Evaluation of Acupuncture Efficacy in Modulating Brain Activity With Periodic-Aperiodic EEG Measurements

Acupuncture is an important therapeutic method of traditional Chinese medicine and can effectively modulate brain disorders. The therapeutic efficacy of acupuncture is hard to evaluate due to lacking of effective measurements of brain activity. In this work, we design an EEG-based monitoring system to evaluate therapeutic effect of acupuncture on human brain by extracting periodic-aperiodic features. Power spectral density is estimated to compute the adjusted power of periodic oscillatory rhythm in EEG under acupuncture stimulation. It is exhibited that the brain activity in alpha band (8–12 Hz) is significantly enhanced during acupuncture, especially in parietal and occipital lobe regions. To probe the modulatory effect of acupuncture on aperiodic brain activity, we calculate the aperiodic exponent based on the parameterization of EEG power spectra. The aperiodic exponent decreases along with acupuncture process, which is more significant in central and frontal lobe regions. Furthermore, sensitivity of different brain regions to acupuncture is assessed by the integration of adjusted power and aperiodic exponent. Experimental results demonstrate the effectiveness of proposed periodic-aperiodic measurements of EEG signals, by which different effects of four acupuncture manipulations are precisely evaluated and a knowledge graph is established. The monitoring system provides a new perspective to quantitatively evaluate acupuncture effect on human brain and improve its therapeutic efficacy in clinical applications for neural disorders.


