Original Article

Split Viewer

Journal of Acupuncture Research 2023; 40(1): 44-52

Published online February 28, 2023

https://doi.org/10.13045/jar.2022.00325

© Korean Acupuncture & Moxibustion Medicine Society

Development and Validation of a Vision-Based Needling Training System for Acupuncture on a Phantom Model

Trong Hieu Luu1 , Hoang-Long Cao1,2 , Duy Duc Pham3 , Le Trung Chanh Tran1 , Tom Verstraten2 ,

1Department of Automation Technology, College of Engineering, Can Tho University, Can Tho, Vietnam
2Brubotics and Flanders Make, Vrije Universiteit Brussel, Brussels, Belgium
3Faculty of Traditional Medicine, Can Tho University of Medicine and Pharmacy, Can Tho, Vietnam

Correspondence to : Hoang-Long Cao
Department of Automation Technology, College of Engineering, Can Tho University, Campus II, 3/2 street, Ninh Kieu District, Can Tho 90000, Vietnam
E-mail: hoang.long.cao@vub.be

Received: November 4, 2022; Accepted: November 25, 2022

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Background: Previous studies have investigated technology-aided needling training systems for acupuncture on phantom models using various measurement techniques. In this study, we developed and validated a vision-based needling training system (noncontact measurement) and compared its training effectiveness with that of the traditional training method.
Methods: Needle displacements during manipulation were analyzed using OpenCV to derive three parameters, i.e., needle insertion speed, needle insertion angle (needle tip direction), and needle insertion length. The system was validated in a laboratory setting and a needling training course. The performances of the novices (students) before and after training were compared with the experts. The technology- aided training method was also compared with the traditional training method.
Results: Before the training, a significant difference in needle insertion speed was found between experts and novices. After the training, the novices approached the speed of the experts. Both training methods could improve the insertion speed of the novices after 10 training sessions. However, the technology-aided training group already showed improvement after five training sessions. Students and teachers showed positive attitudes toward the system.
Conclusion: The results suggest that the technology-aided method using computer vision has similar training effectiveness to the traditional one and can potentially be used to speed up needling training.

Keywords Acupuncture; Computers; Education; Training

Acupuncture has been considered an important therapy in East Asian medicine (i.e., China, Japan, Korea, and Vietnam) and has spread worldwide since a few centuries ago [1-3]. Acupuncture therapy refers to the practice of inserting one or more needles into specific locations (acupoints) on the human body surface [4-6]. To obtain therapeutic effects, practitioners must have a sufficient understanding of acupuncture [7,8]. Incompetent acupuncture skills can cause pain and hematoma and even lead to bent and broken needles [9]. Therefore, novice acupuncturists must follow standard training procedures such as acupoint location training and needling training [10,11].

While acupoint location is outside the patient’s body, needle manipulation is an invasive technique on the body surface, involving different parameters, e.g., needle insertion depth, needle tip direction, and needle movement to obtain deqi [12]. Thus, needling training is often conducted on phantom models to mimic the human body surface. These models are primarily based on silicon, gel, animal skin, apple, and cucumber [11]. Needling training is traditionally executed through an apprenticeship in which the students observe and try to replicate the practice of expert acupuncturists. Then, the performance of the novices is evaluated by the experts. Recently, several studies have developed technology-aided needing training systems with quantification and visualization of needle manipulation parameters. Results showed that training with such systems could improve the needle manipulation skills of novices [12-14]. However, no direct comparison with the traditional training method has been made.

Regarding measurement methods, research in acupuncture needling training has used various sensing technologies to measure the kinetic and kinematic parameters of needle manipulation [11,15]. Attaching sensors directly to the needle is the most widely used technique for real-time measurement of parameters. The most compact sensing device in this technique is Acusensor [16], which has been used in several needling training studies [12-14]. Another technique is using advanced motion tracking systems with physical markers attached to the needle or fingers [9,17-20]. However, these two techniques strongly influence the naturalness of finger movement and sensation while performing needle manipulation. Thus, noncontact measurements have much less influence. A popular noncontact technique is placing pressure sensors attached to the phantom mode; however, this can only measure kinetic parameters, i.e., the axial force and its derived parameters [9,21]. Some proposed using ultrasound images for needle motion tracking [22-25]. The training setup using this technique might not be portable and affordable for training on a large scale. Therefore, needle motion tracking without using physical markers in needling training is an urgent need, which is challenging given the small sizes of acupuncture needles compared with regular injection needles [26,27]. Drawbacks of this technique include measurement of kinematic parameters only and lower measurement accuracy [15]. However, it can be still applicable for certain needling training lessons considering the advantages of noncontact measurement and affordability.

In this study, we present a needling training system for acupuncture using a needle-tracking technique based on computer vision. Needle manipulation parameters were quantified by analyzing needle displacements from images captured by a camera. The system was validated in a needling training course with expert and novice acupuncturists in two studies. The selected training lesson aimed to improve the needle insertion speed within certain limits of needle insertion length and needle insertion angle (needle tip direction). In the first study, the ability to characterize the needle manipulation skills of the proposed system was investigated by comparing the needle manipulation parameters of experts and novices before and after the training. In the second study, the training effectiveness of displaying needle manipulation parameters during training (technology-aided training method) was compared with that of the traditional training method (control group). The technology satisfaction of students and teachers in the technology-aided training group was investigated.

1. Methodology

The proposed vision-based needling training system was developed following a co-creation practice in which all stakeholders were involved in the entire system development [28]. Requirements were first established by expert acupuncturists who practice and train acupuncture skills daily. The targeted needle manipulation parameters are needle insertion speed, needle insertion length, and needle insertion angle (needle tip direction) (Fig. 1B). During the prototyping process, experts iteratively provided feedback to system developers. The final prototype was tested in a laboratory setting before the validation process in the actual needling training environment.

Fig. 1. Vision-based needling training system. (A) The experimental setup includes a phantom model, a camera, a white LED light, and a dark background. Images captured by the camera are processed by a computer to determine needle manipulation parameters. (B) Needle manipulation parameters. (C) The needle handle is detected during needle manipulation and visualized by a red bounding rectangle.

2. System design

1) Setup

The vision-based needling training system is shown in Fig. 1A. The system includes a camera (TD-WC3200, The Deosd), a silicon-based phantom model (Henan Green Medical Tech Co., Ltd., Henan, China), a filiform acupuncture needle (haozhen; Suzhou Tianxie Acupuncture Instruments Co., Ltd., Suzhou, China), a white LED light (1.2 W), and a dark background (PMMA, 3 mm). Images captured by the camera were processed by a computer (Ubuntu 20.4; Canonical Ltd., London, UK) using the OpenCV library in Python [29]. The setup dimensions were 50 cm × 40 cm × 40 cm. The camera was installed at a height of 20.5 cm above the table. The filiform needle was 50 mm in length with the handle part of ϕ1 mm and body part of ϕ0.3 mm. The handle part of the needle was painted blue to create a contrast with the dark background and environment. The white LED light was used to improve the lighting condition if necessary.

2) Needle detection

Detecting the acupuncture needle and its movement is the core function of the system. Since the filiform needle is extremely thin, needle detection was performed through the blue-painted handle. Before capturing images, a camera calibration procedure was performed by applying Zhang’s method to remove distortion [30]. A chessboard was placed perpendicular to the phantom model to set the camera center. At this position, the size of the three squares was 170 × 170 pixels and was equivalent to 30 × 30 mm. The pixel-to-mm ratio was 0.177. Captured images (frames) were converted from RGB to HSV color space (with erosion and dilation) to detect the blue-painted needle handle. A threshold method was then applied with the minimum and maximum threshold parameters of HSV of (0, 0, 0) and (179, 255, 255), respectively. The detected needle handle was visualized by a red bounding rectangle (Fig. 1C). The width, height, minimum angle of rotation, and center point of the bounding rectangle can be estimated to determine needle manipulation parameters.

