond its physical toll, chronic pain is strongly correlated with functional limitations, reduced quality of life, and high rates of comorbid anxiety and depression[4]. For decades, pharmacological interventions, particularly opioid analgesics, have been a cornerstone of severe pain management[2,5,6]. Yet a landmark NIH-commissioned systematic review found “insufficient” evidence for their long-term efficacy, while risks—addiction, overdose, and mortality—have fueled a public health crisis[7]. This therapeutic void has spurred an urgent search for safer, non-addictive alternatives.

In response to this gap, neuromodulation technologies have emerged as important non-pharmacological approaches[8-13]. Conventional modalities such as spinal cord stimulation (SCS) and deep brain stimulation (DBS) typically deliver fixed or manually adjusted “open-loop” stimulation and have been used for refractory neuropathic pain[9,14]. However, static stimulation may fail to account for the dynamic nature of pain processing, neural plasticity, body position, and other physiological changes, contributing to variable efficacy and tolerance over time[15-17]. These limitations have motivated adaptive and closed-loop systems in which sensed neural, physiological, or sensor-derived information is used to modify therapeutic output automatically[14,16-22]. At the interface level, flexible high-density microelectrode arrays are increasingly being developed to provide high-resolution neural recording[23], stimulation[24], or both[24], and represent an important enabling technology for closed-loop brain–machine interfaces[25-27]. Closed-loop neuromodulation, however, is not synonymous with a brain–computer interface. A strict BCI acquires brain-derived signals and translates the resulting information into contingent feedback, device control, neurofeedback, or therapeutic output[9,13,28]. For example, ECAP-controlled SCS represents closed-loop spinal neuromodulation rather than strict BCI because its control signal is derived from spinal neural responses rather than brain activity[11,22]. Emerging pain-related BCI studies have used EEG, MEG, functional neuroimaging, or intracranial recordings for neurofeedback, pain-state decoding, or brain-derived therapeutic control[18-21].

The field is inherently interdisciplinary, bridging neurosciences (pain mechanisms), clinical neurology (patient outcomes), and biomedical engineering (device design and signal processing). Publication volume has increased substantially, yet the intellectual architecture remains incompletely characterized: Who leads globally? Which collaborations drive breakthroughs? How have research hotspots evolved across conventional neuromodulation, adaptive systems, and brain-derived interfaces? To date, bibliometric analyses have focused on related but separate domains: “pain management” broadly[29-31], “BCI in rehabilitation”[32-34], or specific neuromodulation modalities such as spinal cord stimulation[35]. However, the integrated landscape linking pain neuroscience, conventional neuromodulation, closed-loop control, neural decoding, and emerging BCI technologies remains insufficiently characterized.

This study aims to address that gap by providing a comprehensive bibliometric analysis of the global research landscape on this topic from 2000 to 2025. VOSviewer and CiteSpace were used to visualize co-authorship networks, keyword clusters, citation bursts, and disciplinary overlays. Our objectives were threefold: (1) quantify publication trends and global contributions; (2) characterize the relationship among established neuromodulation, adaptive closed-loop systems, and emerging BCI-related technologies; and (3) identify emerging hotspots, such as closed-loop systems, ECAP-guided feedback, and AI-enabled prediction, to guide future research, collaboration, and clinical translation. Because the primary search strategy was intentionally broad, we further performed a focused technology-based sensitivity analysis to distinguish strict BCI/BMI, closed-loop or adaptive neuromodulation, and BCI-enabling pain-decoding publications from the broader bibliometric corpus.

2 Methods

2.1 Data Source and Search Strategy

The Web of Science Core Collection was chosen as the data source based on several factors detailed in the existing literature[36,37]. This study employed a keyword-based search strategy, applying the following terms to the database: TS=(pain OR Neuralgia OR Arthralgia OR Hyperalgesia OR Headache OR Analgesi*) AND (TS=(“brain-computer interface*” OR “brain-machine interface*” OR “neural interface*”) OR (TS=(“transcranial magnetic stimulation” OR “repetitive transcranial magnetic stimulation” OR “deep transcranial magnetic stimulation” OR “single-pulse transcranial magnetic stimulation” OR “rapid rate transcranial magnet stimulation” OR “transcranial direct current stimulation” OR “transcranial alternating current stimulation” OR “motor cortex stimulation” OR “Deep brain stimulation” OR “transcutaneous electrical nerve stimulation” OR “spinal cord stimulation” OR “invasive brain stimulation” OR “noninvasive brain stimulation” OR “non-invasive brain stimulation” OR “invasive neuromodulation” OR “non-invasive neuromodulation” OR “noninvasive neuromodulation”) AND TS=(feedback OR “closed-loop” OR adaptive OR bidirectional OR neurofeedback OR regulate OR modulate))). The literature search was restricted to original research articles and review articles published between January 1, 2000, and October 1, 2025, and the language was restricted to English. The initial search yielded 1,269 records, which included 987 original research articles and 282 reviews. These records, exported using the WoSCC “Full Record and Cited References” option as .txt files, were subsequently subjected to a comprehensive analysis. The specific steps of this process are illustrated in Figure 1.

The search was conducted in the Science Citation Index Expanded (SCI-EXPANDED). Original research articles and reviews were retained as the primary eligible publication types. Records classified solely as proceeding papers, editorial materials, meeting abstracts, corrections, letters, or retracted publications were excluded. A small number of retained Article or Review records carried additional secondary Web of Science document-type tags, including Proceedings Paper, Early Access, or Book Chapter, and remained in the corpus because their primary eligible document type was Article or Review.

Records from separate export files were merged and checked for duplication using the unique Web of Science accession number (UT) as the primary identifier, with DOI and article title used for secondary verification when necessary. Importantly, the 1,269-record corpus was treated as a broad bibliometric landscape rather than as a collection of strict BCI studies. The search strategy was intentionally broad to capture pain-related neuromodulation, neural interfaces, adaptive stimulation, and emerging BCI applications; a focused technology-based sensitivity analysis was therefore added to quantify the relative representation of strict BCI/BMI, closed-loop or adaptive neuromodulation, and BCI-enabling pain-decoding publications.

2.2 Conceptual Scope and Operational Definitions

To address the conceptual heterogeneity of the retrieved literature and to avoid conflating conventional neuromodulation with closed-loop systems or strict BCI technologies, operational definitions were applied for the focused sensitivity analysis and subsequent interpretation.

Conventional or open-loop neuromodulation was defined as electrical or magnetic stimulation delivered using fixed, predefined, or manually adjusted parameters without real-time neural, physiological, behavioral, or sensor-derived information being used to modify stimulation automatically. Representative approaches included conventional spinal cord stimulation (SCS), deep brain stimulation (DBS), repetitive transcranial magnetic stimulation (rTMS), transcranial direct current stimulation (tDCS), and motor cortex stimulation.