Evaluation of Acupuncture Efficacy in Modulating Brain Activity With Periodic-Aperiodic EEG Measurements I. INTRODUCTION
A CUPUNCTURE is an essential therapeutic method in traditional Chinese medicine, which can regulate immune and nervous systems of human [1], [2], [3].It is reported that acupuncture could excite peripheral and central nervous systems by inserting the needle into acupoints to achieve neural modulation combined with various acupuncture manipulations.Clinical studies have demonstrated the remarkable efficacy of acupuncture in modulating brain activity and treating neurological disorders such as Alzheimer's disease (AD) [4], Parkinson's disease [5], and epilepsy [6].Recent research explores the changes in brain activity during and after acupuncture and finds that the effect of acupuncture on brain function persists for several minutes after treatment [6], [7], [8], [9].It is shown that acupuncture could enhance the network connectivity, strengthen the brain's small-world properties, and improve information transmission efficiency in the brain [10], [11], [12], [13], [14], [15].Moreover, different acupuncture manipulations can generate distinct therapeutic effect [16], [17], [18].However, the effectiveness of acupuncture treatment is usually evaluated based on the patient's cognitive and behavior measurements in clinical with low consistency and reliability.Due to the lacking of effective biomarkers, an online monitoring system that can quantitively evaluate the effectiveness of acupuncture is still unestablished now.
In recent years, Brain-Computer Interface (BCI) systems have been widely applied in neuroscience research to record and measure brain activity with various methods, including functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG), and electroencephalography (EEG) [19], [20], [21], [22].It is well-established that EEG is a non-invasive electrophysiological recording technique with the capability to capture real-time neural activity of the cerebral cortex with a millisecond time resolution [23], [24], [25].Recent studies have demonstrated accurate decoding of finger movements and prediction of motor imagery tasks through EEG-based BCI systems [26], [27].It has been shown that oscillation characteristics in EEG signals are relevant to neurophysiological mechanisms underlying cognitive processes [28].Particularly, the frequency of alpha oscillations has been shown to reflect state-related changes in cortical excitability of the brain.Samuel et al. [29] finds the increased frequency of posterior alpha oscillations during the retention period of a memory task, which is accompanied by slower reaction time and decreased neural responses to external stimulation.Mullinger et al. [30] observes a significantly greater increase in EEG alpha power after flickering than after static visual stimulus.Hence, EEG provides a feasible way for evaluating acupuncture effect by the measurement of oscillatory characteristics of brain activity.
Extensive research demonstrates that acupuncture-evoked EEG provide an insight into interpreting the modulatory effect of acupuncture on brain activity [31].Recently, many time-frequency analysis methods have been applied into the measurement of brain response to acupuncture stimulation with EEG signals.As one of mostly used frequency domain analysis methods, power spectral density (PSD) analysis estimates the power spectrum of EEG by converting the amplitude of EEG over time into a spectrum of EEG power over frequency, which allows for the direct observation of the distribution and changes of various rhythmic waves in EEG.Muchtadi et al. [32] finds that acupuncture can enhance PSD in the alpha and theta bands of the EEG, indicating a relaxing effect on the human mind.Using nonlinear and multivariate statistics of EEG signals, Qi et al. [33] finds that acupuncture can activate a wide range of brain regions.Yu et al. [14] records the EEG signal in three acupuncture stages: pre, during, and after stimulation, and investigates its effect on temporal and spatial activity of the brain.It is found that acupuncture at "Zusanli" (ST-36) could significantly improve the oscillation power in delta (1-4 Hz) and alpha (8-12Hz) bands.After acupuncture, the power in the delta band drops back to the value in pre-acupuncture stage, but it remains high in the alpha band, which is known as the "post-effect" of acupuncture.However, previous studies on acupuncture EEG only consider the periodic components of the brain electrical signals.Simply regarding the aperiodic activity as static noise and ignoring it may lead to incomplete evaluation of acupuncture effect.
The PSD of EEG recordings is an intricate combination of periodic oscillatory rhythms and a non-oscillatory activity, which is also termed aperiodic component that typically shown as a 1/f-like curve of power and frequency [34].Specifically, the 1/f-like characteristic has been defined with two parameters: aperiodic offset represents the power at the lowest frequency of the aperiodic curve, and aperiodic exponent defined as the overall decreasing slope of power across frequencies [35].Recent studies have shown that the aperiodic exponent may correspond to physiologically relevant signals and reflect meaningful characteristics of brain function [36].Several experimental findings confirm that aperiodic exponents can dynamically change according to task demands, cognitive states, age, and pathological conditions [37].Voytek et al. [38] finds that the aperiodic exponent exhibits a reduction with increasing age in adults, which providing an electrophysiological support for the correlate of cognitive deficits with aging.Ostlund et al. [39] studies the power spectrum of EEG in a sample of adolescents with and without attention deficit/hyperactivity disorder (ADHD), and finds that the aperiodic exponent is related to the cognition and behavior.Molina et al. [40] finds that the anomalous aperiodic exponent of treated schizophrenia patients can be transiently normalized with memantine, indicating the importance of aperiodic activity for the function and dysfunction of the brain.
Experimental and theoretical studies provide vast evidences showing that the balance between neural excitation and inhibition (E/I balance) is associated with aperiodic exponents, which reflect the neural E/I ratio by the steepness.Colombo et al. [41] finds that the 1/f-like curve is steeper during both sleep and anesthesia compared to wakefulness, indicating an increase in the aperiodic exponent as the E/I balance shifts towards inhibition.By modeling a simulated network including both excitatory and inhibitory populations, Gao et al. [34] finds the crucial role of blocking activity propagation in accelerating the decay of PSD, which suggests that increasing aperiodic exponent fundamentally decreases the efficiency of communications between brain regions.Furthermore, given that periodic components are embedded within aperiodic components, taking aperiodic activity into account can avoid misinterpretations of band-limited power differences.Therefore, extracting both periodic and aperiodic features of EEG signals can represent the brain response to acupuncture stimulation more accurately, and provide a comprehensive measurement for the evaluation of acupuncture effect.
In this work, we design a flexible monitoring system that can evaluate the effect of acupuncture with EEG signals online.The proposed framework records human EEG signals during acupuncture stimulation and extract periodic-aperiodic features to quantitively evaluate the effectiveness of acupuncture.By parameterization of power spectra, the measures of adjusted power and aperiodic exponent are computed to assess the modulatory effect of acupuncture on periodic and aperiodic brain activity, respectively.Furthermore, sensitivity of individual brain regions to acupuncture stimulation is calculated by the integration of adjusted power and aperiodic exponent.Aiming to assess the performance of the proposed system, we design an acupuncture experiment at "Zusanli" acupoint with different acupuncture manipulations.Experimental results demonstrate the accuracy and reliability of the proposed EEG-based monitoring system, with which distinct effects of acupuncture manipulations are accurately evaluated and a knowledge graph for acupuncture is established.