3) Needle manipulation parameters

Three needle manipulation parameters can be determined, i.e., needle insertion speed, needle insertion length, and needle insertion angle. The needle insertion movement was characterized by the starting (xS, yS) and ending (xE, yE) positions of the needle handle, corresponding to the start and end frames (Fig. 2). The time interval between two consecutive frames was 35 ms. The start position was updated when the needle was detected and if its offset position was 5 pixels higher than its previous position. The end position was the last frame when the needle stopped moving. The needle insertion length was determined by the distance between these two points.

Fig. 2. Calculation of needle manipulation parameters based on captured frames. The time interval between two consecutive frames is 35 ms.

The needle insertion angle was determined by the minimum angle of needle rotation with respect to the horizontal axis at the end position (end frame). The needle insertion speed was calculated through the normalized cross-correlation method to improve the accuracy of calculating the center point displacement between two consecutive frames. This method was based on the similarity between the image and the feature [31]. The correlation coefficient Ci of the image (frame i) and the feature (frame i+1) in 35 ms was comparable to the center point displacement. The instant speed at frame i was Ci divided by 35 ms. The needle insertion speed was then calculated with the assumption that it followed the trapezoidal rule.

4) Laboratory test

The system was tested in a laboratory setting. A graphical user interface (GUI) was developed to set up and configure the system for practical usage. After each trial during training, the calculated needle manipulation parameters were displayed on a computer screen (Fig. 3A, B). Novice acupuncturists can instantly monitor their training performance trial by trial. Feedback from the expert acupuncturists confirmed that the developed GUI was applicable for certain needling training lessons. One of the potential training lessons was selected for validation in Sect. 3.

Fig. 3. Laboratory test. (A) An acupuncturist performs needle manipulation skills on the system. (B) The needle manipulation parameters are displayed on the graphical user interface.

3. Validation

The proposed system was validated in a needling training course with expert and novice acupuncturists in two studies to investigate the ability to characterize needle manipulation skills and the effectiveness of the system on needling training. The training lesson was selected to improve the needle insertion speed while controlling the insertion length between 0.5 and 1 cm and the needle insertion angle between 80° and 90° for safety reasons. At these angle limits where the needle is nearly perpendicular to the skin surface, the insertion length and the insertion depth can be used interchangeably. This training is essential because a safe insertion with a high speed can reduce pain in body acupuncture [32,33]. This is even more important with filiform needles when several needles must be inserted (typically from 4 to 20) [34].

Acupuncture activities were performed on a silicon-based phantom model. Needles were not inserted in the human body. No human patients were involved. The study was recognized by the Faculty of Traditional Medicine, conducted in accordance with the ethical guidelines provided by Can Tho University of Medicine and Pharmacy (CTUMP), and pre-registered before data collection on OSF (osf.io/af2uz).

4. Participants

The study recruited 19 expert acupuncturists (6 females, 13 males) from CTUMP. The experts have on average 17.00 (standard deviation [SD], 6.30) years of experience. The novice acupuncturists were 54 students from two classes of the training course (40 females, 14 males; mean age, 22.12; SD, 1.14). They did not follow any acupuncture training before the needing training course. The first class followed traditional needling training through apprenticeship (control group, n = 27). The second class followed a technology-aided needling training using the proposed system (experimental group, n = 27). The students were guided by five teachers (two females, three males).

5. Procedure

1) Experts

All experts received an explanation of the study and provided informed consent. Then, they were asked to perform needle manipulation skills in five trials. After each trial, needle manipulation parameters were measured by the system and recorded by an experimenter.

2) Novices (students)

In the needling training course, all students received an explanation of the study and provided informed consent. The training course lasted for 2 days with five 50-min training sessions each day (8.33 h in total). On training day 1, both groups first received the same theory lecture about acupuncture techniques and familiarized themselves with the needle and phantom model. Before the training, they were asked to perform needle manipulation to determine the baselines (pretest). Then, each group practiced needling manipulation following the assigned training approach. After each day (sessions 5 and 10), needle manipulation skills were evaluated by five valid trials. After each valid trial, needle manipulation parameters were measured by the system and recorded by an experimenter. Finally, at the end of training day 2, students and teachers were asked to complete a technology satisfaction survey using Google Forms.

6. Measurement

Needle manipulation was measured through needle insertion speed (cm/s), needle insertion depth (cm), and needle insertion angle (degree). A trial was considered invalid (or unsafe) if the needle insertion depth and/or the needle insertion angle exceeded the corresponding minimum and maximum limits even if the insertion speed was high. Therefore, valid trials were recorded for equitable comparison purposes. In other words, a very fast insertion but outside the depth and angle limits (unsafe) cannot be used to compare with a slower speed insertion speed within the safety limits.

Within the safety limits, an optimal insertion should have an insertion depth between 0.7 to 0.8 cm and an insertion angle of exactly 90°. The depth “error” of a trial was defined as the distance between the actual depth to the optimal range.

Technology satisfaction was measured by a 5-point Likert scale survey using four constructs of the unified theory of acceptance and use of technology [35]. Perceived usefulness (PU), perceived ease of use (PEOU), and intrinsic motivation (IM) were used for all teachers and students. Additionally, the intention to use (ITU) construct was used for teachers. The wording of these constructs was adapted to the context of needling training.

7. Data analysis

Data were analyzed in Python using pingouin, scipy, and scikit posthocs packages. Normality and homogeneity were checked by the Shapiro–Wilk and Bartlett tests to determine the statistical tests. The significance level was set at 0.05.

1. Study I: characterization of needle manipulation skills

The needle manipulation skills of experts and novices before and after the training were compared (Fig. 4). A clear difference in the mean insertion speeds was found between experts and novices before the training (Fig. 4A). An independent t-test showed that the experts were significantly faster (M = 11.16; SD = 1.52) than the novices (M = 8.39; SD = 2.03; t(71) = 5.36; p < 0.001; d = 1.43). The novices improved their mean insertion speed after the training (48/54, 89%). A paired t-test showed that the mean insertion speed of the novices after the training was significantly improved (M = 10.53; SD = 1.85; t(53) = 7.43; p < 0.001; d = 1.09). An independent t-test showed that this speed was not significantly lower than those of the experts in study 1 (M = 11.16; SD = 1.52; t(71) = 1.31, n.s.).

Fig. 4. Needle manipulation skills of experts and novices. (A) Experts had a higher needle insertion speed than novices before the training. Novices improved their speed after training. (B) No differences were found in the insertion depth error. (C) Statistical differences were found in the insertion angle but were practically negligible. Error bars show the standard deviation.

Similar comparisons were conducted with the needle insertion depth error and insertion angle. No differences were found in the needle insertion depth error (Fig. 4B). Statistically significant but practically negligible differences in the insertion angle were found before (M = 89.25; SD = 1.01) and after (M = 88.90; SD = 0.83; U = 321.5; p < 0.05) training of the experts (M = 88.52; SD = 1.44; W = 349.0; p < 0.05) (Fig. 4C).

2. Study II: training effectiveness

The data of each group indicated that both groups showed improvement after 10 training sessions (traditional: 23/27, 85%; technology-aided: 25/27, 93%). However, the improvement happened at different timescales (Fig. 5A).

Fig. 5. Training results of the traditional training group (Trad.) and the technology-aided group (Tech.). (A) Both groups showed improvement after 10 training sessions (S10). (B) The technology-aided group had a better improvement after five training sessions (S5). No difference in improvement was found in both groups after 10 sessions (S10). Error bars show the 95% confidence interval.
1) Traditional training group

A Friedman test for repeated measures showed a significant difference in the mean insertion speed during training (χ2(2) = 19.23; p < 0.001; W = 0.36). A Nemenyi post-hoc test showed that the mean insertion speed was not significantly improved after five training sessions (S5, M = 9.75; SD = 1.79, n.s.) but only after 10 training sessions (S10, M = 11.12; SD = 1.54; p < 0.01) compared with the pretest value (pretest, M = 8.93; SD = 2.13). The improvement happened between sessions 5 and 10 (p < 0.01).