Closed-loop or adaptive neuromodulation was defined as a system in which neural, physiological, behavioral, or sensor-derived information was acquired and used automatically to modify stimulation or therapeutic output according to the measured state. Closed-loop neuromodulation was not considered synonymous with BCI. For example, evoked compound action potential (ECAP)-controlled SCS was classified as closed-loop neuromodulation because spinal neural responses regulate stimulation automatically, but it was not classified as a strict BCI because the control signal is not brain-derived. Systems using body-position or accelerometry signals, respiration, or other non-brain physiological inputs to adapt stimulation automatically were classified within the same closed-loop/adaptive category.

A strict brain–computer interface or brain–machine interface (BCI/BMI) was defined as a system that: (1) acquires brain-derived signals, including electrophysiological activity such as EEG, ECoG, intracranial local field potentials, or neuronal spiking activity, or other functional brain-derived signals; (2) processes or decodes these signals in real time or near real time; and (3) translates the resulting brain-derived information into contingent feedback, external device control, communication, neurofeedback, or therapeutic output. For the focused pain-related analysis, pain or nociception was additionally required to represent a primary target, outcome, or decoded state.

Studies that used brain-derived signals to identify, classify, or predict pain or nociceptive states but did not use the decoded information to generate subsequent contingent feedback or therapeutic output were classified separately as BCI-enabling pain-decoding studies. This distinction was made because neural decoding alone does not constitute a complete closed-loop BCI system.

These definitions were used to guide the focused sensitivity analysis and the interpretation of the bibliometric findings; they did not alter the composition of the primary 1,269-record corpus used for publication, collaboration, keyword, disciplinary, and citation-network analyses.

2.3 Bibliometric Analysis and Visualization

For subsequent analysis and visualization, the study data were processed with the bibliometric tools provided by CiteSpace and VOSviewer.

2.3.1 VOSviewer Configuration

In VOSviewer(version 1.6.20), diverse nodes were used to depict various entities, including countries, institutions, journals, authors, references, and keywords. The magnitude of each node was directly proportional to its number of publications, citations, or frequency of occurrence. Distinct colors were used to represent different clusters or average years of occurrence, while the thickness of the connecting lines indicated the strength of the relationships between the nodes. The parameters of VOSviewer were modified to perform the following tasks: create maps based on the bibliometric data, read data from the bibliographic database files, use the count function in the full counting method (full counting), and adjust item thresholds according to different analysis items.

2.3.2 CiteSpace Configuration

For the cluster analysis of keywords and cited documents, as well as the detection of citation bursts, CiteSpace (version 6.4.R1, Advanced) was used with the following settings: data source (WoSCC), link retention factor (3.0), lookback years (5), parameter e (1.0), time span (January 2000 to October 1, 2025; 1-year slices), link strength (cosine, intraslice), selection criterion (g-index, k=25), and minimum duration (2 for keywords, 5 for references).

2.4 Sensitivity Analysis, Growth Modeling, and Citation Normalization

2.4.1 Technology-Based Sensitivity Analysis

Because the primary search strategy was intentionally broad, a post hoc technology-based sensitivity analysis was conducted within the 1,269-record corpus. A high-sensitivity prescreen of titles, abstracts, Author Keywords, and Keywords Plus identified 303 publications potentially related to BCI/BMI, neurofeedback, closed-loop or adaptive neuromodulation, or brain-based pain decoding. A subsequent reverse audit using broader adaptive-control terminology identified five additional records, yielding a final candidate set of 308 publications.

The 308 candidate publications were classified according to the operational definitions described in Section 2.2 into four mutually exclusive categories: strict BCI/BMI, closed-loop/adaptive neuromodulation, BCI-enabling pain decoding, or not meeting the focused criteria. Classification was based primarily on titles, abstracts, keywords, and bibliographic information, with full-text or publicly accessible source verification when the technological architecture was unclear. Candidate records were independently evaluated by two investigators according to the predefined criteria, with discrepancies resolved through consensus or adjudication by a third senior investigator. This analysis was intended to quantify the representation of BCI-, decoding-, and closed-loop-related technologies within the broader bibliometric landscape rather than to constitute a systematic review or comparative evaluation of therapeutic efficacy.

Separately, exploratory keyword-based temporal analyses were conducted to characterize annual publication trends associated with closed-loop-related and artificial-intelligence-related terminology; these results are presented in Supplementary Figures 1 and 2.

2.4.2 Publication-Growth Modeling

Annual and cumulative publication counts were used to characterize temporal changes in research activity. To avoid relying solely on the coefficient of determination of a single prespecified model, linear, second-order polynomial, and exponential growth models were compared using goodness-of-fit measures including R2, adjusted R2, and the Akaike information criterion (AIC). The fitted models were used descriptively to characterize historical publication patterns within the observed study period and were not used to extrapolate or forecast future publication output.

2.4.3 Citation Normalization

Because raw citation counts are strongly influenced by publication age, total citation counts were supplemented by a publication-year-adjusted citations-per-year (CPY) metric: CPY = Total citations / (2025 – publication year + 1). Where citation-based comparisons were made across countries, institutions, or other groups, normalized citation measures were interpreted alongside raw citation counts. Because citation distributions are typically right-skewed and may be disproportionately influenced by a small number of highly cited publications, group-level citation metrics were summarized using medians and interquartile ranges where applicable; 95% confidence intervals were additionally reported for normalized estimates when appropriate. Citation-based indicators were interpreted as measures of scholarly visibility and academic attention rather than as direct indicators of methodological quality, clinical efficacy, or overall research quality.

3 Results and Analysis

3.1 Publishing Characteristics

A total of 1,269 publications (including original articles and reviews) related to pain-related neuromodulation, neural interfaces, and BCI-related technologies were included in the analysis. Figure 2 presents the annual and cumulative publication trends from 2000 to October 1, 2025. Because the search was finalized in October 2025, the 2025 publication count represents a partial-year value. Research output remained relatively limited during the early study period but showed an overall upward trend thereafter, despite year-to-year fluctuations. The cumulative number of publications closely followed a second-order polynomial model (y = 2.6742x2 – 23.417x + 55.44). Among the tested models, the second-order polynomial provided the best descriptive fit (R2 = 0.9989; adjusted R2 = 0.9989; AIC = 137.9), compared with the exponential model (R2 = 0.9861; adjusted R2 = 0.9855; AIC = 203.1) and the linear model (R2 = 0.8804; adjusted R2 = 0.8755; AIC = 259.0). These models were used only to characterize historical publication patterns and were not used for forecasting.

3.2 Countries/Regions

A total of 62 countries and regions have contributed to research on pain-related neuromodulation, neural interfaces, and BCI-related technologies. Table 1 lists the top 10 most productive countries or regions, while Figure 3 illustrates a national-level collaboration network comprising entities with at least ten publications. The network includes 23 countries and exhibits a total link strength of 1502, indicating substantial international collaboration within the retrieved literature.

