The authors reviewed literature showing that scalp EEG can track neural oscillation changes during the acupuncture-induced sensation called “Deqi”. EEG’s millisecond temporal resolution allows investigators to observe dynamic, time‑locked activity as the peripheral stimulus unfolds and central networks respond. The article reported that these temporal features are central to efforts to move beyond purely subjective report and toward physiological markers that reflect the evolving neural state during needling.
EEG studies vary in design and outcome measures. Some cohorts examined spectral power and event‑related potentials tied to needle manipulation; others focused on connectivity or time–frequency dynamics. The review synthesized these approaches and framed EEG as the core neurophysiological modality currently best suited to capture the fast dynamics of the “Deqi” process.
Authors analyzed published work on how peripheral signals initiate the cascade that culminates in the reported “Deqi” experience and how the brain responds. Peripheral initiation mechanisms reported in the literature include afferent sensory inputs generated by needle insertion and manipulation, which then engage spinal and supraspinal pathways. The review described central response features as changes in cortical oscillations and network interactions that occur during the induction of “Deqi”.
The study team observed that the peripheral‑to‑central coupling remains incompletely characterized in existing studies. Available papers provide suggestive but heterogeneous evidence about which ascending pathways and cortical targets are most consistently involved. The authors emphasized the need for coordinated protocols to map the timing and directionality of peripheral‑central interactions more precisely.
The review summarized evidence that machine‑learning–based feature decoding applied to EEG signals can classify neural states corresponding to reported “Deqi”. Investigators reported that supervised algorithms extracting spectral, temporal, and connectivity features improve objective identification compared with reliance on subjective sensation alone. These decoding approaches translate continuous EEG dynamics into discrete markers that can, in principle, signal when a subject has entered a target neural state.
The authors noted, however, that decoding accuracy reported across studies remains variable. Heterogeneous feature sets, inconsistent labeling of the “Deqi” event, and differing preprocessing pipelines all contribute to this variability. The review stressed that current ML implementations demonstrate feasibility but not yet robust, generalizable performance suitable for wide clinical deployment.
A major focus of the article was the potential for brain‑computer interface systems to form closed‑loop feedback loops that both detect and actively regulate the “Deqi” state. The concept is straightforward: real‑time EEG decoding detects neural signatures related to “Deqi”, and the BCI delivers feedback or adjusts stimulation parameters to induce or sustain that state.
Authors described the closed‑loop strategy as an innovative path toward precision acupuncture—one where objective neural markers guide intervention timing and parameter adjustments rather than sole dependence on subjective patient report. The review summarized proposed architectures and the conceptual value of such systems but did not provide a single validated protocol; it compiled proposals and early experimental approaches from the literature.
The review cataloged several persistent challenges. Signal artifact interference—stemming from muscle activity, movement, and environmental noise—complicates EEG acquisition during needle manipulation. Authors noted that artifact contamination undermines both feature extraction and downstream decoding.
Another recurring limitation is insufficient decoding accuracy across heterogeneous datasets. The study team reported that standardization is lacking: research paradigms differ in how “Deqi” is defined and labeled, in EEG montage and preprocessing, and in machine‑learning methodology. These inconsistencies hinder cross‑study comparison and the pooling of data for robust algorithm training. The review also pointed out that, because the paper is a synthesis, it does not present new quantitative performance metrics or protocol parameters.
Looking ahead, the authors advocated targeted development in three areas. First, they recommend advancing dynamic network decoding technologies that can map time‑varying interactions rather than static features alone. Second, they urged deeper mechanistic study of peripheral–central coupling to clarify how afferent signals translate into specific cortical states labeled as “Deqi”. Third, they proposed promoting the field’s movement toward greater intelligence and standardization—consistent paradigms, shared preprocessing pipelines, and more rigorous reporting standards to support reproducible decoding.
The review concluded with an explicit statement of current constraints: artifact interference, uneven decoding accuracy, and inconsistent paradigms across studies. The authors did not offer a single validated closed‑loop protocol or quantitative claims of clinical efficacy; instead, they positioned EEG‑BCI as a technical platform that merits further methodological refinement and empirical validation before clinical translation.
Limitations and unanswered questions
Because this article is a review, it synthesizes existing studies rather than reporting new experimental data. The source did not provide specific closed‑loop control algorithms, parameter settings for feedback interventions, or quantitative accuracy statistics. Open questions remain about how best to label “Deqi” for machine learning, which peripheral biomarkers most reliably predict central responses, and how to mitigate real‑world EEG artifact in clinical needling settings. These knowledge gaps frame the next research priorities identified by the authors: dynamic network decoding, deeper peripheral‑central coupling exploration, and standardized, intelligent system development.