2) Technology-aided training group

A Friedman test for repeated measures showed a significant difference in the mean insertion speed during training (χ2(2) = 22.30; p < 0.001; W = 0.41). A Nemenyi post-hoc test showed that the mean insertion speed was already significantly improved after five training sessions (S5, M = 9.77; SD = 1.78; p < 0.01) and was maintained after 10 training sessions (S10, M = 9.94; SD = 1.95; p < 0.01) compared with the pretest value (pretest, M = 7.85; SD = 1.76). The improvement happened within the first five sessions (p < 0.01) and no further improvement within the later sessions.

3) Between-group comparison

Statistical tests on the gained mean scores were conducted because our data did not meet all assumptions for the analysis of covariance (Fig. 5B) [36]. After five training sessions, a Mann–Whitney U test showed that the technology-aided group obtained a significantly higher gain (M = 1.92; SD = 1.72) than the traditional group (M = 0.82; SD = 2.12; U = 236.0; p < 0.05; cles = 0.32). After 10 training sessions, a Mann–Whitney U test showed no significant difference between the technology-aided group (M = 2.09; SD = 2.29) and the traditional group (M = 2.19; SD = 1.88; U = 430.0, n.s.).

4) Technology satisfaction survey

Results from the self-reported technology satisfaction survey showed that students and teachers had positive attitudes toward the system. The technology-aided group rated the system useful (PU: M = 4.69, SD = 0.40) and easy to use (PEOU: M = 4.74, SD = 0.32) and felt motivated while using the system (IM: M = 4.84, SD = 0.29). The ratings from the teachers were also positive (PU: M = 4.43, SD = 0.36; PEOU: M = 4.50, SD = 0.42, IM: M = 4.47, SD = 0.45). They also showed a high ITU the system (ITU: M = 4.40, SD = 0.39).

Results from study I revealed a clear difference in the needle insertion speed between experts and novices before the training, which is in light with findings from previous studies on other kinematic and kinetic parameters [9,17,19,32]. In acupuncture, errors in depth and angle could cause more severe consequences than a low insertion speed. With little time for familiarization with the insertion movement before training, the novices focused on controlling the correct depth and angle. Therefore, it restrained them from getting insertion speed higher than the experts. This finding matches with feedback from the experts and teachers of the training course: (i) a novice needs intensive practice to obtain a higher insertion speed; (ii) controlling a correct depth and a correct angle requires a bit less effort.

After training, the novices could improve their insertion speed and approach the expert speed. This is consistent with findings from several previous studies with other parameters showing improvement after training [12-14]. However, this does not mean the novices reached the expert level in acupuncture skills in general but solely in this specific insertion speed training.

In study II, the technology-aided training approach resulted in a training effect after 10 training sessions similar to the traditional training approach. However, the technology-aided approach showed faster learning progress at the beginning of the training. The technology satisfaction survey indicated two possible explanations for this positive effect. First, both students and teachers considered the system useful and easy to use. Being informed of the needle manipulation parameters during training could potentially help students review their movements after each trial. Second, the students using the system reported high motivation. This can be explained by the novelty effect i.e., students felt excited about training with a new technological system [37]. Assuming that the performance after 10 training sessions is the maximal performance in the Thurstone learning curve in health professional education [38], the technology-aided training group reached their best performance faster. A fine-grained collection of data is required to validate this assumption [39].

We developed and evaluated a vision-based needling training system on a phantom model. The measurement method is noncontact, allowing the acupuncturists to perform needle manipulation skills naturally. The system can quantify three needle manipulation parameters, i.e., needle insertion length, needle insertion angle (needle tip direction), and needle insertion speed. The system was evaluated in a laboratory setting and a needling training course. Results showed that the system could characterize needle manipulation skills and resulted in faster learning progress in the first part of the training course than the traditional training method. More studies are needed to determine if this result can be replicated.

The system has several limitations that can be improved in future work. First, the system tracks motion by using computer vision and is therefore limited to kinematic parameters. Additional sensors will be integrated to quantify kinetic parameters for other training lessons. Second, the positive training effect could partially come from the novelty effect. Future studies should investigate the training effectiveness across multiple training lessons.

The authors would like to thank Tran Minh Man and Le Hong Phuoc for their support in organizing the validation and the experts and students for their participation in this work.

Conceptualization: THL, HLC, DDP, LTCT, TV. Data curation: THL, LTCT, DDP. Data analysis: THL, HLC. Writing original draft: THL, HLC. Writing - review and editing: All authors.

The work leading to these results has received funding from Can Tho University of Medicine and Pharmacy and partially from Flanders Make and VLAIO under the project AUTOCRAFT-SBO.

The authors declare that this research was conducted in full accordance with the Declaration of Helsinki. The study was recognized by the Faculty of Traditional Medicine, Can Tho University of Medicine and Pharmacy (329/QD-BM.YHCT) and followed the ethical guidelines provided by the organization. Needles were not inserted in a human body. No human patients were involved.