Leading the field by a large margin is the United States (USA), contributing over 41% of all publications (n=522). The USA also possesses the highest Total Link Strength (325), indicating both high publication output and extensive international collaboration (Figure 3). Following the US in production volume is China (Rank 2, n=173), which accounts for 13.63% of the total output. China demonstrates an Average Citations per Publication (AC/P) of 17.31. Italy (Rank 3, n=127) and England (Rank 4, n=125) follow closely, with England exhibiting high collaborative connectivity (Total Link Strength 155). Raw average citations per publication were higher in Germany (69.10), France (65.98), and Belgium (61.13) than in several countries with larger publication volumes, including the USA (42.71). After adjustment for publication year, mean citations per year (CPY) also varied across countries; however, confidence intervals were wide for several smaller publication groups, and median CPY values were less markedly different than raw citation averages. These citation indicators were therefore interpreted as measures of scholarly visibility rather than as direct measures of national research quality or clinical impact. Detailed time-normalized citation metrics are provided in Supplementary Table S1.

3.3 Institutions

A total of 1964 research institutions were involved in this field. Table 2 identifies the top 10 institutions based on publication count. In the network visualization, the size of each node indicates the publication output of these institutions. Harvard Medical School and Harvard University were identified as separate bibliometric entities and contributed 42 and 37 publications, respectively, followed by the University of Oxford (n=31), the University of Michigan (n=26), and Aalborg University (n=24).

Figure 4 presents the co-authorship network, indicating strong collaborative ties among institutions in pain-related neuromodulation and BCI-related research. Among the top ten institutions, Johns Hopkins University received the highest unadjusted average citation count. Figure 4b, generated via VOSviewer, maps 165 institutions that met the minimum publication threshold. In this network, the University of California, San Francisco shows the greatest total link strength (693), followed by Harvard University (505). These citation indicators were interpreted as measures of scholarly visibility rather than direct measures of institutional research quality.

3.4 Research Hot Spot

Keywords serve as key indicators of article content. Analyzing their co-occurrence maps the intellectual structure of a field, identifying predominant themes and conceptual linkages. Table 3 lists the frequencies of the top 10 keywords. Analysis of Table 3 reveals the most frequent research themes. The most frequent keywords were “neuropathic pain” (n=284), “spinal cord stimulation” (n=241), “pain” (n=227), “deep brain stimulation” (n=177), and “chronic pain” (n=158). Notably, the generic term “electrical stimulation” exhibits the highest centrality (0.17) among all keywords, followed by “pain” (0.12) and “deep brain stimulation” (0.11). These findings indicate that pain mechanisms and established neuromodulation approaches constitute a substantial part of the intellectual foundation of the retrieved literature.

Utilizing the co-occurrence graph depicted in Figure 5, keyword clustering was performed, forming 11 distinct groups using the log-likelihood ratio (LLR) method. The results, detailed in Table 4, suggest several broad temporal layers based on cluster mean year. (1) The early research foundation (2006–2007) centered on understanding fundamental pain mechanisms, as evidenced by clusters investigating “chronic pain” (#6, 2007) and “opioid receptors” (#7, 2006). (2) Subsequently, prominent clusters involved established neuromodulation technologies, represented by “transcranial magnetic stimulation” (#0, 2010) and “deep brain stimulation” (#1, 2012). (3) More recent cluster mean years were observed for “transcranial direct current stimulation” (#2, 2016), “brain-computer interface” (#4, 2016), “spinal cord stimulation” (#3, 2017), and “pain management” (#8, 2019). Importantly, cluster mean year reflects the temporal concentration of publications within a cluster rather than the date at which a technology was first introduced.

Overall, the clustering results indicate that BCI constitutes a distinct and relatively recent thematic component embedded within a broader research landscape encompassing pain mechanisms and multiple forms of neuromodulation, rather than representing the entire field itself.

3.5 Keywords Burst

The emergence of keywords with high citation bursts illuminates the research frontiers that have attracted significant scholarly attention over time. Figure 6 presents the top 20 keywords with the strongest citation bursts, arranged chronologically. The observed burst patterns can be summarized into three broad temporal patterns:

(1) Foundational Techniques (2006–2014): The focus centered on terms such as “motor cortex stimulation”, “evoked potentials”, and “electrical stimulation”. This period highlights the establishment and exploration of foundational neuromodulation techniques and the objective physiological signals used to measure their effects.

(2) Human & Network Focus (2010–2018): Subsequently, terms like “humans”, “prefrontal cortex”, and “functional connectivity” gained prominence. This pattern indicates increasing attention to human studies and network-level mechanisms of pain modulation.

(3) Recent Clinical Evidence and Safety Focus (2019–2025): More recently, the field has focused on key terms such as “safety”, “meta-analysis”, and “chronic back” (referring to chronic back pain). These terms suggest increasing attention to clinical evidence synthesis, safety evaluation, and specific high-burden pain conditions.

These burst patterns characterize shifts in scholarly attention over time and should not be interpreted as evidence that more recent technologies have replaced established neuromodulation approaches or demonstrated superior clinical efficacy.

3.6 Disciplinary Research

According to CiteSpace, 96 research categories connected by 213 links were identified (Figure 7a). As shown in Table 5, the top ten subject categories in the retrieved pain-related neuromodulation and BCI literature are led by Neurosciences (n=708), Clinical Neurology (n=469), and Anesthesiology (n=117). In terms of betweenness centrality, the leading disciplines are Neurosciences (BC=0.63), Clinical Neurology (BC=0.28), and Engineering, Biomedical (BC=0.27). The field is inherently multidisciplinary, with Neurosciences and Clinical Neurology playing a central role, reflecting the strong neurological and clinical foundations of pain research. Other significant supporting disciplines include Medicine, Research & Experimental, Rehabilitation, and Surgery, among others.

Furthermore, we analyzed trends in disciplinary evolution (Figure 7b). Earlier bursts involved clinically oriented categories such as Rehabilitation, Surgery, and Rheumatology, followed by increasing representation of neuroimaging-related disciplines. Beginning around 2013, disciplines such as Integrative & Complementary Medicine, Multidisciplinary Sciences, and psychology-related categories experienced citation bursts, consistent with increasing disciplinary diversification and the multidimensional nature of pain research. More recent bursts also involved broader biological and multidisciplinary scientific categories. Overall, these findings indicate increasing disciplinary diversification rather than a unidirectional transition toward a single technological domain.

3.7 Reference Analysis

As shown in Table 6, ten highly cited papers were identified within the 1,269-publication primary corpus. The most frequently cited paper (2017 citations) was authored by Zhang W et al. Entitled OARSI recommendations for the management of hip and knee osteoarthritis, Part II: OARSI evidence-based, expert consensus guidelines, it was published in Osteoarthritis and Cartilage in 2008[38]. The prominence of a clinical pain-management guideline among the most highly cited publications further illustrates the breadth of the primary corpus.

