Article Text

other Versions


Motion patterns in acupuncture needle manipulation
  1. Yoonjeong Seo1,2,
  2. In-Seon Lee1,
  3. Won-Mo Jung1,3,
  4. Ho-Sun Ryu1,4,
  5. Jinwoong Lim1,
  6. Yeon-Hee Ryu5,
  7. Jung-Won Kang6,
  8. Younbyoung Chae1,3
  1. 1Acupuncture and Meridian Science Research Center, College of Korean Medicine, Kyung Hee University, Seoul, Korea
  2. 2College of Korean Medicine, Semyung University, Jecheon, Korea
  3. 3Department of Brain and Cognitive Engineering, Korea University, Seoul, Korea
  4. 4College of Korean Medicine, Wonkwang University, Iksan, Korea
  5. 5Acupuncture, Moxibustion and Meridian Research Center, Division of Standard Research, Korea Institute of Oriental Medicine, Daejeon, South Korea
  6. 6Department of Acupuncture and Moxibustion, College of Korean Medicine, Kyung Hee University, Seoul, Korea
  1. Correspondence to Professor Younbyoung Chae, Acupuncture and Meridian Science Research Center, Kyung Hee University, 1 Hoegi-dong, Dongdaemun-gu, Seoul, 130-701, Republic of Korea; ybchae{at}


Background In clinical practice, acupuncture manipulation is highly individualised for each practitioner. Before we establish a standard for acupuncture manipulation, it is important to understand completely the manifestations of acupuncture manipulation in the actual clinic. To examine motion patterns during acupuncture manipulation, we generated a fitted model of practitioners’ motion patterns and evaluated their consistencies in acupuncture manipulation.

Methods Using a motion sensor, we obtained real-time motion data from eight experienced practitioners while they conducted acupuncture manipulation using their own techniques. We calculated the average amplitude and duration of a sampled motion unit for each practitioner and, after normalisation, we generated a true regression curve of motion patterns for each practitioner using a generalised additive mixed modelling (GAMM).

Results We observed significant differences in rotation amplitude and duration in motion samples among practitioners. GAMM showed marked variations in average regression curves of motion patterns among practitioners but there was strong consistency in motion parameters for individual practitioners. The fitted regression model showed that the true regression curve accounted for an average of 50.2% of variance in the motion pattern for each practitioner.

Conclusions Our findings suggest that there is great inter-individual variability between practitioners, but remarkable intra-individual consistency within each practitioner.

  • Acupuncture
  • Medical Education & Training

Statistics from


The word ‘medicine’ originates from the Latin ars medicina, meaning the art of healing. Acupuncture techniques have been used as an art of healing for the treatment of various diseases and symptoms for more than 2500 years.1 Acupuncture manipulation has been considered a fundamental skill for practitioners to elicit the de qi sensation and thereby to maximise the effects of acupuncture treatment in clinical practice.2 Different manipulations of the acupuncture needle can induce different physiological responses in the nervous system3 ,4 and cardiovascular system.5 Recent studies suggested that quantification of the degree of acupuncture stimulation (eg, amplitude, frequency) can be a basic step for research into acupuncture.2 ,6 Using a motion and force sensor, Davis et al7 observed different patterns of needling motion and force between two types of acupuncture manipulation—namely, Chinese acupuncture and Japanese acupuncture. Li et al8 demonstrated that, on repeat testing, experts were more consistent in amplitude and lifting–thrusting velocity than novices based on the kinematic and kinetic features of the acupuncture manipulation.