  1. Kaptchuk TJ. Acupuncture: theory, efficacy, and practice. Ann Intern Med 2002;136:374-383. doi: 10.7326/0003-4819-136-5-200203050-00010.
    Pubmed CrossRef
  2. White A, Ernst E. A brief history of acupuncture. Rheumatology (Oxford) 2004;43:662-663. doi: 10.1093/rheumatology/keg005.
    Pubmed CrossRef
  3. WHO. Acupuncture review and analysis of reports on controlled clinical trials. World Health Organization, 2002.
    CrossRef
  4. Liu T. Acupuncture: what underlies needle administration? Evid Based Complement Alternat Med 2009;6:185-193. doi: 10.1093/ecam/nen002.
    Pubmed KoreaMed CrossRef
  5. Nasir LS. Acupuncture. Prim Care 2002;29:393-405. doi: 10.1016/s0095-4543(01)00007-0.
    Pubmed KoreaMed CrossRef
  6. Ernst E. Acupuncture--a critical analysis. J Intern Med 2006;259:125-137. doi: 10.1111/j.1365-2796.2005.01584.x.
    Pubmed CrossRef
  7. Miao P, Zhang Q, Ji YY, Jiang J, Guo HD, Wang FB, et al. Quantitative assessment of expert's lifting-thrusting manipulation in acupuncture based on haptic device. Appl Mech Mater 2013;397-400:2507-2510. doi: 10.4028/www.scientific.net/AMM.397-400.2507.
    CrossRef
  8. Li J, Wu MX. [Ideas of standardization evaluation on acupuncture skills: enlightened by quantitative appraisal of surgical skills in Europe and North America]. Zhen Ci Yan Jiu 2011;36:449-452. Chinese.
    Pubmed
  9. Li J, Grierson LE, Wu MX, Breuer R, Carnahan H. Perceptual motor features of expert acupuncture lifting-thrusting skills. Acupunct Med 2013;31:172-177. doi: 10.1136/acupmed-2012-010265.
    Pubmed CrossRef
  10. Shin SG, Im IC. Satisfaction level of clinical practice and related variables for students in the department of radiology. J Korea Contents Assoc 2010;10:276-284. doi: 10.5392/JKCA.2010.10.9.276.
    CrossRef
  11. Jang JE, Lee YS, Jang WS, Sung WS, Kim EJ, Lee SD, et al. Trends in acupuncture training research: focus on practical phantom models. J Acupunct Res 2022;39:77-88. doi: 10.13045/jar.2022.00024.
    CrossRef
  12. Lee IS, Lee YS, Park HJ, Lee H, Chae Y. Evaluation of phantom-based education system for acupuncture manipulation. PLoS One 2015;10:e0117992. doi: 10.1371/journal.pone.0117992.
    Pubmed KoreaMed CrossRef
  13. Lim JW, Jung WM, Lee IS, Seo YJ, Ryu HS, Ryu YH, et al. Development of acupuncture manipulation education system. The Acupuncture 2014;31:11-19. doi: 10.13045/acupunct.2014049.
    CrossRef
  14. Jung WM, Lim J, Lee IS, Park HJ, Wallraven C, Chae Y. Sensorimotor learning of acupuncture needle manipulation using visual feedback. PLoS One 2015;10:e0139340. doi: 10.1371/journal.pone.0139340.
    Pubmed KoreaMed CrossRef
  15. Lyu R, Gao M, Yang H, Wen Z, Tang W. Stimulation parameters of manual acupuncture and their measurement. Evid Based Complement Alternat Med 2019;2019:1725936. doi: 10.1155/2019/1725936.
    Pubmed KoreaMed CrossRef
  16. Davis RT, Churchill DL, Badger GJ, Dunn J, Langevin HM. A new method for quantifying the needling component of acupuncture treatments. Acupunct Med 2012;30:113-119. doi: 10.1136/acupmed-2011-010111.
    Pubmed KoreaMed CrossRef
  17. Tang WC, Yang HY, Liu TY, Gao M, Xu G. Motion video-based quantitative analysis of the 'lifting-thrusting' method: a comparison between teachers and students of acupuncture. Acupunct Med 2018;36:21-28. doi: 10.1136/acupmed-2016-011348.
    Pubmed KoreaMed CrossRef
  18. Xu LL, Wang F, Yang HY, Tang WC. Three-dimensional finger motion tracking during needling: a solution for the kinematic analysis of acupuncture manipulation. J Vis Exp 2021;176:e62750. doi: 10.3791/62750.
    Pubmed CrossRef
  19. Xu LL, Xie J, Yang HY, Wang F, Tang WC. Operation stability analysis of basic acupuncture manipulation based on three-dimensional motion tracking data. Wirel Commun Mob Comput 2022;2022:1958984. doi: 10.1155/2022/1958984.
    CrossRef
  20. Zhang A, Yan XK, Liu AG. [An introduction to a newly-developed "Acupuncture Needle Manipulation Training-evaluation System" based on optical motion capture technique]. Zhen Ci Yan Jiu 2016;41:556-559. Chinese.
    Pubmed
  21. Liu TT, Wang ZH, Zhang G, Li H. Application research of neural network in acupuncture training system. Comput Technol Autom 2020;39:102-107.
  22. Leow MQ, Cao T, Lee SH, Cui SL, Tay SC, Ooi CC. Ultrasonography in acupuncture: potential uses for education and research. Acupunct Med 2016;34:320-322. doi: 10.1136/acupmed-2016-011182.
    Pubmed CrossRef
  23. Langevin HM, Konofagou EE, Badger GJ, Churchill DL, Fox JR, Ophir J, et al. Tissue displacements during acupuncture using ultrasound elastography techniques. Ultrasound Med Biol 2004;30:1173-1183. doi: 10.1016/j.ultrasmedbio.2004.07.010.
    Pubmed CrossRef
  24. Konofagou EE, Langevin HM. Using ultrasound to understand acupuncture. Acupuncture needle manipulation and its effect on connective tissue. IEEE Eng Med Biol Mag 2005;24:41-46. doi: 10.1109/memb.2005.1411347.
    Pubmed CrossRef
  25. Chen YL, Hou MC, Chang SC, Chuang KW, Lee PY, Huang CC. Development and evaluation of inexpensive ultrasound using A-mode and M-mode signals to identify lung depth and avoid risk of pneumothorax in acupuncture. J Med Biol Eng 2021;41:251-259. doi: 10.1007/s40846-020-00584-x.
    CrossRef
  26. Chou L, Hong JY, Hong C. A new technique for acupuncture therapy and its effectiveness in treating fibromyalgia syndrome: a case report. J Musculoskelet Pain 2008;16:193-198. doi: 10.1080/10582450802161978.
    CrossRef
  27. Chou LW, Hsieh YL, Kuan TS, Hong CZ. Needling therapy for myofascial pain: recommended technique with multiple rapid needle insertion. Biomedicine (Taipei) 2014;4:13. doi: 10.7603/s40681-014-0013-2.
    Pubmed KoreaMed CrossRef
  28. Sanders EBN, Stappers PJ. Co-creation and the new landscapes of design. CoDesign 2008;4:5-18. doi: 10.1080/15710880701875068.
    CrossRef
  29. Bradski G, Kaehler A. Learning OpenCV: computer vision with the OpenCV library. O'Reilly Media, 2008.
    CrossRef
  30. Zhang Z. Flexible camera calibration by viewing a plane from unknown orientations. IEE, pp 666-673, 1999.
    CrossRef
  31. Lewis JP. Fast normalized cross-correlation. In: Denis Laurendeau, ed. Vision Interface. Canadian Image Processing and Pattern Recognition Society, pp 120-123, 1995.
    CrossRef
  32. Ikai Y, Migaki M, Kida N, Iwamoto H, Hamada H. Analysis of the skills to acupuncture. In: Duffy V, ed. Digital Human Modeling. Applications in Health, Safety, Ergonomics and Risk Management: Human Modeling. Springer, pp 66-73, 2015.
    CrossRef
  33. Yin CS, Kim JH, Park HJ. High-velocity insertion of acupuncture needle is related to lower level of pain. J Altern Complement Med 2011;17:27-32. doi: 10.1089/acm.2010.0120.
    Pubmed CrossRef
  34. Lao L. Acupuncture techniques and devices. J Altern Complement Med 1996;2:23-25. doi: 10.1089/acm.1996.2.23.
    Pubmed CrossRef
  35. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information technology: toward a unified view. MIS Q 2003;27:425-478. doi: 10.2307/30036540.
    CrossRef
  36. Wright DB. Comparing groups in a before-after design: when t test and ANCOVA produce different results. Br J Educ Psychol 2006;76(Pt 3):663-675. doi: 10.1348/000709905X52210.
    Pubmed CrossRef
  37. Clark RE. Reconsidering research on learning from media. Rev Educ Res 1983;53:445-459. doi: 10.3102/00346543053004445.
    CrossRef
  38. Howard NM, Cook DA, Hatala R, Pusic MV. Learning curves in health professions education simulation research: a systematic review. Simul Healthc 2021;16:128-135. doi: 10.1097/SIH.0000000000000477.
    Pubmed CrossRef
  39. Pusic MV, Boutis K, Hatala R, Cook DA. Learning curves in health professions education. Acad Med 2015;90:1034-1042. doi: 10.1097/ACM.0000000000000681.
    Pubmed CrossRef

Article

Original Article

Journal of Acupuncture Research 2023; 40(1): 44-52

Published online February 28, 2023 https://doi.org/10.13045/jar.2022.00325

Copyright © Korean Acupuncture & Moxibustion Medicine Society.

Development and Validation of a Vision-Based Needling Training System for Acupuncture on a Phantom Model

Trong Hieu Luu1 , Hoang-Long Cao1,2 , Duy Duc Pham3 , Le Trung Chanh Tran1 , Tom Verstraten2 ,

1Department of Automation Technology, College of Engineering, Can Tho University, Can Tho, Vietnam
2Brubotics and Flanders Make, Vrije Universiteit Brussel, Brussels, Belgium
3Faculty of Traditional Medicine, Can Tho University of Medicine and Pharmacy, Can Tho, Vietnam

Correspondence to:Hoang-Long Cao
Department of Automation Technology, College of Engineering, Can Tho University, Campus II, 3/2 street, Ninh Kieu District, Can Tho 90000, Vietnam
E-mail: hoang.long.cao@vub.be

Received: November 4, 2022; Accepted: November 25, 2022

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Background: Previous studies have investigated technology-aided needling training systems for acupuncture on phantom models using various measurement techniques. In this study, we developed and validated a vision-based needling training system (noncontact measurement) and compared its training effectiveness with that of the traditional training method.
Methods: Needle displacements during manipulation were analyzed using OpenCV to derive three parameters, i.e., needle insertion speed, needle insertion angle (needle tip direction), and needle insertion length. The system was validated in a laboratory setting and a needling training course. The performances of the novices (students) before and after training were compared with the experts. The technology- aided training method was also compared with the traditional training method.
Results: Before the training, a significant difference in needle insertion speed was found between experts and novices. After the training, the novices approached the speed of the experts. Both training methods could improve the insertion speed of the novices after 10 training sessions. However, the technology-aided training group already showed improvement after five training sessions. Students and teachers showed positive attitudes toward the system.
Conclusion: The results suggest that the technology-aided method using computer vision has similar training effectiveness to the traditional one and can potentially be used to speed up needling training.