II. MONITORING SYSTEM AND EVALUATION METHODS
The architecture of proposed monitoring and evaluation system for acupuncture efficacy is presented in Fig. 1.The real-time monitoring system records EEG signals of the subject under acupuncture and its modulatory effect on brain activity is then evaluated by EEG processor.The pipeline of EEG processing mainly includes raw data preprocessing, features extraction, measurement and evaluation (Fig. 2).Acupuncture-evoked EEG signals are recorded and digitized by means of a 64 channel EEG amplifier.Then preprocessing procedures are adopted to decrease noises from environment and equipment and to remove artifacts of eye movements.In order to extract features of brain activity, time-frequency analysis and PSD methods are implied to each channel of EEG signals.Furthermore, three evaluation measurements of EEG signals are obtained by parameterizing PSD into periodic and aperiodic components.Each part of this EEG-based online monitoring system for acupuncture efficacy is detailed as follows.

A. EEG Recordings
EEG signals are continuously recorded by an EEG amplifier, which is an international 10-20 standard electrodes system, including 60 electrodes evenly distributed along the scalp and 4 electrooculography (EOG) electrodes.The sampling rate of 64-channel EEG signals is 1000 Hz.

B. Data Preprocessing
To enhance the signal quality and remove artifacts, the EEG signals are preprocessed to remove high-frequency noise, mains electrical interference, and baseline drifting.The recorded EEG signals are filtered by a bandpass finite impulse digital filter with a range of 0.5 Hz to 30 Hz to eliminate mains electrical interference and high-frequency noise.Channels that have missing or abnormal EEG signals are marked as artifactual and interpolated to minimize the influence of faulty channels on subsequent analysis.Then the baseline correction is applied to remove the temporal drifts across electrodes.We applied the average re-referencing method to increase the signal-to-noise ratio and enhance the quality of the EEG data.Specifically, we subtract the averaged signal of all electrodes from the signal of each electrode.Independent components analysis (ICA) is performed to identify EOG artifacts for correction.By default, only channels whose correlation with EOG electrodes is higher than 50% are subtracted from the data.Additionally, we employed the ICA method from EEGLAB to detect and remove components associated with muscle and heart artifacts.The quality of EEG data is improved through separating and excluding these artifacts.Previous EEG studies has indicated that the configuration of the EEG recording system and the data preprocessing methods used in this study are suitable for EEG data analysis [42], [43].

C. Features Extraction
1) Time-Frequency Analysis: In order to extract the dynamic features of brain activity, the continuous wavelet transform (CWT) method is employed to perform time-frequency analysis of acupuncture EEG signals.The CWT of the square integrable function x(t) concerning a given wavelet function can be expressed as follows: where the ϕ( t−b a ) indicates the dilated and translated version of ϕ(t) with a being the scale parameter and b being the transition parameter.Here we choose the Morlet wavelet as ϕ(t), which is defined as By applying the CWT method to EEG signals, we obtain the wavelet coefficients that have both time and frequency domain Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
characteristics and can be used as one of the main features of brain activity.
2) Power Spectral Density: The PSD is estimated for each electrode of EEG using the Classical Welch method, which can minimize the interference of high amplitude transients and artifacts compared with other methods, and thus improve the resolution and variance of classical spectrum estimation.The calculation process is as follows.
Assuming that the EEG signal with a length of N is marked as x(n), n = 0, 1, . . ., N − 1, and it is divided into L nonoverlapping segments.The length of each segment is M, and the data in the segment i can be expressed as Then calculate the periodogram of each segment respectively.The periodogram is defined as follows: where w(n) is a window function, which can help reduce the variance of the periodogram.In this work, a hamming window of 1 s and 50% overlap is adopted as the window function, which can be expressed as and U is normalization factor that can be calculated as Assuming that the periodogram of each segment is uncorrelated to each other, the definition of PSD is as follows:

