How Medi-Chi Is Bridging the AI Gap in Acupuncture: A Response to the 2025 SAR Roadmap

In late 2025 the Journal of Integrative and Complementary Medicine published a Society for Acupuncture Research roadmap titled "Artificial Intelligence and Digital Health in Acupuncture Practice" [1]. It surveyed 46 practitioners across multiple countries, ran SWOT analyses for clinicians, researchers and patients, and concluded with six concrete recommendations for closing what the authors called the AI gap in acupuncture.

The gap is real. Most acupuncture is still documented as free-text in incompatible systems. Across the SAR survey, only 52.2% of practitioners reported recording point names in a structured way and just 45.7% captured electrical stimulation parameters; items like rationale for point selection, anatomical region, laterality and Chinese-character codes were each documented by only a single respondent [1]. Pattern differentiation rarely makes it into structured fields at all. That makes it nearly impossible to feed acupuncture data into modern machine-learning pipelines — or even to compare two clinics across the street from each other.

Medi-Chi was built, in large part, to close that gap. This article walks through each of the six SAR recommendations and shows where Medi-Chi already meets them, and where work is still ongoing.

Recommendation 1 — Interoperable core data standards

The roadmap is explicit that "data are the foundational material for AI systems to learn and provide informed recommendations" and that without interoperable core data standards acupuncture records cannot be meaningfully combined for personalised, AI-supported care [1]. HL7 FHIR is the dominant open standard for health-record exchange [2], and the WHO's ICD-11 Chapter 26 now codifies Traditional Medicine pattern diagnoses in a globally interoperable form [3].

How Medi-Chi addresses it. Every diagnosis generated by Medi-Chi is stored against a structured schema: the patient's symptoms map to discrete TCM patterns, the patterns map to organ systems, and the organ systems map to point prescriptions across fourteen named acupuncture styles. Western diagnoses are cross-referenced through the Western Symptoms Guide so the same case can be looked at from either side. The internal data model is FHIR-compatible at the field level, and we are working toward exporting ICD-11 TM2 codes for pattern diagnoses so that research collaborators can ingest Medi-Chi exports directly.

Recommendation 2 — Use structured data templates

The SAR survey found that only 52.2% of respondents recorded point names in a structured way and only 45.7% recorded stimulation parameters, with open-text entry the dominant pattern across every other documentation category surveyed [1]. The authors note this matters because "open-access-text data entry can introduce biases and impact the accuracy, reliability, and validity of data analysis and decision-making processes" and explicitly recommend templates aligned with EHR standards [1]. Outcome capture is similarly fragmented: 56.5% of respondents use numeric rating scales, 41.3% use visual analog scales and 41.3% rely on verbal expressions, with only 28.3% using validated PROMs [1].

How Medi-Chi addresses it. The Medi-Chi Symptom Questionnaire is twelve structured steps. Pulse, tongue, sleep, digestion, emotional state and pain location are all captured as discrete fields. The diagnosis engine returns a structured prescription that includes:

• the pattern differentiation with classical name and pinyin
• the acupuncture style used (Five Element, Stems & Branches, Saam, Master Tung, Dr. Tan balance method, Esoteric, and others)
• the point list with point code, pinyin, Chinese characters, laterality, depth, technique (tonification/even/reduction) and needling cautions
• any electrical stimulation parameters, retention time and adjunct modalities such as moxa, cupping or gua sha
• the clinical reasoning that links symptoms to pattern to points

These map directly onto Table 3 of the SAR paper — safety and clean-needle procedures, needle specifications and technique, timing logs, adjunct therapies, patient responses and clinical triggers [1]. The Clinical Handout System then renders that structured record as a patient-facing PDF for record-keeping continuity.

Recommendation 3 — Increase digital literacy on AI

The roadmap argues that clinicians, students and patients all need plain-language education about how AI in healthcare actually works — what it can do, what it cannot, where it fails, and how to keep it accountable, naming transparency and liability as the headline risks practitioners need to understand before adopting AI tools [1]. The SWOT analysis (Table 6) lists "limited digital literacy among health care providers" and "suboptimal or poorly designed user interfaces" among the core weaknesses currently holding the field back [1].

How Medi-Chi addresses it. The Medi-Chi blog is the practical layer of this education. Recent articles cover the AI TCM Diagnosis Tool, Pattern Differentiation Software and the Acupuncture Treatment Plan Generator — each written for working clinicians rather than computer scientists. Every diagnosis page now carries an Evidence and Sources section with strength-of-evidence tags (guideline-endorsed, mixed-evidence, limited-evidence, traditional-use) so practitioners and patients can see at a glance how strong the supporting research is for the recommendation they are looking at. That kind of transparency is what the SAR paper means by digital literacy in practice.

Recommendation 4 — Strengthen interdisciplinary collaboration

The SAR authors argue AI in acupuncture only works when "experts from various fields, for example, medicine (both biomedicine and Traditional East Asian Medicine), computer science, ethics, and policymaking" are all in the room, and frame interdisciplinary collaboration as essential rather than optional given the complexity of AI technologies [1].