Keywords: Acupuncture, Computers, Education, Training

INTRODUCTION

Acupuncture has been considered an important therapy in East Asian medicine (i.e., China, Japan, Korea, and Vietnam) and has spread worldwide since a few centuries ago [1-3]. Acupuncture therapy refers to the practice of inserting one or more needles into specific locations (acupoints) on the human body surface [4-6]. To obtain therapeutic effects, practitioners must have a sufficient understanding of acupuncture [7,8]. Incompetent acupuncture skills can cause pain and hematoma and even lead to bent and broken needles [9]. Therefore, novice acupuncturists must follow standard training procedures such as acupoint location training and needling training [10,11].

While acupoint location is outside the patient’s body, needle manipulation is an invasive technique on the body surface, involving different parameters, e.g., needle insertion depth, needle tip direction, and needle movement to obtain deqi [12]. Thus, needling training is often conducted on phantom models to mimic the human body surface. These models are primarily based on silicon, gel, animal skin, apple, and cucumber [11]. Needling training is traditionally executed through an apprenticeship in which the students observe and try to replicate the practice of expert acupuncturists. Then, the performance of the novices is evaluated by the experts. Recently, several studies have developed technology-aided needing training systems with quantification and visualization of needle manipulation parameters. Results showed that training with such systems could improve the needle manipulation skills of novices [12-14]. However, no direct comparison with the traditional training method has been made.

Regarding measurement methods, research in acupuncture needling training has used various sensing technologies to measure the kinetic and kinematic parameters of needle manipulation [11,15]. Attaching sensors directly to the needle is the most widely used technique for real-time measurement of parameters. The most compact sensing device in this technique is Acusensor [16], which has been used in several needling training studies [12-14]. Another technique is using advanced motion tracking systems with physical markers attached to the needle or fingers [9,17-20]. However, these two techniques strongly influence the naturalness of finger movement and sensation while performing needle manipulation. Thus, noncontact measurements have much less influence. A popular noncontact technique is placing pressure sensors attached to the phantom mode; however, this can only measure kinetic parameters, i.e., the axial force and its derived parameters [9,21]. Some proposed using ultrasound images for needle motion tracking [22-25]. The training setup using this technique might not be portable and affordable for training on a large scale. Therefore, needle motion tracking without using physical markers in needling training is an urgent need, which is challenging given the small sizes of acupuncture needles compared with regular injection needles [26,27]. Drawbacks of this technique include measurement of kinematic parameters only and lower measurement accuracy [15]. However, it can be still applicable for certain needling training lessons considering the advantages of noncontact measurement and affordability.

In this study, we present a needling training system for acupuncture using a needle-tracking technique based on computer vision. Needle manipulation parameters were quantified by analyzing needle displacements from images captured by a camera. The system was validated in a needling training course with expert and novice acupuncturists in two studies. The selected training lesson aimed to improve the needle insertion speed within certain limits of needle insertion length and needle insertion angle (needle tip direction). In the first study, the ability to characterize the needle manipulation skills of the proposed system was investigated by comparing the needle manipulation parameters of experts and novices before and after the training. In the second study, the training effectiveness of displaying needle manipulation parameters during training (technology-aided training method) was compared with that of the traditional training method (control group). The technology satisfaction of students and teachers in the technology-aided training group was investigated.

MATERIALS AND METHODS

1. Methodology

The proposed vision-based needling training system was developed following a co-creation practice in which all stakeholders were involved in the entire system development [28]. Requirements were first established by expert acupuncturists who practice and train acupuncture skills daily. The targeted needle manipulation parameters are needle insertion speed, needle insertion length, and needle insertion angle (needle tip direction) (Fig. 1B). During the prototyping process, experts iteratively provided feedback to system developers. The final prototype was tested in a laboratory setting before the validation process in the actual needling training environment.

Figure 1. Vision-based needling training system. (A) The experimental setup includes a phantom model, a camera, a white LED light, and a dark background. Images captured by the camera are processed by a computer to determine needle manipulation parameters. (B) Needle manipulation parameters. (C) The needle handle is detected during needle manipulation and visualized by a red bounding rectangle.

2. System design

1) Setup

The vision-based needling training system is shown in Fig. 1A. The system includes a camera (TD-WC3200, The Deosd), a silicon-based phantom model (Henan Green Medical Tech Co., Ltd., Henan, China), a filiform acupuncture needle (haozhen; Suzhou Tianxie Acupuncture Instruments Co., Ltd., Suzhou, China), a white LED light (1.2 W), and a dark background (PMMA, 3 mm). Images captured by the camera were processed by a computer (Ubuntu 20.4; Canonical Ltd., London, UK) using the OpenCV library in Python [29]. The setup dimensions were 50 cm × 40 cm × 40 cm. The camera was installed at a height of 20.5 cm above the table. The filiform needle was 50 mm in length with the handle part of ϕ1 mm and body part of ϕ0.3 mm. The handle part of the needle was painted blue to create a contrast with the dark background and environment. The white LED light was used to improve the lighting condition if necessary.

2) Needle detection

Detecting the acupuncture needle and its movement is the core function of the system. Since the filiform needle is extremely thin, needle detection was performed through the blue-painted handle. Before capturing images, a camera calibration procedure was performed by applying Zhang’s method to remove distortion [30]. A chessboard was placed perpendicular to the phantom model to set the camera center. At this position, the size of the three squares was 170 × 170 pixels and was equivalent to 30 × 30 mm. The pixel-to-mm ratio was 0.177. Captured images (frames) were converted from RGB to HSV color space (with erosion and dilation) to detect the blue-painted needle handle. A threshold method was then applied with the minimum and maximum threshold parameters of HSV of (0, 0, 0) and (179, 255, 255), respectively. The detected needle handle was visualized by a red bounding rectangle (Fig. 1C). The width, height, minimum angle of rotation, and center point of the bounding rectangle can be estimated to determine needle manipulation parameters.

3) Needle manipulation parameters

Three needle manipulation parameters can be determined, i.e., needle insertion speed, needle insertion length, and needle insertion angle. The needle insertion movement was characterized by the starting (xS, yS) and ending (xE, yE) positions of the needle handle, corresponding to the start and end frames (Fig. 2). The time interval between two consecutive frames was 35 ms. The start position was updated when the needle was detected and if its offset position was 5 pixels higher than its previous position. The end position was the last frame when the needle stopped moving. The needle insertion length was determined by the distance between these two points.

Figure 2. Calculation of needle manipulation parameters based on captured frames. The time interval between two consecutive frames is 35 ms.

The needle insertion angle was determined by the minimum angle of needle rotation with respect to the horizontal axis at the end position (end frame). The needle insertion speed was calculated through the normalized cross-correlation method to improve the accuracy of calculating the center point displacement between two consecutive frames. This method was based on the similarity between the image and the feature [31]. The correlation coefficient Ci of the image (frame i) and the feature (frame i+1) in 35 ms was comparable to the center point displacement. The instant speed at frame i was Ci divided by 35 ms. The needle insertion speed was then calculated with the assumption that it followed the trapezoidal rule.

