We are translating millennia of TCM wisdom with a mathematical language called "tensor–hypergraph".
The modernisation and internationalisation of traditional Chinese medicine has long faced a core paradox: it holds massive real-world data, yet because "pattern differentiation and treatment" is highly individualised and holistic, it resists quantification and verification by modern data science. As the national "Data Element ×" strategy names TCM data a high-value data resource, making this data both "computable" and "trustworthy" has become a defining research problem.
A joint team from Guangdong University of Technology, Hong Kong Baptist University, The Education University of Hong Kong and the Guangdong Digital Industry Research Institute offers their answer — HoloTSH (Holographic Tensor-Spectrum Hypergraph framework). It is both a technical breakthrough and a paradigm shift: giving every clinical TCM record a "digital identity" so that data flows are compliant, trustworthy and traceable, directly serving the national data strategy.
Millennia of pattern-differentiation experience are essentially a precision "black-box system". A master's "intuition" is the emergent product of decades of clinical practice. But when that intuition must become inheritable, auditable digital assets, three obstacles appear:
Data silos: TCM clinical data is scattered across hospitals, research institutes and pharma companies with no unified standard.
Trust deficit: classic "rule engines" are rigid; modern "deep learning" is a black box — the former cannot adapt to dynamic TCM reasoning, the latter fails compliance audit.
Circulation: the state asks data to be "supplied, flowing and well used", yet TCM data — privacy-sensitive and logically complex — has struggled to enter the data-element market.
HoloTSH's core innovation is two mathematical tools acting as "translator" and "auditor" for TCM data.
1. Tensor decomposition: mathematical coordinates for "patterns".
The team maps TCM "patterns" (e.g. "liver constraint with spleen deficiency") to points in a high-dimensional tensor manifold. Through tensor decomposition, the once-fuzzy "intuition" becomes a computable, measurable mathematical object — for the first time, schools of TCM from different regions share a unified "mathematical grammar".
2. Hypergraph audit: a "black box recorder" for data flows.
Tensors express; hypergraphs supervise. HoloTSH tracks every flow of data through hypergraph structure, so that any call, modification or sharing of pattern data leaves tamper-proof topological traces — a flight recorder and electronic ID card for TCM data.
Case 1: an accelerator for drug discovery. Working with a top-tier TCM hospital, HoloTSH consolidated scattered clinical data into a computable "pattern manifold", precisely identifying the "pattern drift" of a classic formula in modern populations — a key target for new-drug development, compressing years of pattern mining into weeks.
Case 2: a passport for cross-border regulation. For regulatory mutual recognition — the biggest hurdle of TCM internationalisation — HoloTSH's "manifold audit reports" describe a formula's logic in pure mathematics, giving cross-border regulators objective evidence beyond culture and language.
The team plans to extend HoloTSH to herb traceability and integrated clinical-research platforms, so that every herb and every formula carries a verifiable "data passport". When TCM's "intuition" is translated into mathematics, it is not diluted — it gains a passport to the world.
As the programme lead puts it: "We are not remaking TCM with Western medicine; we are respecting TCM with mathematics. HoloTSH's mission is to let millennia of wisdom flourish in the digital era."
HoloTSH means far more than technology. It proves that true modernisation of traditional medicine is not shallow "integration of Chinese and Western medicine" but a rigorous bridge of modern mathematics beneath traditional wisdom. Under the national "Data Element ×" strategy, HoloTSH offers a Chinese solution for TCM data-element circulation — and a paradigm for the digital transformation of traditional medicine worldwide.
This article answers a question the public often asks: can AI learn a TCM master's intuition? The answer is neither "yes" nor "no" — the problem must first be translated into mathematics: patterns as tensor manifolds, experience as computable objects. Only then can AI learn and be audited. It also shows that "AI safety" is broader than defence against hackers: making AI's judgments explainable, traceable, auditable is part of safety too — our third lock, "trustworthiness".
Note the "hypergraph audit" design: every data flow leaves tamper-proof topological traces. For general readers this is exactly how "data-element circulation" and "personal-privacy protection" coexist — data must be usable and leave a trail. For the technical detail, see the HoloTSH preprint on our publications page (TechRxiv, DOI 10.36227/techrxiv.177130202.27796764).