Notably, among these ten highly-cited papers, several address the fields of neuromodulation and pain management. For instance, Ossipov et al. (2014) summarized that dysfunction of the descending pain-modulatory circuits — particularly the attenuation of inhibitory effects and predominance of facilitatory influences — represents a key mechanism underlying the transition from acute to chronic pain. Integrating evidence from animal models, human brain imaging and clinical pharmacology, they elucidated the interactions between these circuits and higher-order brain centres, and indicated that mechanistic insights in this regard may inform novel therapeutic strategies for chronic pain[39]. Deer et al. (2017) compared dorsal root ganglion stimulation versus conventional spinal cord stimulation for chronic neuropathic pain of the lower extremities. Their results demonstrated that dorsal root ganglion stimulation yielded a significantly higher success rate of pain relief (≥50% pain reduction) at 3 months, with more precise therapeutic targeting and less postural-driven paresthesia fluctuation[40]. Osborn et al. (2018) developed a bio-inspired multilayer electronic skin (e-dermis). Via neuromorphic-like interfaces and transcutaneous electrical nerve stimulation (TENS), the system enabled amputees with phantom-limb sensation to perceive graded tactile and nociceptive inputs ranging from innocuous to noxious, allowing discrimination of object curvature and sharpness, and delivering more natural multi-dimensional sensory feedback for prosthetic hands[41].

These highly-cited papers reflect research hotspots in neuromodulation, non-invasive brain stimulation, pain mechanisms and prosthetic technologies. Together, they illustrate the heterogeneous scientific foundations of the field rather than a literature composed exclusively of strict BCI studies; citation prominence was interpreted as scholarly visibility rather than evidence of clinical superiority.

CiteSpace was used to identify the top 25 references with the strongest citation bursts (Fig. 8). The three publications with the highest burst-strength values were as follows: Mekhail N (2020), The Lancet Neurology, “Long-term safety and efficacy of closed-loop spinal cord stimulation to treat chronic back and leg pain (Evoke): a double-blind, randomised, controlled trial”(burst strength: 14.07, burst period: 2022–2025)[11]; Lefaucheur JP (2020), “Evidence-based guidelines on the therapeutic use of rTMS: An update (2014–2018)” (burst strength: 13.9, burst period: 2022–2025)[42]; Kapural L (2015), “Novel 10-kHz high-frequency therapy (HF10 therapy) is superior to traditional low-frequency spinal cord stimulation for the treatment of chronic back and leg pain: the SENZA-RCT randomized controlled trial” (burst strength: 13.5, burst period: 2017–2020)[43].

These burst references represent conceptually distinct forms of neuromodulation. In particular, the Mekhail study represents closed-loop spinal neuromodulation rather than a strict BCI because stimulation is regulated using spinal ECAP feedback rather than brain-derived signals. The Lefaucheur and Kapural publications represent established non-invasive and spinal stimulation approaches, respectively. Their coexistence among influential burst references supports interpretation of the dataset as a broad neuromodulation and neural-interface landscape in which closed-loop systems and BCI-related approaches constitute more recent but distinct research streams.

Furthermore, the reference with the highest betweenness-centrality identified by CiteSpace was produced by Lefaucheur JP (2014, Clin Neurophysiol)[12], a comprehensive evidence-based guideline on the therapeutic applications of rTMS developed by a panel of European experts. This foundational publication systematically appraised clinical evidence across multiple neurological and psychiatric disorders, and provided a Grade A (definite efficacy) recommendation for the analgesic effects of high-frequency rTMS over the contralateral primary motor cortex. Its high betweenness centrality indicates a bridging role across research domains involving neurophysiology, clinical neurology, pain management, and non-invasive neuromodulation; centrality was interpreted as structural connectivity within the citation network rather than as evidence of clinical superiority.

3.8 Focused Sensitivity Analysis of BCI and Closed-Loop Technologies

A focused technology-based sensitivity analysis was conducted to determine the relative representation of strict BCI/BMI, closed-loop or adaptive neuromodulation, and brain-based pain decoding within the broader 1,269-publication corpus.

The initial high-sensitivity prescreen identified 303 potentially relevant publications. A subsequent reverse audit of the remaining records, using broader adaptive-control terminology, identified five additional candidate publications, resulting in a final focused screening set of 308 records. This focused analysis provides a more specific assessment of the representation of BCI-related and adaptive neuromodulation technologies within the broader bibliometric corpus.

Following structured record-level classification using the predefined operational criteria, 80 publications met at least one of the three focused technological definitions, corresponding to 26.0% of the 308 candidate publications and 6.3% of the complete 1,269-record corpus.

Twenty-nine publications were classified as strict BCI/BMI, representing 9.4% of the focused candidate set and 2.3% of the complete corpus. These publications used brain-derived signals within real-time or near-real-time contingent architectures for neurofeedback, device control, or therapeutic intervention in which pain or nociception represented a primary target, outcome, or decoded state.

Forty-two publications were classified as closed-loop or adaptive neuromodulation, representing 13.6% of the focused candidate set and 3.3% of the complete corpus. These systems automatically modified therapeutic output according to non-brain neural, physiological, behavioral, or sensor-derived information. Representative control signals included spinal evoked compound action potentials, body-position or accelerometry signals, respiration, and other physiological feedback. These systems were therefore distinguished from strict BCI despite sharing an adaptive feedback-control architecture.

Nine publications were classified as BCI-enabling pain-decoding studies, representing 2.9% of the focused candidate set and 0.7% of the complete corpus. These studies used brain-derived signals to classify, predict, or decode pain or nociceptive states but did not use the decoded information to control a subsequent therapeutic or feedback output.

The remaining 228 candidate publications (74.0%) did not meet the focused criteria. Common reasons included conventional fixed or manually adjusted neuromodulation, use of EEG or neuroimaging only as offline outcome measures, BCI applications directed primarily toward motor rehabilitation or communication rather than pain, and neural or physiological biomarkers proposed for future adaptive systems without implementation of automatic feedback control.

The focused technological categories were concentrated in the more recent literature. Among strict BCI/BMI publications, 22 of 29 (75.9%) were published between 2020 and 2025, with a median publication year of 2022. Among closed-loop or adaptive neuromodulation publications, 33 of 42 (78.6%) were published during the same period, with a median year of 2023. Five of the nine BCI-enabling pain-decoding publications (55.6%) were published between 2020 and 2025, with a median publication year of 2023.

Overall, the sensitivity analysis confirmed that strict BCI/BMI represented only a small component of the primary bibliometric corpus. Even when strict BCI/BMI, closed-loop or adaptive neuromodulation, and BCI-enabling pain decoding were considered together, these focused technological categories accounted for only 6.3% of the complete dataset. The primary 1,269-publication corpus should therefore be interpreted as a broad pain-related neuromodulation and neural-interface landscape within which strict BCI, adaptive closed-loop systems, and brain-based pain decoding constitute smaller but increasingly visible technological streams.

Exploratory keyword-based temporal analyses also showed increasing annual publication activity associated with closed-loop-related and artificial-intelligence-related terminology in recent years (Supplementary Figures 1 and 2). These descriptive trends were analyzed separately from the technology-based classification described above and should not be interpreted as equivalent to the strict BCI/BMI or closed-loop/adaptive categories.

4 Discussion

This bibliometric analysis provides a comprehensive mapping of the intellectual structure and research landscape linking pain, neuromodulation, neural interfaces, and emerging brain–computer interface (BCI) technologies from 2000 to 2025. By integrating quantitative metrics from 1,269 publications, we characterize key trends, interdisciplinary synergies, and global patterns while distinguishing established neuromodulation from newer adaptive and brain-derived interface technologies. The following sections synthesize these findings, contextualize them within broader pain research, and outline implications for future investigations.