4) Laboratory test

The system was tested in a laboratory setting. A graphical user interface (GUI) was developed to set up and configure the system for practical usage. After each trial during training, the calculated needle manipulation parameters were displayed on a computer screen (Fig. 3A, B). Novice acupuncturists can instantly monitor their training performance trial by trial. Feedback from the expert acupuncturists confirmed that the developed GUI was applicable for certain needling training lessons. One of the potential training lessons was selected for validation in Sect. 3.

Figure 3. Laboratory test. (A) An acupuncturist performs needle manipulation skills on the system. (B) The needle manipulation parameters are displayed on the graphical user interface.

3. Validation

The proposed system was validated in a needling training course with expert and novice acupuncturists in two studies to investigate the ability to characterize needle manipulation skills and the effectiveness of the system on needling training. The training lesson was selected to improve the needle insertion speed while controlling the insertion length between 0.5 and 1 cm and the needle insertion angle between 80° and 90° for safety reasons. At these angle limits where the needle is nearly perpendicular to the skin surface, the insertion length and the insertion depth can be used interchangeably. This training is essential because a safe insertion with a high speed can reduce pain in body acupuncture [32,33]. This is even more important with filiform needles when several needles must be inserted (typically from 4 to 20) [34].

Acupuncture activities were performed on a silicon-based phantom model. Needles were not inserted in the human body. No human patients were involved. The study was recognized by the Faculty of Traditional Medicine, conducted in accordance with the ethical guidelines provided by Can Tho University of Medicine and Pharmacy (CTUMP), and pre-registered before data collection on OSF (osf.io/af2uz).

4. Participants

The study recruited 19 expert acupuncturists (6 females, 13 males) from CTUMP. The experts have on average 17.00 (standard deviation [SD], 6.30) years of experience. The novice acupuncturists were 54 students from two classes of the training course (40 females, 14 males; mean age, 22.12; SD, 1.14). They did not follow any acupuncture training before the needing training course. The first class followed traditional needling training through apprenticeship (control group, n = 27). The second class followed a technology-aided needling training using the proposed system (experimental group, n = 27). The students were guided by five teachers (two females, three males).

5. Procedure

1) Experts

All experts received an explanation of the study and provided informed consent. Then, they were asked to perform needle manipulation skills in five trials. After each trial, needle manipulation parameters were measured by the system and recorded by an experimenter.

2) Novices (students)

In the needling training course, all students received an explanation of the study and provided informed consent. The training course lasted for 2 days with five 50-min training sessions each day (8.33 h in total). On training day 1, both groups first received the same theory lecture about acupuncture techniques and familiarized themselves with the needle and phantom model. Before the training, they were asked to perform needle manipulation to determine the baselines (pretest). Then, each group practiced needling manipulation following the assigned training approach. After each day (sessions 5 and 10), needle manipulation skills were evaluated by five valid trials. After each valid trial, needle manipulation parameters were measured by the system and recorded by an experimenter. Finally, at the end of training day 2, students and teachers were asked to complete a technology satisfaction survey using Google Forms.

6. Measurement

Needle manipulation was measured through needle insertion speed (cm/s), needle insertion depth (cm), and needle insertion angle (degree). A trial was considered invalid (or unsafe) if the needle insertion depth and/or the needle insertion angle exceeded the corresponding minimum and maximum limits even if the insertion speed was high. Therefore, valid trials were recorded for equitable comparison purposes. In other words, a very fast insertion but outside the depth and angle limits (unsafe) cannot be used to compare with a slower speed insertion speed within the safety limits.

Within the safety limits, an optimal insertion should have an insertion depth between 0.7 to 0.8 cm and an insertion angle of exactly 90°. The depth “error” of a trial was defined as the distance between the actual depth to the optimal range.

Technology satisfaction was measured by a 5-point Likert scale survey using four constructs of the unified theory of acceptance and use of technology [35]. Perceived usefulness (PU), perceived ease of use (PEOU), and intrinsic motivation (IM) were used for all teachers and students. Additionally, the intention to use (ITU) construct was used for teachers. The wording of these constructs was adapted to the context of needling training.

7. Data analysis

Data were analyzed in Python using pingouin, scipy, and scikit posthocs packages. Normality and homogeneity were checked by the Shapiro–Wilk and Bartlett tests to determine the statistical tests. The significance level was set at 0.05.

RESULTS

1. Study I: characterization of needle manipulation skills

The needle manipulation skills of experts and novices before and after the training were compared (Fig. 4). A clear difference in the mean insertion speeds was found between experts and novices before the training (Fig. 4A). An independent t-test showed that the experts were significantly faster (M = 11.16; SD = 1.52) than the novices (M = 8.39; SD = 2.03; t(71) = 5.36; p < 0.001; d = 1.43). The novices improved their mean insertion speed after the training (48/54, 89%). A paired t-test showed that the mean insertion speed of the novices after the training was significantly improved (M = 10.53; SD = 1.85; t(53) = 7.43; p < 0.001; d = 1.09). An independent t-test showed that this speed was not significantly lower than those of the experts in study 1 (M = 11.16; SD = 1.52; t(71) = 1.31, n.s.).

Figure 4. Needle manipulation skills of experts and novices. (A) Experts had a higher needle insertion speed than novices before the training. Novices improved their speed after training. (B) No differences were found in the insertion depth error. (C) Statistical differences were found in the insertion angle but were practically negligible. Error bars show the standard deviation.

Similar comparisons were conducted with the needle insertion depth error and insertion angle. No differences were found in the needle insertion depth error (Fig. 4B). Statistically significant but practically negligible differences in the insertion angle were found before (M = 89.25; SD = 1.01) and after (M = 88.90; SD = 0.83; U = 321.5; p < 0.05) training of the experts (M = 88.52; SD = 1.44; W = 349.0; p < 0.05) (Fig. 4C).

2. Study II: training effectiveness

The data of each group indicated that both groups showed improvement after 10 training sessions (traditional: 23/27, 85%; technology-aided: 25/27, 93%). However, the improvement happened at different timescales (Fig. 5A).

Figure 5. Training results of the traditional training group (Trad.) and the technology-aided group (Tech.). (A) Both groups showed improvement after 10 training sessions (S10). (B) The technology-aided group had a better improvement after five training sessions (S5). No difference in improvement was found in both groups after 10 sessions (S10). Error bars show the 95% confidence interval.
1) Traditional training group

A Friedman test for repeated measures showed a significant difference in the mean insertion speed during training (χ2(2) = 19.23; p < 0.001; W = 0.36). A Nemenyi post-hoc test showed that the mean insertion speed was not significantly improved after five training sessions (S5, M = 9.75; SD = 1.79, n.s.) but only after 10 training sessions (S10, M = 11.12; SD = 1.54; p < 0.01) compared with the pretest value (pretest, M = 8.93; SD = 2.13). The improvement happened between sessions 5 and 10 (p < 0.01).

2) Technology-aided training group

A Friedman test for repeated measures showed a significant difference in the mean insertion speed during training (χ2(2) = 22.30; p < 0.001; W = 0.41). A Nemenyi post-hoc test showed that the mean insertion speed was already significantly improved after five training sessions (S5, M = 9.77; SD = 1.78; p < 0.01) and was maintained after 10 training sessions (S10, M = 9.94; SD = 1.95; p < 0.01) compared with the pretest value (pretest, M = 7.85; SD = 1.76). The improvement happened within the first five sessions (p < 0.01) and no further improvement within the later sessions.

