Traditional Chinese Medicine (TCM) is a form of complementary medicine that was first developed over two thousand years ago. The TCM concepts, such as ‘Yin’(阴) and ‘Yang’(阳), that are used to describe groups of conditions have no direct translation in modern medicine, resulting in a gap between TCM and modern medicine that needs to be addressed via evidence-based research.
In the broad sense, ‘Yin’ and ‘Yang’ are two ends of a spectrum. They can be used on a wide variety of levels, such as personality (introverts as ‘Yin’ and extroverts as ‘Yang’) and gender (females as ‘Yin’ and males as ‘Yang’). A balance between ‘Yin’ and ‘Yang’ is one of the crucial definitions of a healthy individual, and an imbalance can lead to diseases. TCM clinicians treat patients by correcting the ‘Yin’ and ‘Yang’ imbalance by replenishing ‘Yin’ or ‘Yang’ and diminishing ‘Yin’ or ‘Yang’. 1
TCM robots
AI-powered TCM robots emulate the traditional four-examination method used by human physicians. By analysing facial and tongue characteristics, checking wrist pulses, and asking patients questions, these robots can distinguish various health imbalances and physical conditions, offering personalised health improvement advice accordingly.
Photo: Imago / Alamy Stock Photo
TCM views a person’s health as a dynamic balance of the different syndromes, including ‘Yin’ and ‘Yang’. TCM clinicians first identify the symptoms present in the patient using the four-examination method (望闻问切). From the information collected, they deduce the disease and syndrome of the patient, which they then use to devise a treatment plan for the patient.
The four-examination method consists of inspection (望), auscultation and olfaction (闻), inquiry (问), and palpation (切). Inspection is to observe the patient using sight, including the patient’s physique, range of movement, and tongue colour. Auscultation and olfaction are conducted to detect any abnormalities by listening or smelling. Wheezing, increased bowel sounds, and bad breath are abnormalities recorded through auscultation and olfaction. Inquiry is to ask the patient about the progression of the disease from the start to the present, the medical history, and other relevant information. Palpation is to use the sense of touch to feel any irregularity. It is most commonly used to observe the pulse at the wrist. The information gathered from each way of the four-examination method contributes to the accuracy of the diagnosis.2
TCM AND AI
TCM has accumulated a wealth of knowledge over the course of its development, and it is humanly impossible to learn from all the past experiences of earlier TCM clinicians. With AI, it is possible to analyse all the past TCM records to generalise the knowledge by discovering patterns and methods used for diagnosis and treatment. Machine learning models can also be built to mimic the diagnosis and treatment procedure of an experienced TCM clinician.
TCM clinicians use the four-examination method to gather information about the patient, which can result in subjectivity. To give an example of subjectivity in TCM practice, when TCM clinicians identify the symptoms present in the patient, they use their senses to observe the colour of the tongue and face. Colour is a subjective feature, and TCM clinicians may perceive the same colour differently based on their observation and experience. AI can come into play to analyse and diagnose the patient objectively and accurately. AI can use data obtained from images and sensors to quantify the symptoms, such as the redness of the tongue or face, and determine the symptoms in the patient more objectively.
A TCM powerhouse
Launched in January 2024, China’s Shuzhi Qihuang (数智岐黄) is a leading Large Language Model in TCM. Trained on a vast library of over 1,000 classical TCM texts and ancient literature, it utilises advanced AI techniques and offers functionalities
such as prescription recommendations, herbal property interpretation and syndrome diagnosis. The model’s capabilities were further validated by surpassing GPT-4 in a simulated TCM practitioner qualification exam.