D. Measurement and Evaluation
Periodic and aperiodic features of PSD are characterized using the fitting-oscillations-and-one-over-f (FOOOF) toolbox [35].Based on the parameterization method, the PSD of all EEG electrodes can be conceptualized as a combination of periodic and aperiodic components (Fig. 3).The aperiodic component describes the 1/f-like characteristic of power spectra and is fit as a function within the frequency range of the spectrum, while the periodic components are characterized as frequency regions of power exceed the aperiodic component.Each of these periodic components is modeled with a Gaussian.PSD are parameterized for the frequency range of 0.5-30Hz, with 0.5 Hz frequency resolution.The PSD is modeled as follows: where T is the aperiodic component, G is the Gaussian that represent a periodic component, and N refers to the number of Gaussian distributions.Each G n can be defined by where p represents the adjusted power of the putative periodic component, in log 10 (power) values; c is the central frequency and w is the width of the Gaussian, equivalent to the bandwidth of the oscillation rhythm, both of them are measured in Hz; and F refers to the vector of input frequencies (Hz).
A Lorentzian function is adopted to model the aperiodic component T , which can be written as where, d is the aperiodic offset, χ is the aperiodic exponent and the parameter k controls the bending of the aperiodic component, with F as the array of frequencies of the power spectrum.
In this work, the aperiodic exponent value and the adjusted power value are extracted from the brain signal for each EEG electrode.These values are then combined to calculate a composite feature measurement called "sensitivity index", which can be represented as where, S is the sensitivity index of brain regions; p is the variation of adjusted power in each electrode between two acupuncture stages; and χ represents the variation of aperiodic exponent in each electrode between different stages of acupuncture.To sum up, three measurements calculated from EEG signals during acupuncture stimulation, including adjusted power, aperiodic exponent, and sensitivity index, are adopted to evaluate the effect of acupuncture.

A. Experiment Setup
To assess the effectiveness of proposed monitoring system and periodic-aperiodic measurements, we design the acupuncture experiment and record EEG signals of human by applying different acupuncture manipulations, including lifting-thrusting (LT) and twirling-rotating (TR), at "Zusanli" acupoint with different frequencies.In LT manipulation, the needle is vertically lifted and thrusted at an appropriate depth, and in TR manipulation the needle is horizontally rotated at an appropriate force.The frequencies for each acupuncture manipulation are 60 times/min or 90 times/min, which is called low frequency manipulation (Low) and high frequency manipulation (High), respectively.
The "Zusanli" acupoint is located in the lower limb, one finger's breadth from the anterior crest of the tibia, 3 centimeters below the knee joint [44].The application of acupuncture at the "Zusanli" acupoint has emerged as a promising therapeutic approach in the rehabilitation of neurological disorders.Research has substantiated its utility in facilitating neural function recovery during spinal tethering procedures, enhancing motor and sensory rehabilitation in stroke survivors, and supporting the management of diabetic peripheral neuropathy [45].Additionally, clinical studies have shown that acupuncture at "Zusanli" is effective for brain diseases, such as sleep disorders, cerebral infarction and Alzheimer's disease [46], [47], [48].Resent study reveals that stimulating "Zusanli" activates the vagus nerve-adrenal anti-inflammatory pathway, mediated by PROKR2-Cre marked sensory neurons in the dorsal root ganglia (DRG), essential for the activation of this pathway by low-intensity acupuncture stimulation [49].These findings provide a scientific basis for the application of "Zusanli" acupuncture in neurological rehabilitation and uncover the neuroanatomical mechanisms behind acupuncture treatment.To comprehensively elucidate the mechanisms and full scope of acupuncture's benefits in neurological rehabilitation, additional investigation is warranted.
A total of twenty subjects aged 21-26 (mean age 23), who have no experience of acupuncture treatment and no significant difference in physical condition, are recruited for the acupuncture experiment.All recruited subjects meet the criteria for acupuncture and agree to participate in the study, which is conducted with the permission of the ethics committee of the First Teaching Hospital of Tianjin University of Traditional Chinese Medicine.Acupuncture is carried out as following: firstly, the subject maintains resting and awake for 10 minutes, which is the pre-acupuncture (Pre-Acup) stage.Then a specialized acupuncturist inserts the needle into the acupoint till "qi" is felt to achieve and the subject has some needling sensations.There is a two-minute interval before the application of acupuncture treatments to eliminate the influence of needle insertion to the body on the EEG recording.Following that, Low LT, High LT, Low TR and High TR manipulations are performed in turn for 20 seconds, which is the during-acupuncture (Dur-Acup) stage.There is an interval of ten minutes between different manipulations, called post-acupuncture (Post-Acup) stage.Finally, pulled out the needle and the experiment finished.In order to further remove the influence on signal caused by insertion and withdrawal of needle, middle epoch of 20-second EEG data in Pre-Acup and Post-Acup stages is selected for each subject.The recorded EEG is preprocessed and analyzed by the monitoring and evaluation system proposed in this work.Three measurements, adjusted power, aperiodic exponent and sensitivity index, are calculated to evaluate the modulatory effect of different acupuncture manipulations on brain activity.In addition, we investigate the spatiotemporal characteristics of dynamical response of different brain regions to acupuncture stimulation.Based on above results, a knowledge graph is established to help evaluate the acupuncture effect generated by different manipulations in each brain region.