It is sometimes claimed that the parameters of acupuncture manipulation should be standardised for the education system. To compare and evaluate acupuncture manipulation, Liu et al9 developed a novel acupuncture parameter analysis software, Acupuncture Manipulation Information Analysis System, by capturing the real-time position of needles during acupuncture manipulation. This software was shown to assist novices in learning the skills of acupuncture manipulation by quantifying their motion parameters.10 Furthermore, an intelligent three-dimensional interactive virtual reality system supported a force-feedback interface for needle insertion.11 However, many practitioners still depend solely on description in classical Chinese medical texts or verbal instruction from peers without objective information regarding quantitative parameters for acupuncture manipulation. Without a quantified standard for acupuncture manipulations, significant variations in specific physical parameters among practitioners are likely to remain.12

To propose a standard for acupuncture manipulation, we need to understand the manifestations of acupuncture manipulation in the actual clinic. Generally, acupuncture manipulation is dependent on each practitioner in clinical practice. Although practitioners are trained with a standardised protocol, they tend to establish their own motion pattern for acupuncture manipulation as they accumulate experience. Hence, considerable variation probably exists in acupuncture manipulation skills among practitioners, although there may be strong consistency in these embodied movements for a given practitioner. However, there has been no detailed exploration of motion pattern of experienced practitioners during acupuncture needle manipulations. Thus, it is important to analyse the motion patterns of each practitioner during acupuncture manipulation. To extract motion pattern units for each practitioner, it is necessary to construct a fitted model for a given motion pattern such as the generalised additive mixed model (GAMM).

In this study we used motion sensors to gather motion parameters for acupuncture manipulations, such as amplitude and frequency, as experienced practitioners rotated the acupuncture needle at the LI11 Quchi acupuncture point in a manner typically performed in the clinic. Based on the quantification of motion parameters during acupuncture manipulation, we generated a fitted model of motion patterns for each practitioner and evaluated the consistency of acupuncture manipulation.

Materials and methods


Eight right-handed licensed acupuncture practitioners (one female and seven male doctors of Korean medicine; mean age 39.4±0.7 years, range 37–43 years) and one healthy volunteer (man aged 27 years, height 175 cm, weight 70 kg) were recruited through poster advertisements on bulletin boards in the local community. All practitioners and the volunteer received a detailed explanation of the study and written informed consent was obtained.

Measurement of acupuncture manipulation consistency

We acquired real-time motion wave data (80.3 Hz sampling rate) from each practitioner during acupuncture manipulation using a motion sensor (Acusensor 2, Stromatec, Vermont, USA). We attached the motion sensor to the right LI11 of the healthy volunteer. LI11 is located on the lateral aspect of the elbow at the midpoint of the line connecting LU5 with the lateral epicondyle of the humerus.13 Practitioners were asked to manipulate an acupuncture needle at LI11 of the participant using a rotating technique in their own way as they would do normally on this acupuncture point. Practitioners inserted an acupuncture needle (J-type Japanese Seirin needle; 0.25×40 mm, Seirin, Japan) into the skin at LI11 through the motion sensor. After the motion sensors were calibrated using a two-axis actuator, each practitioner performed acupuncture manipulation for about 30 s (figure 1A) while the motion pattern of their acupuncture manipulation was measured. We evaluated only rotation-related information from the motion sensor.

Figure 1

Summary of data acquisition, processing and analysis of motion patterns during acupuncture manipulation by Practitioner 1. (A) Data acquisition: real-time motion wave data from each practitioner during acupuncture rotation manipulation at the LI11 acupuncture point using a motion sensor. (B) Preprocessing with infinite impulse response Butterworth filter. (C) Identification of motion unit based on unique rhythmic manipulating movement by the practitioner. (D) Identified sample motion unit. (E) Normalisation with resampling of the number of observed data to a specific number (100). (F) Normalisation with rescaling of the rotation amplitude between 0 and 1. (G) Generalised additive mixed modelling (GAMM) for the rotating motion M for a practitioner. GAM, generalised additive mixed.

Data processing

Based on preprocessing, the infinite impulse response (IIR) Butterworth filter was applied to the raw signal as a band pass filter to filter out fluctuating low-frequency signals (<0.2 Hz) and high-frequency noise signals (>5 Hz) (figure 1B). After preprocessing, a wave of raw signal was indicative of a rotating movement—that is, one cycle of relative movement of the first and second digit holding a needle. Raw signals for all practitioners were separated into repeated motion units because of practitioners’ unique rhythmic manipulations (figure 1C). A sampled motion unit consisted of a specific number of waves varying according to each practitioner's motion. The identification of a motion unit for each practitioner was confirmed by a recorded video clip of each practitioner's hand movement and a verbal report from the practitioner (figure 1D).