3) Between-group comparison

Statistical tests on the gained mean scores were conducted because our data did not meet all assumptions for the analysis of covariance (Fig. 5B) [36]. After five training sessions, a Mann–Whitney U test showed that the technology-aided group obtained a significantly higher gain (M = 1.92; SD = 1.72) than the traditional group (M = 0.82; SD = 2.12; U = 236.0; p < 0.05; cles = 0.32). After 10 training sessions, a Mann–Whitney U test showed no significant difference between the technology-aided group (M = 2.09; SD = 2.29) and the traditional group (M = 2.19; SD = 1.88; U = 430.0, n.s.).

4) Technology satisfaction survey

Results from the self-reported technology satisfaction survey showed that students and teachers had positive attitudes toward the system. The technology-aided group rated the system useful (PU: M = 4.69, SD = 0.40) and easy to use (PEOU: M = 4.74, SD = 0.32) and felt motivated while using the system (IM: M = 4.84, SD = 0.29). The ratings from the teachers were also positive (PU: M = 4.43, SD = 0.36; PEOU: M = 4.50, SD = 0.42, IM: M = 4.47, SD = 0.45). They also showed a high ITU the system (ITU: M = 4.40, SD = 0.39).

DISCUSSION

Results from study I revealed a clear difference in the needle insertion speed between experts and novices before the training, which is in light with findings from previous studies on other kinematic and kinetic parameters [9,17,19,32]. In acupuncture, errors in depth and angle could cause more severe consequences than a low insertion speed. With little time for familiarization with the insertion movement before training, the novices focused on controlling the correct depth and angle. Therefore, it restrained them from getting insertion speed higher than the experts. This finding matches with feedback from the experts and teachers of the training course: (i) a novice needs intensive practice to obtain a higher insertion speed; (ii) controlling a correct depth and a correct angle requires a bit less effort.

After training, the novices could improve their insertion speed and approach the expert speed. This is consistent with findings from several previous studies with other parameters showing improvement after training [12-14]. However, this does not mean the novices reached the expert level in acupuncture skills in general but solely in this specific insertion speed training.

In study II, the technology-aided training approach resulted in a training effect after 10 training sessions similar to the traditional training approach. However, the technology-aided approach showed faster learning progress at the beginning of the training. The technology satisfaction survey indicated two possible explanations for this positive effect. First, both students and teachers considered the system useful and easy to use. Being informed of the needle manipulation parameters during training could potentially help students review their movements after each trial. Second, the students using the system reported high motivation. This can be explained by the novelty effect i.e., students felt excited about training with a new technological system [37]. Assuming that the performance after 10 training sessions is the maximal performance in the Thurstone learning curve in health professional education [38], the technology-aided training group reached their best performance faster. A fine-grained collection of data is required to validate this assumption [39].

CONCLUSION

We developed and evaluated a vision-based needling training system on a phantom model. The measurement method is noncontact, allowing the acupuncturists to perform needle manipulation skills naturally. The system can quantify three needle manipulation parameters, i.e., needle insertion length, needle insertion angle (needle tip direction), and needle insertion speed. The system was evaluated in a laboratory setting and a needling training course. Results showed that the system could characterize needle manipulation skills and resulted in faster learning progress in the first part of the training course than the traditional training method. More studies are needed to determine if this result can be replicated.

The system has several limitations that can be improved in future work. First, the system tracks motion by using computer vision and is therefore limited to kinematic parameters. Additional sensors will be integrated to quantify kinetic parameters for other training lessons. Second, the positive training effect could partially come from the novelty effect. Future studies should investigate the training effectiveness across multiple training lessons.

ACKNOWLEDGMENTS

The authors would like to thank Tran Minh Man and Le Hong Phuoc for their support in organizing the validation and the experts and students for their participation in this work.

AUTHOR CONTRIBUTIONS

Conceptualization: THL, HLC, DDP, LTCT, TV. Data curation: THL, LTCT, DDP. Data analysis: THL, HLC. Writing original draft: THL, HLC. Writing - review and editing: All authors.

CONFLICTS OF INTEREST

The authors have no conflicts of interest to declare.

FUNDING

The work leading to these results has received funding from Can Tho University of Medicine and Pharmacy and partially from Flanders Make and VLAIO under the project AUTOCRAFT-SBO.

ETHICAL STATEMENT

The authors declare that this research was conducted in full accordance with the Declaration of Helsinki. The study was recognized by the Faculty of Traditional Medicine, Can Tho University of Medicine and Pharmacy (329/QD-BM.YHCT) and followed the ethical guidelines provided by the organization. Needles were not inserted in a human body. No human patients were involved.

Fig 1.

Figure 1.Vision-based needling training system. (A) The experimental setup includes a phantom model, a camera, a white LED light, and a dark background. Images captured by the camera are processed by a computer to determine needle manipulation parameters. (B) Needle manipulation parameters. (C) The needle handle is detected during needle manipulation and visualized by a red bounding rectangle.
Journal of Acupuncture Research 2023; 40: 44-52https://doi.org/10.13045/jar.2022.00325

Fig 2.

Figure 2.Calculation of needle manipulation parameters based on captured frames. The time interval between two consecutive frames is 35 ms.
Journal of Acupuncture Research 2023; 40: 44-52https://doi.org/10.13045/jar.2022.00325

Fig 3.

Figure 3.Laboratory test. (A) An acupuncturist performs needle manipulation skills on the system. (B) The needle manipulation parameters are displayed on the graphical user interface.
Journal of Acupuncture Research 2023; 40: 44-52https://doi.org/10.13045/jar.2022.00325

Fig 4.

Figure 4.Needle manipulation skills of experts and novices. (A) Experts had a higher needle insertion speed than novices before the training. Novices improved their speed after training. (B) No differences were found in the insertion depth error. (C) Statistical differences were found in the insertion angle but were practically negligible. Error bars show the standard deviation.
Journal of Acupuncture Research 2023; 40: 44-52https://doi.org/10.13045/jar.2022.00325

Fig 5.

Figure 5.Training results of the traditional training group (Trad.) and the technology-aided group (Tech.). (A) Both groups showed improvement after 10 training sessions (S10). (B) The technology-aided group had a better improvement after five training sessions (S5). No difference in improvement was found in both groups after 10 sessions (S10). Error bars show the 95% confidence interval.
Journal of Acupuncture Research 2023; 40: 44-52https://doi.org/10.13045/jar.2022.00325