B. Enhanced Periodic Oscillatory Power of EEG With Acupuncture
A typical electrophysiological feature associated with acupuncture treatment is fluctuations in specific frequency bands of EEG signals, which is a phenomenon that can directly reflect the modulatory effect of acupuncture on brain activity.The effectiveness of four acupuncture manipulations on the spectral power is assessed by means of wavelet time frequency distribution diagram of EEG signals.The acupuncture process of each manipulation has been divided into three stages (Pre-Acup, Dur-Acup and Post-Acup), on the basis of which we draw the time-frequency diagram within 0.5-30Hz band for each acupuncture manipulation, as shown in Fig. 4. It is found that the power of the alpha band (8-12Hz) is significantly enhanced in the Dur-Acup stage with respect to the Pre-Acup stage, and then decreases after acupuncture stimulation.It is noteworthy that the spectral power of alpha band in the postacupuncture, as compared to the Pre-Acup stage, still retains enhancement effect.
We further calculate PSD to quantify the modulatory effect of four acupuncture manipulations on the periodic oscillatory activity of human brain.The obtained results are exhibited in Fig. 5.It can be observed that there are power peaks in alpha band for each stage of Pre-Acup, Dur-Acup and Post-Acup.Importantly, for each acupuncture manipulation, the peak value Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.  of Dur-Acup stage is much higher than that of Pre-Acup and Post-Acup stages.To remove the influence of aperiodic component on the power spectrum analysis, we calculate the adjusted power based on power spectrum parametrization method.Furthermore, we compare the adjusted power in alpha band at different stages of acupuncture with four manipulations (Fig. 6).The one-way analysis of variance (ANOVA) is applied to determine whether there are any statistically significant differences among three acupuncture stages.
With adjusted power spectrum, the efficacy of acupuncture on modulating periodic oscillatory activity is clearly manifested.Different level of enhanced adjusted power in To investigate spatiotemporal characteristics of oscillatory response of the brain to external acupuncture stimulation, we calculate the adjusted power in alpha band for each EEG electrode (Fig. 7).It can be observed that each acupuncture manipulation can activate the whole brain.However, the response of different brain regions shows a significant heterogeneity.The highest variation of adjusted alpha power during and after acupuncture is found in parietal and occipital lobe regions (Pz, P2, P4, P5, P6, P7, P8, POz, PO3, PO4, PO7, PO8, Oz and O2).There are several brain regions recorded by p1 and p3 electrodes, whose oscillatory activity is more sensitive to high-frequency LT and TR manipulations, indicating a widespread acupuncture effect on the brain.In addition, the most significant enhancement of adjusted alpha power after acupuncture is evoked by the high-frequency TR manipulation.
In conclusion, acupuncture has a strong modulatory effect on periodic oscillatory activities of human brain.By parameterizing the PSD of EEG signals, it is determined that acupuncture mainly adjust the power in alpha band, which is relevant to cognitive functions [48], [49], [50].Furthermore, acupuncture manipulations with high-frequency exhibit stronger modulatory effect.Compared to LT manipulations, TR manipulations demonstrate higher efficacy in modulating periodic activities of the whole brain, showing superior immediate and sustained acupuncture effect.Specifically, the parietal and occipital lobe regions are more sensitive to acupuncture stimulations.