To generate a pattern template of rotating motion for each practitioner, we normalised the duration (length of sampled motion unit) and amplitude (magnitude of rotation) in every sampled unit using the Python SciPy package ( For the normalisation of duration we used the resampling method to adjust the number of observed data for each motion unit to a specific number (100) (figure 1E). For the normalisation of amplitude we rescaled the rotation amplitude of each motion unit between 0 and 1 (figure 1F).

Data analysis

Before normalisation we calculated the average amplitude and duration of a sampled motion unit for each practitioner, and the normalised motion units were analysed for each practitioner using a GAMM with methods implemented in the mgcv R package (

Generalised additive modelling (GAM) can handle both parametric and non-parametric effects of many predictors. GAMM, which is based on the GAM method, allows specification of smooth functions (non-parametric regression) within a mixed model framework.14 ,15 Considering the complex non-linearity of motion units, we conducted non-parametric regression allowing the functional form between y and x to be determined based on the data alone. In the GAMM, we modelled the rotating motion (M) for a practitioner as the sum of smooth function in time t and random intercepts for individual units (u) (figure 1G):Embedded Imagewhere M is motion for a practitioner, s(t) is a smooth function in time, and b(u) is random intercepts for individual units.

To evaluate the fitted models (pattern templates), we applied the adjusted R2 values for the motion pattern of each practitioner. Deviance explained is a quality-of-fit statistic based on a generalisation of the idea of using the sum of squares of residuals in ordinary least squares to cases where model fitting is achieved using maximum likelihood.


Baseline characteristics of practitioners

All practitioners had over 10 years of clinical experience in the acupuncture field, with a mean of 13.9 years (range 10–19 years) (table 1). Questionnaires completed by the practitioners indicated that they used different types of acupuncture manipulations: (a) rotating the needle (Rotation; 100%); (b) lifting and thrusting the needle (Displacement; 50%); (c) inserting the needle with or against the direction of the meridian/channel course (Direction; 62.5%); and (d) inserting and extracting the needle in coordination with the patient's respiration (Respiration; 37.5%). Rotating the needle was the most common acupuncture manipulation in our practitioners.

Table 1

Baseline characteristics of practitioners regarding acupuncture manipulation

Profiles of motion parameters during acupuncture manipulation

The needle motion profiles of all practitioners are shown in figure 2. We identified particular motion units for each practitioner during the entire acupuncture manipulation. In terms of amplitude, large differences were noted in rotation amplitude, with a mean of 0.99 revolutions (range 0.23–1.85). Large differences in the duration of rotation per motion pattern were also observed with a mean of 1.51 s (range 0.78–2.87 s).

Figure 2

Needle motion profiles of all practitioners. Blue lines separated by red dots are sampled motion units from raw signals. In terms of amplitude, considerable variation was observed in rotation amplitude: mean 0.99 revolutions; range 0.23–1.85 revolutions (orange bar). In terms of duration, strong inconsistency in rotation duration per pattern was identified: mean 1.51 s; range 0.78–2.87 s (blue bar).

Fitted model of motion patterns for each practitioner

Based on GAMM, we estimated the true regression curve of motion patterns for each practitioner during acupuncture manipulation (figure 3). In terms of inter-individual variability, considerable variation was identified in the average regression curve and the CI among practitioners. In contrast, significant intra-individual consistency in motion parameters was noted during acupuncture manipulation (mean adjusted R2=0.495, range 0.235–0.739). The fitted regression model showed that the true regression curves accounted for an average of 50.2% of the variance in the motion pattern for each practitioner (deviance explained; mean 50.2%, range 24.6–74.3%). It is possible that the more regular the motion patterns were, the greater the deviance explained became. Based on the consistency data, we could quantitatively estimate the consistency of acupuncture manipulations in each practitioner: Practitioner 1 (P1) (0.407, 41.7%); P2 (0.696, 69.9%); P3 (0.739, 74.3%); P4 (0.362, 37.1%); P5 (0.322, 33.1%); P6 (0.616, 62.0%); P7 (0.235, 24.6%); P8 (0.581, 58.7%).