How Medi-Chi addresses it. Medi-Chi is built by a registered Chinese medicine practitioner (Derek Doran, AHPRA CMR0002211465) working with software, AI and design specialists, and the product itself is explicitly bilingual at the conceptual level — Western diagnoses on one side, classical pattern differentiation on the other, with an audit trail showing how one was translated into the other. The Modern Nèi Jīng project takes the same translational stance for the classical literature: classical Chinese characters preserved, modern English explanation alongside, and clinical relevance flagged at the chapter level. That cross-disciplinary practice is encoded into the platform rather than bolted onto it.

Recommendation 5 — Privacy and ethical aspects are important

Patient acupuncture records contain sensitive health information. The SAR SWOT analysis flags "concerns regarding data privacy and ethical implications" and "loss of individual control over how health information is used or shared" as threats common to clinicians, researchers and patients alike, and the roadmap warns that AI tools must respect the same privacy obligations as any other electronic health record [1].

How Medi-Chi addresses it. Medi-Chi is designed against the Australian Privacy Principles under the Privacy Act 1988 [5] and the AHPRA / Chinese Medicine Board of Australia codes governing record-keeping and scope of practice [4]. Concretely:

• The backend uses row-level security so each practitioner can only read and write their own records.
• Patient identifying data is kept in the practitioner's authenticated account, separated from the inference pipeline.
• The diagnosis engine operates on structured symptom inputs rather than free-text identifiers, which limits the surface area for accidental disclosure.
• Every clinical output carries a visible disclaimer that Medi-Chi is adjunct decision support, not a replacement for a registered practitioner, and that serious or undiagnosed conditions belong with a medical doctor.
• The platform is built to be AHPRA and TGA compliant: no testimonials, no outcome guarantees, no restricted-herb recommendations.

Recommendation 6 — Strengthen research infrastructure

Finally, the roadmap calls for an end to data left on the floor. The researcher-facing SWOT highlights opportunities for "expansion of research scope beyond traditional randomized controlled trials", "integration of diverse and multimodal data sources" and "enhanced scalability and reach of remote or decentralized clinical trials" — all of which depend on having structured, interoperable acupuncture data to draw on in the first place [1].

How Medi-Chi addresses it. Because Medi-Chi captures symptoms, pattern differentiation, treatment style and point selection in structured fields, properly de-identified and ethics-approved exports can be made available for research collaborations. The Research Database and Tongue Diagnosis modules are designed with this in mind: every tongue image contributed by a consenting practitioner is tagged with its corresponding pattern differentiation, which over time produces exactly the kind of multimodal, labelled dataset the SAR paper says the field needs.

What is still ahead

The roadmap is a roadmap, not a finish line. There are places where Medi-Chi still has work to do — full FHIR API parity, end-to-end ICD-11 TM2 coding on every prescription, formal pilot studies with university partners, and a published data dictionary so external researchers can map their own exports onto ours. Those are on the build queue, and the Europe 2026 roadmap sets the timeline.

The point is not that Medi-Chi has solved the AI gap in acupuncture. The point is that the gap is solvable, and that a small, focused, clinician-led platform can implement most of the SAR roadmap inside two years — provided it is built from day one around structured data, transparent evidence, and the regulatory framework practitioners already work within.

Try it

• Generate a structured diagnosis — twelve-step questionnaire, fourteen acupuncture styles, evidence-tagged output
• Acupuncture Points Reference — every point with pinyin, characters, depth, technique and cautions
• Conditions Index — browse by strength-of-evidence tag
• Pattern Differentiation — structured TCM pattern engine
• Modern Nèi Jīng — classical foundations in plain English

Footnotes

1. Witt CM, Murphy C, Graca S, Angell C, Gale MK, et al. Artificial Intelligence in Acupuncture: Recommendations from the Society for Acupuncture Research Special Interest Group Artificial Intelligence and Digital Health. Journal of Integrative and Complementary Medicine (2025). doi:10.1177/27683605251389440 — survey of 46 practitioners, SWOT analyses, and six recommendations cited throughout this article (Tables 2, 3, 5 and 6; Recommendations 1–6).

2. HL7 International. FHIR Overview (2024) — the open standard for interoperable health-record exchange referenced in Recommendation 1.

3. World Health Organization. ICD-11 Chapter 26 — Traditional Medicine Conditions (Module 1) (2024) — globally interoperable coding of TCM pattern diagnoses.

4. AHPRA / Chinese Medicine Board of Australia. Codes and Guidelines (2024) — record-keeping, advertising and scope-of-practice obligations Medi-Chi is built to satisfy.

5. Office of the Australian Information Commissioner. Australian Privacy Principles (Privacy Act 1988, 2024) — the privacy framework Medi-Chi's data handling is designed against.