C. Decreased Aperiodic Exponent of EEG Signals With Acupuncture
To assess the variation of aperiodic activity of the brain with different acupuncture manipulations, we calculate the aperiodic measurement in three acupuncture stages.The statistical results are shown in Fig. 8.We can observe the reduction of aperiodic exponents induced by acupuncture stimulations.For Low LT, median: 1.27 µV 2 Hz −1 in Pre-Acup vs. 1.12 µV 2 Hz −1 in Dur-Acup, p = 2.8 × 10 −3 ; for High LT, median: 1.18 µV 2 Hz −1 in Pre-Acup vs. 0.92 µV 2 Hz −1 in Dur-Acup, p = 3.1 × 10 −4 ; for Low TR, median: 1.30 µV 2 Hz −1 in Pre-Acup vs. 1.14 µV 2 Hz −1 in Dur-Acup, The aperiodic activity is slightly recovered after acupuncture, but is still weaker than that before stimulation, which is manifested by significant differences of aperiodic exponent between Pre-Acup stages and Post-Acup stages.For Low LT, median: 1.In addition, the recovery level of aperiodic exponents in high frequency acupuncture manipulations are much lower compared to that in low frequency acupuncture manipulations., which also demonstrates the high modulatory efficacy of high frequency stimulation.
The obtained results offer new insights into dynamic fluctuations of aperiodic exponent of EEG signals.Spatiotemporal analysis has revealed that aperiodic activity is distributed throughout the brain and concentrated in the frontal and central lobe regions in resting state.Acupuncture can transform the aperiodic activity in the parietal and occipital lobe regions into periodic oscillations with alpha rhythms (Fig. 7).Meanwhile, due to the transmission of neural activity across different brain regions through network, aperiodic activity in other brain regions also partly transformed into periodic activity, especially in the frontal and central lobe regions.Considering the post-acupuncture effect, the recovery of aperiodic activity in the occipital and parietal lobe regions is significantly slower than that in the frontal lobe, which may due to the interaction between these brain regions.
In summary, acupuncture leads to a reduction of the aperiodic exponent in the whole brain, which is a manifestation that the excitatory-inhibition balance of the brain shifts towards to high excitability with acupuncture stimulation.In resting state, aperiodic activity is mainly distributed in frontal and central lobe regions, which exhibit significant decline of aperiodic exponents during acupuncture.In addition, the aperiodic exponents of the parietal and occipital lobe regions also decrease during acupuncture, but the recovery of aperiodic activity in these regions after acupuncture is slow.It is further demonstrated that high frequency acupuncture stimulation performs superior modulatory efficacy on aperiodic brain activity.

D. Modulatory Effect of Distinct Acupuncture Manipulations on Brain Activity
In order to further analyze the modulatory efficacy of acupuncture stimulation on brain activity, we calculate the sensitivity index of various brain regions to different acupuncture manipulations by the integration of periodic and periodic measurements.Firstly, we compute the variations of adjusted power and aperiodic exponent values between Pre-Acup and Dur-Acup stages in each EEG electrode.Then, the average values of the two measurements for five brain regions are obtained according to the distribution of EEG electrodes, by which the sensitivity index of each brain region to acupuncture stimulation can be measured.The sensitivity index, which takes periodic and aperiodic features of EEG signals into account, exhibits the influence of acupuncture stimulation on brain activity in a comprehensive way.With the sensitivity index, the modulatory effect on difference brain regions can Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.

TABLE I THREE EVALUATION MEASUREMENTS OF FIVE BRAIN REGIONS TO FOUR ACUPUNCTURE MANIPULATIONS
and it is found that different acupuncture manipulations exert varying influences on distinct brain regions.Table I shows the measurements of adjusted power, aperiodic exponent and sensitivity index in five brain regions with four acupuncture manipulations.
Above results indicate that variations of periodic and aperiodic brain activities induced by different acupuncture manipulations are distinct from each other.The statistical data reveals that the parietal and occipital lobe regions are the regions where periodic brain activity is significantly influenced by acupuncture stimulation, while the effect on aperiodic brain activity is mainly concentrated in frontal and central lobe regions.According to the comprehensive measurement, the frontal lobes are the region that exhibits the strongest modulatory effect on brain activity with the Low LT and Low TR manipulations, with sensitivity index values of 0.026 and 0.039, respectively.In high frequency acupuncture manipulations, the parietal lobes are most sensitive to acupuncture stimulation.The sensitivity index values of parietal lobes with High LT and High TR are 0.055 and 0.057, respectively.
Based on above evaluation measurements, including periodic power and aperiodic exponent, a knowledge graph for acupuncture is drawn in Fig. 10.It integrates the variation of periodic and aperiodic neural activities with independent t-statistics for each brain region during stimulation on "Zusanli" acupoint, which serves as an estimation of acupuncture's effect size.Comparison between different acupuncture manipulations reveals that group-level increases of periodic power in alpha band and decreases of aperiodic exponents over brain regions, showing distinct efficacy of various acupuncture stimulations.It provides a guidance for optimizing acupuncture manipulations and improving the therapeutic effect in clinical practice.When acupuncture treatment is applied to neural diseases associated with specific brain regions, the acupuncture knowledge map can assist acupuncturists in selecting an optimized acupuncture manipulation or stimulation frequency to modulate the brain activity.