Figure 3

Fitted model of motion pattern for each practitioner. With regard to inter-individual variability, considerable variation was observed in averaged regression curve (red line) and the CI (orange range) among practitioners. With regard to intra-individual variability, there were strong consistencies in motion parameters in individual practitioners during acupuncture manipulation (deviance explained: mean 50.2%; range 24.6–74.3%). GAM, generalised additive mixed.


The present study explored the motion patterns—specifically, amplitude and frequency—among practitioners during acupuncture manipulations in a simulated clinical situation. We observed significant differences in rotation amplitude and duration per sampled motion among practitioners as they manipulated acupuncture needles. In contrast, remarkable consistency was observed in motion patterns for each practitioner based on GAMM analysis.

Acupuncture manipulation skills have been passed down from generation to generation through apprentice education, and it is difficult for practitioners to convey information on specific physical motion parameters. Over time, motion patterns in acupuncture needle manipulations in the clinic varied. Results from previous studies demonstrated significant variations in specific physical parameters among practitioners in mainland China.12 A comparative study found significant differences in the lifting–thrusting and rotating amplitudes in needling performed by practitioners of two different acupuncture styles.7 In the current study, significant variations were noted in physical motion parameters such as amplitude and duration among practitioners as they performed acupuncture manipulation in their own fashion. Our findings further support previous studies that identified noticeable variability in physical motion parameters during acupuncture manipulations across individual practitioners.

The consistency of acupuncture manipulation skills was enhanced after novice medical students practised their skills.10 After dozens of training sessions with intelligent virtual environments for Chinese acupuncture learning and training, students improved the accuracy and consistency of needle insertion and manipulation.11 Practitioners are generally trained in their professional skills for acupuncture manipulations following standardised protocols established over time, after which they consolidate and retain their own pattern during clinical practice with accumulated experience. These enhanced motor performances in experienced acupuncture practitioners could be explained by modulation of patterns of local resting state activity in the brain due to increased strength of regional clustering involved in advanced local information processing efficiency.16 In the current study, we observed significant consistency in motion parameters in individual practitioners using GAMM methods (deviance explained: mean 50.2%, range 24.6–74.3%). For example, practitioners 2 and 3 showed greater consistency of motion pattern during acupuncture manipulation (figure 3, P2 and P3) and the fitted model for the motion pattern accounted for 69.9% and 74.3% of the variance in their motion patterns, respectively. On the other hand, practitioners 5 and 7 showed less consistency in their motion patterns during acupuncture manipulation (figure 3, P5 and P7) and the fitted model only accounted for 33.1% and 24.6% of the variance in their motion patterns, respectively. Based on analysis of the consistencies in motion parameters, we can propose a novel evaluation model for the stabilities of acupuncture manipulations for each practitioner. Quantification of the motion pattern during acupuncture manipulations would help people explore whether or not acupuncture manipulations contribute to the clinical outcomes in a future study.

The analysis of motion patterns during acupuncture manipulation revealed great inter-individual variability and remarkable intra-individual consistency within practitioners. Acupuncture experts consolidate their own patterns of acupuncture manipulation with cumulative clinical experience. In a future study it would be interesting to compare the motion patterns during acupuncture manipulation between experienced practitioners and novices, and to explore the process of visuomotor learning when the novices acquire the motion patterns of acupuncture manipulation from the master.

Summary points

  • Needle manipulation has been linked to physiological response.

  • Considerable variation exists in acupuncture manipulation techniques among practitioners.

  • Individual practitioners showed high consistency of their embodied movements.


View Abstract


  • Contributors YS, H-SR and YC designed the study, analysed the data and drafted and revised the paper. I-SL and W-MJ co-designed the study, monitored data collection and analysed the data. JL, J-WK and Y-HR drafted relevant sections of the paper.

  • Funding This research was supported by the 2014 KIOM Undergraduate Research Program funded by the Korea Institute of Oriental Medicine.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval All procedures in this study were conducted in accordance with the guidelines of the Declaration of Helsinki (1954) and approved by the Institutional Review Board of Korea University, Seoul, Republic of Korea.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.