References

  1. Kaptchuk TJ. Acupuncture: theory, efficacy, and practice. Ann Intern Med 2002;136:374-383. doi: 10.7326/0003-4819-136-5-200203050-00010.
    Pubmed CrossRef
  2. White A, Ernst E. A brief history of acupuncture. Rheumatology (Oxford) 2004;43:662-663. doi: 10.1093/rheumatology/keg005.
    Pubmed CrossRef
  3. WHO. Acupuncture review and analysis of reports on controlled clinical trials. World Health Organization, 2002.
    CrossRef
  4. Liu T. Acupuncture: what underlies needle administration? Evid Based Complement Alternat Med 2009;6:185-193. doi: 10.1093/ecam/nen002.
    Pubmed KoreaMed CrossRef
  5. Nasir LS. Acupuncture. Prim Care 2002;29:393-405. doi: 10.1016/s0095-4543(01)00007-0.
    Pubmed KoreaMed CrossRef
  6. Ernst E. Acupuncture--a critical analysis. J Intern Med 2006;259:125-137. doi: 10.1111/j.1365-2796.2005.01584.x.
    Pubmed CrossRef
  7. Miao P, Zhang Q, Ji YY, Jiang J, Guo HD, Wang FB, et al. Quantitative assessment of expert's lifting-thrusting manipulation in acupuncture based on haptic device. Appl Mech Mater 2013;397-400:2507-2510. doi: 10.4028/www.scientific.net/AMM.397-400.2507.
    CrossRef
  8. Li J, Wu MX. [Ideas of standardization evaluation on acupuncture skills: enlightened by quantitative appraisal of surgical skills in Europe and North America]. Zhen Ci Yan Jiu 2011;36:449-452. Chinese.
    Pubmed
  9. Li J, Grierson LE, Wu MX, Breuer R, Carnahan H. Perceptual motor features of expert acupuncture lifting-thrusting skills. Acupunct Med 2013;31:172-177. doi: 10.1136/acupmed-2012-010265.
    Pubmed CrossRef
  10. Shin SG, Im IC. Satisfaction level of clinical practice and related variables for students in the department of radiology. J Korea Contents Assoc 2010;10:276-284. doi: 10.5392/JKCA.2010.10.9.276.
    CrossRef
  11. Jang JE, Lee YS, Jang WS, Sung WS, Kim EJ, Lee SD, et al. Trends in acupuncture training research: focus on practical phantom models. J Acupunct Res 2022;39:77-88. doi: 10.13045/jar.2022.00024.
    CrossRef
  12. Lee IS, Lee YS, Park HJ, Lee H, Chae Y. Evaluation of phantom-based education system for acupuncture manipulation. PLoS One 2015;10:e0117992. doi: 10.1371/journal.pone.0117992.
    Pubmed KoreaMed CrossRef
  13. Lim JW, Jung WM, Lee IS, Seo YJ, Ryu HS, Ryu YH, et al. Development of acupuncture manipulation education system. The Acupuncture 2014;31:11-19. doi: 10.13045/acupunct.2014049.
    CrossRef
  14. Jung WM, Lim J, Lee IS, Park HJ, Wallraven C, Chae Y. Sensorimotor learning of acupuncture needle manipulation using visual feedback. PLoS One 2015;10:e0139340. doi: 10.1371/journal.pone.0139340.
    Pubmed KoreaMed CrossRef
  15. Lyu R, Gao M, Yang H, Wen Z, Tang W. Stimulation parameters of manual acupuncture and their measurement. Evid Based Complement Alternat Med 2019;2019:1725936. doi: 10.1155/2019/1725936.
    Pubmed KoreaMed CrossRef
  16. Davis RT, Churchill DL, Badger GJ, Dunn J, Langevin HM. A new method for quantifying the needling component of acupuncture treatments. Acupunct Med 2012;30:113-119. doi: 10.1136/acupmed-2011-010111.
    Pubmed KoreaMed CrossRef
  17. Tang WC, Yang HY, Liu TY, Gao M, Xu G. Motion video-based quantitative analysis of the 'lifting-thrusting' method: a comparison between teachers and students of acupuncture. Acupunct Med 2018;36:21-28. doi: 10.1136/acupmed-2016-011348.
    Pubmed KoreaMed CrossRef
  18. Xu LL, Wang F, Yang HY, Tang WC. Three-dimensional finger motion tracking during needling: a solution for the kinematic analysis of acupuncture manipulation. J Vis Exp 2021;176:e62750. doi: 10.3791/62750.
    Pubmed CrossRef
  19. Xu LL, Xie J, Yang HY, Wang F, Tang WC. Operation stability analysis of basic acupuncture manipulation based on three-dimensional motion tracking data. Wirel Commun Mob Comput 2022;2022:1958984. doi: 10.1155/2022/1958984.
    CrossRef
  20. Zhang A, Yan XK, Liu AG. [An introduction to a newly-developed "Acupuncture Needle Manipulation Training-evaluation System" based on optical motion capture technique]. Zhen Ci Yan Jiu 2016;41:556-559. Chinese.
    Pubmed
  21. Liu TT, Wang ZH, Zhang G, Li H. Application research of neural network in acupuncture training system. Comput Technol Autom 2020;39:102-107.
  22. Leow MQ, Cao T, Lee SH, Cui SL, Tay SC, Ooi CC. Ultrasonography in acupuncture: potential uses for education and research. Acupunct Med 2016;34:320-322. doi: 10.1136/acupmed-2016-011182.
    Pubmed CrossRef
  23. Langevin HM, Konofagou EE, Badger GJ, Churchill DL, Fox JR, Ophir J, et al. Tissue displacements during acupuncture using ultrasound elastography techniques. Ultrasound Med Biol 2004;30:1173-1183. doi: 10.1016/j.ultrasmedbio.2004.07.010.
    Pubmed CrossRef
  24. Konofagou EE, Langevin HM. Using ultrasound to understand acupuncture. Acupuncture needle manipulation and its effect on connective tissue. IEEE Eng Med Biol Mag 2005;24:41-46. doi: 10.1109/memb.2005.1411347.
    Pubmed CrossRef
  25. Chen YL, Hou MC, Chang SC, Chuang KW, Lee PY, Huang CC. Development and evaluation of inexpensive ultrasound using A-mode and M-mode signals to identify lung depth and avoid risk of pneumothorax in acupuncture. J Med Biol Eng 2021;41:251-259. doi: 10.1007/s40846-020-00584-x.
    CrossRef
  26. Chou L, Hong JY, Hong C. A new technique for acupuncture therapy and its effectiveness in treating fibromyalgia syndrome: a case report. J Musculoskelet Pain 2008;16:193-198. doi: 10.1080/10582450802161978.
    CrossRef
  27. Chou LW, Hsieh YL, Kuan TS, Hong CZ. Needling therapy for myofascial pain: recommended technique with multiple rapid needle insertion. Biomedicine (Taipei) 2014;4:13. doi: 10.7603/s40681-014-0013-2.
    Pubmed KoreaMed CrossRef
  28. Sanders EBN, Stappers PJ. Co-creation and the new landscapes of design. CoDesign 2008;4:5-18. doi: 10.1080/15710880701875068.
    CrossRef
  29. Bradski G, Kaehler A. Learning OpenCV: computer vision with the OpenCV library. O'Reilly Media, 2008.
    CrossRef
  30. Zhang Z. Flexible camera calibration by viewing a plane from unknown orientations. IEE, pp 666-673, 1999.
    CrossRef
  31. Lewis JP. Fast normalized cross-correlation. In: Denis Laurendeau, ed. Vision Interface. Canadian Image Processing and Pattern Recognition Society, pp 120-123, 1995.
    CrossRef
  32. Ikai Y, Migaki M, Kida N, Iwamoto H, Hamada H. Analysis of the skills to acupuncture. In: Duffy V, ed. Digital Human Modeling. Applications in Health, Safety, Ergonomics and Risk Management: Human Modeling. Springer, pp 66-73, 2015.
    CrossRef
  33. Yin CS, Kim JH, Park HJ. High-velocity insertion of acupuncture needle is related to lower level of pain. J Altern Complement Med 2011;17:27-32. doi: 10.1089/acm.2010.0120.
    Pubmed CrossRef
  34. Lao L. Acupuncture techniques and devices. J Altern Complement Med 1996;2:23-25. doi: 10.1089/acm.1996.2.23.
    Pubmed CrossRef
  35. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information technology: toward a unified view. MIS Q 2003;27:425-478. doi: 10.2307/30036540.
    CrossRef
  36. Wright DB. Comparing groups in a before-after design: when t test and ANCOVA produce different results. Br J Educ Psychol 2006;76(Pt 3):663-675. doi: 10.1348/000709905X52210.
    Pubmed CrossRef
  37. Clark RE. Reconsidering research on learning from media. Rev Educ Res 1983;53:445-459. doi: 10.3102/00346543053004445.
    CrossRef
  38. Howard NM, Cook DA, Hatala R, Pusic MV. Learning curves in health professions education simulation research: a systematic review. Simul Healthc 2021;16:128-135. doi: 10.1097/SIH.0000000000000477.
    Pubmed CrossRef
  39. Pusic MV, Boutis K, Hatala R, Cook DA. Learning curves in health professions education. Acad Med 2015;90:1034-1042. doi: 10.1097/ACM.0000000000000681.
    Pubmed CrossRef
JAR
Feb 29, 2024 Vol.41 No.1, pp. 1~73

Stats or Metrics

Share this article on

  • line

Journal of Acupuncture Research

pISSN 2586-288X
eISSN 2586-2898
qr-code Download