IV. CONCLUSION
In this work, we have designed an EEG-based monitoring system to evaluate the modulatory efficacy of acupuncture stimulation on brain activity with the periodic and aperiodic features.Three unique measurements of EEG signals, including adjusted power, aperiodic exponent and sensitivity index of each brain region, are calculated as the evaluation indicators for acupuncture effect.The performance and effectiveness of the proposed system are assessed by acupuncture experiments that involves four types of acupuncture manipulations.Experimental results demonstrate that periodic brain activity in alpha band is significantly enhanced during acupuncture, specifically in the parietal and occipital lobe regions.Compared to LT manipulations, TR manipulations demonstrate higher modulatory efficacy with sustained post-acupuncture effects.Furthermore, acupuncture decreases the aperiodic exponent value of EEG signal by modulating E/I balance of the brain.The significant decline of aperiodic exponents mainly occurs in frontal and central lobe regions during acupuncture, and high frequency acupuncture performs superior modulatory efficacy on aperiodic brain activity.By the integration of adjusted power and aperiodic exponent, the sensitivity index of each brain region to acupuncture stimulation is computed, by which distinct modulatory effects of four acupuncture manipulations is quantitatively assessed and a knowledge graph for acupuncture effect is established.The proposed system is valuable for assisting physicians in making accurate and rapid decisions on acupuncture and improve the therapeutic effect on brain disorders.

Fig. 1 .
Fig.1.The monitoring and evaluation system of acupuncture efficacy, consisting of "Zusanli" acupuncture experiment, EEG signal recording and EEG data processing.

Fig. 2 .
Fig. 2. The EEG processing pipeline that includes data recording, data preprocessing, features extraction, as well as the measurement and evaluation of acupuncture modulatory effect.

Fig. 3 .
Fig. 3. Periodic and aperiodic measurements of EEG power spectrum.Three periodic parameters and two aperiodic parameters are obtained by fitting the original PSD with FOOOF toolbox.

Fig. 4 .
Fig. 4. The time-frequency spectrograms of EEG signals with four acupuncture manipulations.(a) Low LT, (b) High LT, (c) Low TR, (d) High TR.The power of alpha band significantly increases in Dur-Acup and Post-Acup stages compared to the Pre-Acup stage.

Fig. 5 .
Fig. 5.The power spectral density of EEG signals with different acupuncture manipulations.(a) Low LT, (b) High LT, (c) Low TR, (d) High TR.There are power peaks in alpha band for each acupuncture stage and the peak value of Dur-Acup stage is the highest for each acupuncture manipulation.

Fig. 6 .
Fig. 6.The periodic measurements of adjusted alpha power across the whole brain with different acupuncture manipulations.(a) Low LT, (b) High LT, (c) Low TR, (d) High TR.Significance level: p < 0.05 ( * ), p < 0.01 ( * * ), and p < 0.001 ( * * * ).The adjusted power in alpha band is significantly enhanced by acupuncture, and the stimulation with high frequency exhibits a stronger modulatory effect on periodic brain activity.

Fig. 7 .
Fig. 7. Comparison of the spatial distribution of adjusted alpha power across the whole brain.Acupuncture stimulation mainly modulates the periodic brain activities in parietal and occipital lobe regions, and the modulation effect of high frequency acupuncture manipulations is more profound compared to that of low frequency acupuncture manipulations.

Fig. 8 .
Fig. 8.The measurements of aperiodic exponent across the whole brain with different acupuncture manipulations.(a) Low LT, (b) High LT, (c) Low TR, (d) High TR.The aperiodic exponent significantly decreases in Dur-Acup and Post-Acup stages.

Fig. 9 .
Fig. 9. Comparison of the spatial distribution of aperiodic exponent across the whole brain.The frontal and central lobe regions exhibit stronger modulatory effect of acupuncture stimulation on aperiodic brain activity.

Fig. 10 .
Fig.10.graph of acupuncture effect on different brain regions with independent t-statistics of periodic-aperiodic measurements between Pre-Acup and Dur-Acup stages.It exhibits the variation of periodic and aperiodic neural activities in each brain region during stimulation on "Zusanli" acupoint with various acupuncture manipulations.Colors in each electrode correspond to the independent t-statistics.Significant group-level increases and decreases are shown in yellow and in blue, respectively.