Imagine a treatment for knee osteoarthritis (KOA) that combines traditional wisdom with cutting-edge technology—making personalized, effective care possible like never before. But here's where it gets controversial: Can artificial intelligence truly revolutionize acupuncture, a centuries-old practice rooted in centuries-old theories? And how might this new approach improve patient outcomes while overcoming the limitations of conventional methods?
Background
Knee osteoarthritis (KOA) is a long-lasting degenerative joint condition marked by the gradual breakdown of cartilage in the knee, which leads to persistent pain, stiffness, and difficulty moving.1 As we age, the chances of developing KOA rise sharply, affecting roughly 10% of men and 13% of women over 60.2 This disease doesn’t just impede physical activity; it diminishes quality of life, often results in increased healthcare costs, and adds a hefty economic burden on society.
Traditional Chinese Medicine (TCM) has practiced acupuncture for thousands of years as a way to relieve pain and restore bodily balance.3–6 This involves inserting very thin needles into specific points called acupoints, believed to stimulate healing and energy flow.7 Studies have shown that acupuncture can effectively reduce pain and improve joint function, possibly through mechanisms such as endogenous opioid release, reduction of inflammation,10 and better blood flow in the affected area. Despite promising results, a major hurdle persists: variability in how acupoints are chosen, which limits uniformity and widespread acceptance.
Typically, acupuncturists select points based on their experience and TCM theory.11–12 This approach, however, leads to differences from one practitioner to another, and lacks standardization, making it harder to scientifically validate effects.13 Efforts to categorize acupoint choices in KOA usually involve local points (like ST36 or ST34), distal points (such as LI11), or combinations of both.14–17 Yet, these methods often rely heavily on empirical knowledge rather than solid scientific frameworks, which hinders broader applicability.
According to TCM, acupoints linked to specific conditions can become 'sensitized'—showing signs like warmth, redness, swelling, or tenderness, which indicate underlying issues.19,20 Our research team previously used infrared thermography to measure skin temperature patterns in KOA patients, revealing correlations between thermal changes, clinical symptoms, and inflammatory markers.21 These findings suggested a pathway for individualizing acupuncture by targeting thermally sensitive acupoints.
The Promise of New Technologies
Recently, advances in artificial intelligence (AI) have opened new doors for personalized medicine. By integrating AI algorithms with infrared imaging, clinicians can now objectively analyze thermal patterns to select the most appropriate acupoints for each patient—reducing subjective bias and variability. Portable infrared devices can quickly generate tailored acupoint prescriptions based on thermal signatures, promising more consistent and effective treatment outcomes.22
This study aims to compare AI-assisted, infrared imaging-based acupuncture point selection with traditional experience-based methods for KOA management. Our goal is to establish a standardized, reproducible approach that maximizes therapeutic benefits and broadens clinical application.
Study Design and Methodology
This is a multi-center, randomized, placebo-controlled clinical trial, involving four hospitals: Shanghai Yangzhi Rehabilitation Hospital, Shuguang Hospital affiliated with Shanghai University of Traditional Chinese Medicine, Ruijin Hospital’s Hainan branch, and Zhongshan Hospital of Traditional Chinese Medicine. The study has been registered with the Chinese Clinical Trial Registry (ChiCTR2400087106). A flow diagram (Figure 1) illustrates the process, and Table 1 provides a detailed schedule for enrollment, intervention, and assessments.
Participants will be recruited via outpatient clinics, community outreach, and online advertisements. Interested individuals will contact the research team or be followed up through screening records to determine eligibility.
Inclusion criteria involve being aged 50 or older, having persistent knee pain (>3 months) with a pain score of at least 4/10, radiologically confirmed KOA (Kellgren-Lawrence grade II or III), and meeting certain diagnostic standards. Exclusion criteria comprise conditions like systemic arthritis, recent knee surgery, other lower limb disorders, recent knee injections, blood clotting issues, recent acupuncture, or other health problems that could interfere with participation.
Patients who meet these criteria and consent to participate will be randomly assigned to one of three groups: AI-assisted personalized acupuncture, conventional acupuncture, or sham (placebo) acupuncture, with allocation concealed via opaque envelopes. Blinding is maintained for participants, outcome assessors, and data analysts, although acupuncturists will be aware of treatment assignments to ensure proper execution.
Sample Size and Statistical Power
Based on previous data, detecting meaningful improvements in pain and function requires around 32 participants per group. To account for possible dropouts (~20%), we will enroll 120 participants (40 per group). This ensures our study has enough statistical power to confidently assess whether AI-assisted acupuncture is superior.
Interventions involve three groups:
- The AI-assisted group uses infrared thermal images analyzed by an AI system to determine ideal acupoints.
- The conventional group relies on traditional expert knowledge and anatomical standards.
- The sham group receives superficial needling at non-acupuncture points without stimulation.
All patients will undergo treatment twice weekly for 8 weeks, with each session lasting 30 minutes. Standardized equipment, training, and protocols across all centers aim to ensure consistency.
AI-based Acupoint Selection involves analyzing skin temperature across five knee regions (middle, upper inner/outer, lower inner/outer) to identify thermally sensitive sites. The AI system then maps these areas to relevant meridians and prescribes appropriate acupoints based on traditional and clinical principles. Needles are inserted vertically to depths of 25–40 mm, with stimulation until Deqi—the characteristic sensation—appears.
Outcome Evaluations include primary measures like knee function (via WOMAC scores) and pain severity (using Numeric Rating Scale). Secondary outcomes include stiffness, quality of life (SF-12), knee Joint Range of Motion (ROM), and traditional Chinese medicine symptom scores. Additionally, inflammatory markers such as IL-1β, IL-6, and TNF-α are monitored to explore biological effects.
Data Analysis will follow an intention-to-treat approach, comparing change scores across groups using advanced statistical models, adjusting for baseline differences, and correcting for multiple comparisons. Subgroup analyses will explore factors like age, severity, and prior acupuncture experience, shedding light on who benefits most.
Discussion
Our innovative trial bridges the gap between ancient Chinese healing arts and modern AI technology, aiming to improve treatment consistency, efficacy, and scientific validation. The inclusion of sham controls and rigorous methodology seeks to produce reliable results, potentially transforming acupuncture from an art rooted in subjective expertise to a science grounded in evidence. We believe that this strategy could encourage wider acceptance and integration of acupuncture in mainstream medicine.
But here’s the interesting part—can an AI system truly capture the complexity of traditional diagnostics, or does it oversimplify a nuanced practice? And will this fusion of tech and tradition indeed lead to better patient results, or could it diminish the human element that often makes acupuncture effective? Share your thoughts in the comments—do you see this as a breakthrough or a step too far?