4.1 General Overview and Core Findings

Chronic pain represents a formidable challenge to global health, with traditional pharmacotherapies often proving inadequate or fraught with risk[2,6]. In this context, neuromodulation and BCI-related technologies represent an expanding therapeutic and research landscape rather than a single unified treatment paradigm. The present analysis of 1,269 publications from the Web of Science Core Collection (2000–Oct 2025) showed substantial growth in research activity, with the second-order polynomial model providing the best descriptive fit to the observed cumulative publication trajectory. However, the focused sensitivity analysis demonstrated that only 29 publications (2.3% of the complete corpus) met the strict BCI/BMI definition, 42 (3.3%) were classified as closed-loop or adaptive neuromodulation, and 9 (0.7%) as BCI-enabling pain decoding. Together, these three categories accounted for 80 publications, or 6.3% of the complete corpus. Accordingly, the primary literature is better interpreted as a broad pain-related neuromodulation and neural-interface landscape in which adaptive control, neural decoding, and strict BCI constitute smaller but increasingly visible technological streams. The interdisciplinary structure remains evident, with Neurosciences (centrality 0.63), Clinical Neurology (0.28), and Engineering, Biomedical (0.27) occupying prominent positions in the disciplinary network. These bibliometric patterns describe scholarly activity and intellectual organization; they do not establish comparative therapeutic efficacy or clinical superiority.

4.2 From Conventional Neuromodulation to Closed-Loop Systems and Brain–Computer Interfaces

The intellectual evolution identified by keyword clustering and citation-burst analysis is best interpreted as a technological continuum comprising several overlapping layers rather than as a linear succession in which newer technologies replace earlier approaches. Cluster mean years indicate the temporal concentration of publications within thematic groups and therefore should not be interpreted as the dates at which individual technologies were introduced or as direct evidence of a completed paradigm shift.

Phase I (c. 2006–2007): Foundational pain mechanisms and clinical context. The earliest clusters focused on core pain mechanisms, including “#6: chronic pain” (Mean Year 2007) and “#7: opioid receptors” (Mean Year 2006). These themes developed during a period of increasing clinical concern regarding the limitations and risks of long-term opioid therapy[44-48]. This historical context helped sustain interest in non-pharmacological alternatives, but the temporal coexistence of these developments should not be interpreted as evidence that the opioid crisis directly caused the growth of neuromodulation research.

Phase II (c. 2010–2012): Rise of established open-loop neuromodulation. A second thematic layer involved the development and clinical consolidation of stimulation technologies, reflected by clusters for “#0: transcranial magnetic stimulation” (2010), “#1: deep brain stimulation” (2012), and “#5: multiple waveforms” (2012), together with the keyword burst for “motor cortex stimulation”. The operation of these therapies was characterized predominantly by an open-loop architecture, meaning that stimulation was delivered using parameters that were pre-set or intermittently adjusted rather than continuously controlled by real-time biological feedback. These approaches remain clinically relevant in their own right and provide much of the mechanistic and clinical foundation on which adaptive systems have subsequently been developed. The maturation of this era is exemplified by the high-centrality guidelines of Lefaucheur et al.[12], which systematically established an evidence base for rTMS in pain management. Here, citation-network centrality is interpreted as a measure of structural connectivity rather than therapeutic superiority.

Phase III (c. 2016–2019): Emergence of adaptive closed-loop systems and BCI-related research. More recent clusters include “#4: brain-computer interface” (2016), “#3: spinal cord stimulation” (2017), and “#8: pain management” (2019), indicating increasing scholarly attention to adaptive control and neural-interface technologies. One of the clearest clinically developed examples is ECAP-controlled spinal cord stimulation. Mekhail et al.[11] reported a randomized controlled trial in which spinal evoked compound action potentials were used as a feedback signal to adjust stimulation automatically. This architecture represents closed-loop spinal neuromodulation rather than a strict BCI because the control signal is derived from spinal neural responses rather than brain activity. The sensitivity analysis reinforces this distinction: 42 publications were classified as closed-loop or adaptive neuromodulation, compared with 29 strict BCI/BMI publications and 9 BCI-enabling pain-decoding studies. Strict pain-related BCI remains less clinically mature and is more frequently represented by neurofeedback, feasibility, protocol, and proof-of-concept studies. For example, Shirvalkar et al.[49] demonstrated individualized prediction of spontaneous chronic pain states from intracranial recordings in four participants, an important proof of concept for brain-derived pain decoding but not evidence of therapeutic efficacy for a brain-signal-driven closed-loop intervention. Thus, the more defensible interpretation is technological convergence among conventional neuromodulation, physiological feedback control, neural decoding, and brain-derived interfaces rather than a completed transition from open-loop stimulation to BCI-based therapy. Longer-term randomized follow-up of the EVOKE cohort reported sustained differences between ECAP-controlled closed-loop and fixed-output open-loop SCS through 36 months[50]. A later prospective subgroup analysis also evaluated ECAP-controlled closed-loop SCS in chronic nonsurgical refractory back pain[51].

4.3 The Interdisciplinary Landscape: Bridging Engineering, Neuroscience, and Psychology

The convergence of established neuromodulation, adaptive closed-loop systems, and emerging BCI technologies is fundamentally interdisciplinary. Our analysis of subject categories (Table 5) reveals the field’s collaborative structure. Neurosciences (Centrality 0.63) and Clinical Neurology (0.28) provide major mechanistic and clinical foundations, whereas Engineering, Biomedical (Centrality 0.27) shows comparatively high network centrality despite a smaller publication volume (n=69). This pattern is consistent with a bridging role for biomedical engineering across device design, signal acquisition, signal processing, feedback algorithms, and clinical translation. Furthermore, the disciplinary burst analysis (Figure 7b) highlights the field’s expanding perspective. The burst of Psychology and related fields beginning around 2013 is consistent with growing recognition of the biopsychosocial and multidimensional nature of pain. For future adaptive interfaces, this implies that clinically useful systems may need to account not only for sensory-discriminative features but also for affective and cognitive dimensions of the pain experience.

4.4 Global Research Landscape: Contributions, Collaboration, and Citation Visibility

The global distribution of research (Table 1) reveals distinct patterns of publication output and collaborative connectivity. The United States contributed the largest number of publications (n=522) and had the highest total link strength (TLS=325), indicating both high research activity and extensive international collaboration. China, Italy, and England also contributed substantial publication volumes, while institutional analyses identified Harvard Medical School and Harvard University as separate bibliometric entities rather than a single combined institution.

Several European countries, including Germany, France, and Belgium, showed relatively high unadjusted average citation counts. However, these differences should not be interpreted as evidence of superior national research quality. Citation accumulation is influenced by publication age, disciplinary composition, document type, journal visibility, collaboration patterns, and a small number of highly cited publications. After publication-year adjustment, mean citations per year remained heterogeneous, but wide confidence intervals and smaller differences in median CPY indicated substantial skewness. Accordingly, publication volume, collaborative connectivity, raw citation impact, and time-normalized citation performance should be regarded as complementary rather than interchangeable dimensions of research activity.