Photo: John H / iStock
CURRENT AREAS OF AI APPLICATION IN TCM
AI has been used to create clinical decision support systems (CDSS) to support TCM clinicians. There are many different types of CDSS, each trained to perform one or more specific tasks in diagnosis or treatment. The CDSS can help identify patients’ symptoms, diagnose them using their medical records, and propose optimal treatment plans.3
The CDSS has often been used to identify facial, tongue, and wrist pulse symptoms obtained from inspection and palpation. The facial and tongue information is captured in images, while the wrist pulse symptoms are measured in pulse waves. The CDSS is able to analyse such data and detect the respective symptoms. The TCM clinician enters other information from the inquisition, auscultation and olfaction in the four-examination method. After compiling all the relevant information, the CDSS diagnoses the patient and proposes a treatment plan accordingly. TCM clinicians have tested this tool and have received it well. However, there is still work to be done to improve the accuracy and explainability of its results.4 5
AI has also been used to discover drugs in TCM. Network pharmacology is used to study the interactions between drugs, diseases, and biological targets. In TCM, network pharmacology builds a network of interactions between TCM herbs and diseases and their related target proteins. Stronger interactions between the active components in the TCM herb and disease targets indicate higher therapeutic efficacy. AI is able to utilise information from the network and deduce active components in herbs that may be potential drug candidates to treat the specified disease. 6
AI can also be used for treatments. Currently, AIdriven acupuncture and ‘Tuina’ robots are being developed to treat patients based on a treatment plan provided by a TCM clinician. These robots can identify the acupoints on a patient’s body and perform the required treatment accordingly. Since such treatments are traditionally time-consuming and labour-intensive, TCM clinicians can diagnose and treat more patients with AI machines. This will make TCM more accessible and affordable to more patients.
AI-driven acupuncture and ‘Tuina’ robots are being developed to treat patients based on a treatment plan provided by a TCM clinician.
In Singapore, the AI-powered ‘Tuina’ robot EMMA developed by AiTreat is planning randomised controlled trials (RCT) to test the efficacy of the ‘Tuina’ robot compared to manual ‘Tuina’. The EMMA robot uses sensors and 3D vision to detect muscle stiffness and proposes an appropriate ‘Tuina’ plan. After inputting specifications and gaining confirmation from the TCM clinician, EMMA can perform ‘Tuina’ on the patient using its robotic arm.7 8
Another research team from the National Cheng Kung University in Taiwan built a robotcontrolled acupuncture machine that is able to locate the acupoints and stimulate the acupoints accurately, but it is still difficult to detect the ‘deqi’ (得气) sensation, which is a criterion for the acupuncture to take effect.9
EMMA, the robotic ‘Tuina’ therapist
EMMA, or “Expert Manipulative Massage Automation,” is a robotic ‘Tuina’ device
designed to reduce therapists’ workload and improve treatment quality. Equipped with AI and three-dimensional visioning technology to locate the relevant acupoints on a patient’s body, it allows a qualified TCM practitioner to use its touchscreen to prescribe and finetune a treatment plan based on TCM theories and the patient’s syndromes. EMMA then performs the customised ‘Tuina’ procedures with high precision and measures the effectiveness with its sensors.
Source: AiTreat
CHALLENGES
Although the application of AI in TCM has its benefits, AI faces some challenges due to its inherent nature. For example, traditional clinical TCM data are not ideal for AI learning. AI requires a systematic approach to data collection, but there is no standardised method in TCM record keeping. Thus, when TCM records between different clinics and hospitals are integrated, incompatible and missing information would weaken the effectiveness of AI.
As AI needs to be trained on a large amount of data, proper data protections must be implemented to mitigate the risks of a data breach, as data mishandling can compromise patients’ privacy and confidentiality. Informed consent must also be obtained from the participating patients to avoid data infringements.
Since AI is still susceptible to errors, TCM clinicians must be aware of the risks associated with using AI in clinical practice and be equipped to deal with any issues that may arise. If an error that causes harm to a patient occurs, clear regulations and guidelines are required to resolve the legal conflicts between the parties involved.
AI-driven acupuncture and ‘Tuina’ robots are being developed to treat patients based on a treatment plan provided by a TCM clinician.