China ranked second in publication output (n=173) but showed lower unadjusted average citations per publication (17.31) and a lower total link strength (51) than several other highly productive countries. These observations identify differences in citation visibility and international connectivity rather than a simple quality hierarchy. For emerging strict BCI and adaptive neuromodulation research, broader multicenter collaboration may be particularly valuable for external validation of neural biomarkers, replication of decoding algorithms, recruitment of adequately powered cohorts, and assessment across heterogeneous patient populations.

4.5 Clinical Translation and Future Directions

Clinical translation of adaptive neuromodulation and pain-related BCI depends on substantially more than technological feasibility. The evidence base remains heterogeneous across technological categories: established neuromodulation and selected closed-loop SCS systems have reached randomized clinical evaluation, whereas strict pain-related BCI is more commonly represented by feasibility, neurofeedback, protocol, and proof-of-concept studies. Major translational challenges include long-term signal stability, appropriate clinical trial design, device safety, regulatory validation, health-economic implementation, usability, neural-data governance, privacy, cybersecurity, and equitable access. The US Food and Drug Administration has issued guidance on nonclinical testing and clinical-study considerations for implanted BCI devices in paralysis or amputation[52]. Although this guidance is not pain-specific, it illustrates the regulatory complexity of systems that combine chronic neural recording, software, and device-mediated output.

One important research direction is the development of brain-derived sensing and decoding architectures. First-in-human work using bidirectional intracranial recording systems has demonstrated that individualized chronic pain states can be decoded from activity in regions including the orbitofrontal and anterior cingulate cortices[49]. These findings support the feasibility of brain-derived biomarkers as potential inputs for future adaptive systems, but successful decoding should be distinguished from demonstration of therapeutic benefit. At the hardware level, advances in implantable recording systems and flexible, high-density microelectrode arrays may improve signal acquisition and support more stable long-term neural interfaces[26]. Translation will ultimately require prospective testing that links engineering performance to clinically meaningful patient outcomes. Broader literature on DBS and non-invasive brain stimulation for chronic pain further illustrates the heterogeneous translational maturity of brain-directed neuromodulation approaches[53,54].

Machine learning (ML) and artificial intelligence (AI) are also increasingly applied to EEG, intracranial electrophysiology, functional imaging, and multimodal physiological data for pain-related classification and prediction[55-59]. These approaches may eventually provide useful control signals for adaptive interventions; however, high offline classification accuracy alone does not establish clinical validity. Small sample sizes, high-dimensional feature spaces, inter-individual variability, repeated observations from the same participants, and changes in neural signals over time can all inflate apparent performance or limit generalizability. Future studies should therefore emphasize participant-level train/test separation when cross-patient inference is intended, transparent preprocessing and feature selection, independent external validation, prespecified evaluation metrics, and prospective real-time testing. The distinction between personalized within-subject decoding and population-level cross-subject prediction should also be explicit, because these approaches serve different clinical purposes and require different standards of generalizability. A systematic review of EEG-based machine-learning studies for pain similarly identified heterogeneous reporting, potential bias, and insufficient external validation as barriers to clinical interpretation[60]. Related BCI and machine-learning reviews further emphasize the need to distinguish exploratory algorithmic performance from validated clinical utility[57].

Meanwhile, the capability to infer or classify pain-related states from neural data raises important ethical challenges regarding neurorights and mental privacy[61,62]. Neural biomarkers should complement rather than replace patient self-report, because pain remains intrinsically subjective and multidimensional. A failure of an algorithm to identify an expected neural signature should not be interpreted as evidence that a patient’s reported pain is absent or invalid. Future regulatory and governance frameworks should address data ownership, informed consent, cybersecurity, secondary use of neural data, algorithmic bias, and the ethical implications of continuous neural monitoring[61-63]

4.6 Limitations of the Study

While this analysis provides a comprehensive structural overview, several limitations must be acknowledged. First, the reliance solely on the WoSCC database, although it provides standardized bibliographic and citation metadata, necessarily excludes publications indexed only in other major databases, such as Scopus, PubMed, or specialized engineering repositories (e.g., IEEE Xplore), which may be highly relevant to BCI technology. Second, our search strategy, though designed to be comprehensive by including both BCI and feedback-based neuromodulation terms, may not have captured all nascent or highly specialized subdomains due to the rapid evolution of terminology. Third, the inclusion of only English-language publications might introduce a bias, undervaluing significant contributions from non-English-speaking research communities. Fourth, bibliometric metrics are time-sensitive; the citation counts and impact of the most recent publications (2023–2025) are likely underrepresented, reflecting a natural citation accrual lag rather than their true scholarly value.

Fifth, the primary search was deliberately broad and generated a conceptually heterogeneous 1,269-record parent corpus. The focused sensitivity analysis partially addressed this issue by identifying 308 candidate publications and distinguishing strict BCI/BMI, closed-loop or adaptive neuromodulation, and BCI-enabling pain decoding. However, this secondary classification was post hoc and was intended to improve technological interpretation rather than to replace a formal systematic review of individual interventions.

Sixth, the focused classification relied primarily on titles, abstracts, keywords, and bibliographic information, with full-text or publicly accessible source verification used when the technological architecture was unclear. Classification therefore depended partly on the level of technical detail reported in each publication, and uncommon terminology may still have resulted in missed or misclassified records. The categories represent publication-level technological focus rather than formal assessments of study quality, clinical maturity, or therapeutic efficacy.

Seventh, the parent retrieval strategy covered pain-related research broadly rather than applying a chronic-pain-only eligibility criterion. Although chronic and neuropathic pain were prominent within the corpus, some records addressed experimental pain, procedural pain, headache, or other pain-related states. The findings should therefore be interpreted as characterizing a broader pain-related neuromodulation and neural-interface landscape.

Finally, bibliometric prominence should not be equated with clinical evidence. Publication growth, keyword frequency, network centrality, and citation bursts identify scholarly attention and intellectual connectivity but do not establish therapeutic effectiveness, comparative benefit, safety, or readiness for routine clinical adoption. This study did not perform a formal risk-of-bias assessment, meta-analysis, or certainty-of-evidence evaluation.

5 Conclusion

This bibliometric analysis characterizes the global research landscape linking pain, neuromodulation, neural interfaces, adaptive closed-loop systems, and emerging brain–computer interface technologies from 2000 to 2025.

The findings do not indicate a completed transition from conventional neuromodulation to BCI-based therapy. Rather, they reveal a broad technological continuum in which established approaches such as SCS, DBS, rTMS, and tDCS continue to form the major scientific and clinical foundation of the field, while physiological sensing, adaptive feedback control, neural pain decoding, and brain-derived closed-loop systems represent more recent and distinct areas of development. The focused sensitivity analysis reinforces this interpretation: only 29 of the 1,269 publications were classified as strict BCI/BMI, 42 as closed-loop or adaptive neuromodulation, and 9 as BCI-enabling pain decoding. Together, these 80 publications represented 6.3% of the complete corpus.

Increasing publication activity and bibliometric prominence therefore reflect growing scientific attention, interdisciplinary engagement, and technological exploration rather than verified clinical efficacy or therapeutic superiority. Future translation will depend on prospective and external validation of neural biomarkers and decoding algorithms, appropriately designed multicenter studies, multidimensional patient-centered outcomes, transparent analytical pipelines, and early consideration of regulatory, health-economic, privacy, cybersecurity, and ethical issues. The eventual role of strict BCI and adaptive closed-loop systems in pain management will be determined by reproducible evidence and meaningful patient benefit rather than by publication growth alone.

References

[1]

Raja SN, Carr DB, Cohen M, et al. The revised International Association for the Study of Pain definition of pain: concepts, challenges, and compromises. Pain. 2020;161(9):1976–1982.

[2]

Cohen SP, Vase L, Hooten WM. Chronic pain: an update on burden, best practices, and new advances. Lancet. 2021;397(10289):2082–2097.

[3]

Ferreira ML, de Luca K, Haile LM, et al. Global, regional, and national burden of low back pain, 1990–2020, its attributable risk factors, and projections to 2050: a systematic analysis of the Global Burden of Disease study 2021. Lancet Rheumatol. 2023;5(6):e316–e329.

[4]

Sheng J, Liu S, Wang Y, Cui R, Zhang X. The link between depression and chronic pain: neural mechanisms in the brain. Neural Plast. 2017;2017:9724371.

[5]

Volkow ND, McLellan AT. Opioid abuse in chronic pain — misconceptions and mitigation strategies. N Engl J Med. 2016;374(13):1253–1263.

[6]

Malafoglia V, Ilari S, Vitiello L, et al. The interplay between chronic pain, opioids, and the immune system. Neuroscientist. 2022;28(6):613–627.

[7]

Chou R, Turner JA, Devine EB, et al. The effectiveness and risks of long-term opioid therapy for chronic pain: a systematic review for a National Institutes of Health pathways to prevention workshop. Ann Intern Med. 2015;162(4):276–286.

[8]

Chaudhary U, Birbaumer N, Ramos-Murguialday A. Brain–computer interfaces for communication and rehabilitation. Nat Rev Neurol. 2016;12(9):513–525.

[9]

Ramadan RA, Altamimi AB. Unraveling the potential of brain-computer interface technology in medical diagnostics and rehabilitation: a comprehensive literature review. Health Technol (Berl). 2024;14(2):263–276.

[10]

Patrick-Krueger KM, Burkhart I, Contreras-Vidal JL. The state of clinical trials of implantable brain–computer interfaces. Nat Rev Bioeng. 2025;3(1):50–67.

[11]

Mekhail N, Levy RM, Deer TR, et al. Long-term safety and efficacy of closed-loop spinal cord stimulation to treat chronic back and leg pain (evoke): a double-blind, randomised, controlled trial. Lancet Neurol. 2020;19(2):123–134.

[12]

Lefaucheur JP, André-Obadia N, Antal A, et al. Evidence-based guidelines on the therapeutic use of repetitive transcranial magnetic stimulation (rTMS). Clin Neurophysiol. 2014;125(11):2150–2206.

[13]

Moxon KA, Foffani G. Brain-machine interfaces beyond neuroprosthetics. Neuron. 2015;86(1):55–67.

[14]

Shirvalkar P, Veuthey TL, Dawes HE, Chang EF. Closed-loop deep brain stimulation for refractory chronic pain. Front Comput Neurosci. 2018;12:18.

[15]

Schultz DM, Webster L, Kosek P, Dar U, Tan Y, Sun M. Sensor-driven position-adaptive spinal cord stimulation for chronic pain. Pain Physician. 2012;15(1):1–12.

[16]

Patil AS, Levasseur B, Gupta M. Neuromodulation and habituation: a literature review and conceptional analysis of sustaining therapeutic efficacy and mitigating habituation. Biomedicines. 2024;12(5):930.

[17]

Mangano N, Torpey A, Devitt C, Wen GA, Doh C, Gupta A. Closed-loop spinal cord stimulation in chronic pain management: mechanisms, clinical evidence, and emerging perspectives. Biomedicines. 2025;13(5):1091.

[18]

Diotaiuti P, Corrado S, Tosti B, et al. Evaluating the effectiveness of neurofeedback in chronic pain management: a narrative review. Front Psychol. 2024;15:1369487.

[19]

Chmiel J, Kopańska M, Leszek J, Trojniak J, Ros T. EEG neurofeedback for the treatment of neuropathic pain in the elderly—a mechanistic review. Geroscience. 2026;48(3):4893–4922.

[20]

Knotkova H, Hamani C, Sivanesan E, et al. Neuromodulation for chronic pain. Lancet. 2021;397(10289):2111–2124.

[21]

Wang J, Chen ZS. Closed-loop neural interfaces for pain: where do we stand? Cell Rep Med. 2024;5(10):101662.

[22]

Russo M, Cousins MJ, Brooker C, et al. Effective relief of pain and associated symptoms with closed-loop spinal cord stimulation system: preliminary results of the Avalon study. Neuromodulation. 2018;21(1):38–47.

[23]

Liu X, Sanchez SW, Gong Y, et al. An insect-based bioelectronic sensing system combining flexible dual-sided microelectrode array and insect olfactory circuitry for human lung cancer detection. Biosens Bioelectron. 2025;281:117356.

[24]

Wu Q, Gong Y, Liu X. Bridging the gap: integrated high-density microelectrode arrays for cellular, organoid, and clinical electrophysiology. Micromachines. 2026;17(5):611.

[25]

Liu X. Development and application of flexible dual-sided microelectrode arrays (MEAs) for advanced bioelectronic sensing and neural interfacing [dissertation]. 2025.

[26]

Liu X, Gong Y, Jiang Z, Stevens T, Li W. Flexible high-density microelectrode arrays for closed-loop brain–machine interfaces: a review. Front Neurosci. 2024;18:1348434.

[27]

Araki T, Bongartz LM, Kaiju T, et al. Flexible neural interfaces for brain implants—the pursuit of thinness and high density. Flex Print Electron. 2020;5(4):043002.

[28]

McFarland DJ, Wolpaw JR. Brain-computer interfaces for communication and control. Commun ACM. 2011;54(5):60–66.

[29]

Lee IS, Lee H, Chen YH, Chae Y. Bibliometric analysis of research assessing the use of acupuncture for pain treatment over the past 20 years. J Pain Res. 2020;13:367–376.

[30]

Park J, Kim H, Kim KW, Cho JH, Chung WS, Song MY. Bibliometric analysis of research trends on acupuncture for neck pain treatment over the past 20 years. J Pain Res. 2021;14:3207–3221.

[31]

Zhang L, Zhang H, Zhou X, Zhao J, Wang X. Bibliometric analysis of research on migraine-stroke association from 2013 to 2023. J Pain Res. 2023;16:4089–4112.

[32]

Chio N, Quiles-Cucarella E. A bibliometric review of brain-computer interfaces in motor imagery and steady-state visually evoked potentials for applications in rehabilitation and robotics. Sensors (Basel). 2025;25(1):154.