Another challenge of using AI in TCM is its inability to process human emotions and provide empathetic patient care. Considerations such as economic status, a patient’s response to treatment, and a patient’s state of mind should also contribute to the decision-making of TCM treatments. Expensive herbs can be replaced with cheaper ones with similar effects for low-income patients. A patient who fears needles can be treated with herbs instead of acupuncture. A terminally ill patient may prefer a treatment that mitigates the symptoms to one that causes undesirable side effects. Thus, TCM clinicians must intervene and make the final decision that is best suited for the patient.10 11
Although there are efforts to mitigate the risks of using AI, some may still harbour doubts and mistrust towards AI for good reasons. As a researcher using AI, I know that AI may only be dependable when it has been rigorously tested and proven. However, the main objective of integrating AI into TCM is not for AI to independently function as a TCM clinician but to support TCM clinicians in their work with proper supervision
CONCLUSION
AI has a wide range of applications in TCM and has shown potential to advance the field of TCM in clinical practice and research. On a larger scale, AI is already playing significant roles in healthcare, from abnormality detection in chest X-rays to the detection of diabetic retinopathy. Quoting Professor Kenneth Mak, the Director-General of Health at the Ministry of Health, “Most people are not aware that AI is already part of our normal conversations because a lot of the applications are behind the scenes, under the surface; but already (it) has a big and significant influence in how we deliver care at the population level, not necessarily to the individual level”. In Singapore, AI has already been implemented in healthcare.12 Thus, AI integration into TCM should follow suit to improve healthcare quality with effective and reliable TCM options. ∞
KON WEN XUAN
Kon Wen Xuan is a PhD student in the Nanyang Technological University (NTU) Graduate College, under the Interdisciplinary Graduate Programme (IGP AI-X). He obtained his Double Degree in Bachelor of Science in Biomedical Sciences and Bachelor of Chinese Medicine at NTU. During his course of study, Wen Xuan studied TCM theories from professors in Beijing University of Chinese Medicine (BUCM) and built machine learning models to predict TCM clinicians’ diagnosis based on the patient records in the NTU TCM clinic. Currently, he is continuing his research on AI and TCM.
JULY 2024 | ISSUE 12
NAVIGATING THE AI TERRAIN
- “Yin/Yang Theory — TCM World.” TCM World, Feb 2024. https://www.tcmworld.org/what-is-tcm/yin-yang-theory
- Sui, Dong, et al. “Data-Driven Based Four Examinations in TCM: A Survey.” Digital Chinese Medicine, vol 5, no 4, Dec 2022, pp 377–385. <br>https://doi.org/10.1016/j.dcmed.2022.12.004.
- Wang, Yulin, et al. “The Impact of Artificial Intelligence On Traditional Chinese Medicine.” The American Journal of Chinese Medicine, vol 49, no 06, Jan 2021, pp 1297–1314. https://doi.org/10.1142/ s0192415x21500622.
- Tao, Liyuan, et al. “Accuracy and Effects of Clinical Decision Support Systems Integrated With BMJ Best Practice-Aided Diagnosis: Interrupted Time Series Study.” JMIR Medical Informatics, vol 8, no 1, Jan 2020, e16912. https://doi.org/10.2196/16912.
- Shen, Ying, et al. “Decision Support System for Acupuncture Treatment of Ischemic Stroke.” Advanced Data Mining and Applications Lecture Notes in Computer Science, 2020, pp 582-594. Springer International Publishing. https://doi.org/10.1007/978-3- 030-65390-3_44.
- Lin, Yumeng, et al. “Computers, Especially AI-Assisted Drug Virtual Screening And Design In Traditional Chinese Medicine.” Phytomedicine, vol 107, Dec 2022, 154481. https://doi.org/10.1016/j. phymed.2022.154481.
- “AI-Powered Massage Robot To Start Trials in United States and Singapore.” Nanyang Technological University Singapore, 13 Feb 2023. https://www.ntu.edu.sg/news/detail/ai-powered-massagerobot-to-start-trials-in-united-states-and-singapore
- “A New Option: Tui Na Massage by a Robot Masseuse.” Gleneagles Hospital Singapore, 10 Jun 2021. https://www.gleneagles.com.sg/ health-plus/article/tui-na-chinese-robot-massage
- Lan, Kun-Chan, and Gerhard Litscher. “Robot-Controlled Acupuncture — An Innovative Step Towards Modernisation of The Ancient Traditional Medical Treatment Method.” Medicines (Basel), vol 6, no 3, 10 Aug 2019, pp 87. https://doi.org/10.3390/ medicines6030087.
- Wang, Yulin, et al. “The Impact of Artificial Intelligence On Traditional Chinese Medicine.” The American Journal of Chinese Medicine, vol 49, no 06, Jan 2021, pp 1297–1314. https://doi.org/10.1142/ s0192415x21500622.
- Adams, Jon, et al. Traditional, Complementary and Integrative Medicine: An International Reader. Bloomsbury Publishing, 2017.
- “AI Already Playing Bigger Roles Behind The Scenes In Healthcare: Kenneth Mak.” The Straits Times, 6 Aug 2023. https://www. straitstimes.com/singapore/health/ai-already-playing-bigger-rolesbehind-the-scenes-in-healthcare-kenneth-mak.