[33]

Liu MY, Fang MZ, Zhang BH, et al. Bibliometric analysis of brain-computer interface research in spinal cord injury: current landscape and future directions. Int J Surg. 2025;111(7):4804–4806.

[34]

Li F, Zhang D, Chen J, Tang K, Li X, Hou Z. Research hotspots and trends of brain-computer interface technology in stroke: a bibliometric study and visualization analysis. Front Neurosci. 2023;17:1243151.

[35]

Zhou XC, Huang YB, Liu Z, et al. Bibliometric analysis of functional magnetic resonance imaging studies on manual therapy analgesia from 2002–2022. J Pain Res. 2023;16:2115–2129.

[36]

Zhu H, Huang Y, Wang X, Xiang W, Xie Y. Trends and hotspots in research related to tumor immune escape: bibliometric analysis and future perspectives. Front Immunol. 2025;16:1604216.

[37]

Yang S, Zhao S, Ye Y, Jia L, Lou Y. Global research trends on the links between gut microbiota and cancer immunotherapy: a bibliometric analysis (2012-2021). Front Immunol. 2022;13:952546.

[38]

Zhang W, Moskowitz RW, Nuki G, et al. OARSI recommendations for the management of hip and knee osteoarthritis, part II: OARSI evidence-based, expert consensus guidelines. Osteoarthritis Cartilage. 2008;16(2):137–162.

[39]

Ossipov MH, Morimura K, Porreca F. Descending pain modulation and chronification of pain. Current Opinion in Supportive & Palliative Care. 2014;8(2):143–151.

[40]

Deer TR, Levy RM, Kramer J, et al. Dorsal root ganglion stimulation yielded higher treatment success rate for complex regional pain syndrome and causalgia at 3 and 12 months: a randomized comparative trial. Pain. 2017;158(4):669–681.

[41]

Osborn LE, Dragomir A, Betthauser JL, et al. Prosthesis with neuromorphic multilayered e-dermis perceives touch and pain. Sci Robot. 2018;3(19):eaat3818.

[42]

Lefaucheur JP, Aleman A, Baeken C, et al. Evidence-based guidelines on the therapeutic use of repetitive transcranial magnetic stimulation (rTMS): an update (2014–2018). Clin Neurophysiol. 2020;131(2):474–528.

[43]

Kapural L, Yu C, Doust MW, et al. Novel 10-kHz high-frequency therapy (HF10 therapy) is superior to traditional low-frequency spinal cord stimulation for the treatment of chronic back and leg pain: the SENZA-RCT randomized controlled trial. Anesthesiology. 2015;123(4):851–860.

[44]

Volkow ND, Blanco C. The changing opioid crisis: development, challenges and opportunities. Mol Psychiatry. 2021;26(1):218–233.

[45]

The Lancet Digital Health. Turning the tide on the opioid crisis. Lancet Digit Health. 2022;4(6):e398.

[46]

Shurman J, Koob GF, Gutstein HB. Opioids, pain, the brain, and hyperkatifeia: a framework for the rational use of opioids for pain. Pain Med. 2010;11(7):1092–1098.

[47]

Florence C, Luo F, Rice K. The economic burden of opioid use disorder and fatal opioid overdose in the United States, 2017. Drug Alcohol Depend. 2021;218:108350.

[48]

Compton WM, Valentino RJ, DuPont RL. Polysubstance use in the U.S. opioid crisis. Mol Psychiatry. 2021;26(1):41–50.

[49]

Shirvalkar P, Prosky J, Chin G, et al. First-in-human prediction of chronic pain state using intracranial neural biomarkers. Nat Neurosci. 2023;26(6):1090–1099.

[50]

Mekhail NA, Levy RM, Deer TR, et al. ECAP-controlled closed-loop versus open-loop SCS for the treatment of chronic pain: 36-month results of the EVOKE blinded randomized clinical trial. Reg Anesth Pain Med. 2024;49(5):346–354.

[51]

Hunter CW, Raskin JS, Mekhail NA, et al. ECAP-controlled closed-loop spinal cord stimulation for chronic nonsurgical refractory back pain: subgroup analysis from two prospective multicenter clinical trials. Spine. 2025;50(23):1637–1647.

[52]

US Food and Drug Administration. Implanted brain-computer interface (BCI) devices for patients with paralysis or amputation - non-clinical testing and clinical considerations: guidance for industry and Food and Drug Administration staff. Center for Devices and Radiological Health; 2021. Accessed July 25, 2025. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/implanted-brain-computer-interface-bci-devices-patients-paralysis-or-amputation-non-clinical-testing

[53]

Shaheen N, Shaheen A, Elgendy A, et al. Deep brain stimulation for chronic pain: a systematic review and meta-analysis. Front Hum Neurosci. 2023;17:1297894.

[54]

Sveva V, Cruciani A, Mancuso M, et al. Cerebellar non-invasive brain stimulation: a frontier in chronic pain therapy. J Pers Med. 2024;14(7):675.

[55]

Al-Nafjan A, Alshehri H, Aldayel M. Objective pain assessment using deep learning through EEG-based Brain–Computer interfaces. Biology. 2025;14(2):210.

[56]

Wager TD, Atlas LY, Lindquist MA, Roy M, Woo CW, Kross E. An fMRI-based neurologic signature of physical pain. N Engl J Med. 2013;368(15):1388–1397.

[57]

Williams C, Anik FI, Hasan MM, et al. Advancing brain-computer interface closed-loop systems for neurorehabilitation: systematic review of AI and machine learning innovations in biomedical engineering. JMIR Biomed Eng. 2025;10:e72218.

[58]

Ryabkin D, Pustovit G, Gerasimenko AY. Machine learning-based detection of pain from EEG signals. In: 2025 systems and technologies of the digital HealthCare (STDH). IEEE; 2025:74–77.

[59]

Alshehri H, Al-Nafjan A, Aldayel M. Decoding pain: a comprehensive review of computational intelligence methods in electroencephalography-based Brain–Computer interfaces. Diagnostics. 2025;15(3):300.

[60]

Mari T, Henderson J, Maden M, Nevitt S, Duarte R, Fallon N. Systematic review of the effectiveness of machine learning algorithms for classifying pain intensity, phenotype or treatment outcomes using electroencephalogram data. J Pain. 2022;23(3):349–369.

[61]

Ruiz S, Valera L, Ramos P, Sitaram R. Neurorights in the constitution: from neurotechnology to ethics and politics. Philos Trans R Soc Lond B Biol Sci. 2024;379(1915):20230098.

[62]

Brown CML. Neurorights, mental privacy, and mind reading. Neuroethics. 2024;17(2):34.

[63]

Di Salvo M. The protection of neural rights in the age of neurotechnologies and AI. The ethical challenge for law and neuroscience. Russ J Econ Law. 2025;19(1):202–233.

Rights & permissions

2097-7816/2026 The Third Affiliated Hospital of Sun Yat-Sen University and Higher Education Press Limited Company. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

PDF (10013KB)

0

Accesses

0

Citation

Detail

Sections
Recommended

/

〈 〉