藏医六点位置特异模型的单点判别力跨脏器异质性检验:哪些点位的判别力显著优于随机基准

Written by

in

Language

Evidence level: C

把藏医「腕上六点对应脏腑」的古老指派,翻译成一张可一次性裁决的判别力矩阵:行是点位,列是脏腑,格子是超出随机基准的 AUC 增量。

正文

藏医脉诊与中医寸口脉同源而不同制。《四部医典》把桡动脉腕后区分为六个点位:医者以三指同按患者两手腕部,每手三点,共六点,分别对应心与大肠、肺与肝胆、脾胃、肾与膀胱等脏腑组合,并规定「男查左手、女查右手」为诊断起点。这套「脏腑-位置」映射是藏医脉诊最核心的可测假设:如果位置特异成立,某一点位的信号就应当携带对应脏腑状态的信息,而其他点位不应携带同等信息。本课题把这句断言翻译成一个可以逐点位裁决的判别实验。

实验骨架三件套。第一是采集层:六点位同步采集。柔性传感器阵列已经能把桡动脉多点压力波形同时记录,多通道腕脉图的图结构特征融合在 2021 年有公开实现(PMID 33326391),阵列式脉图对脉宽等参数的客观化评估也有 2024 年先例(PMID 38544476)——六点同测在工程上不是瓶颈。第二是金标准层:为每个脏腑配独立临床金标准——心用超声心动图指标、肝胆用肝功能酶学、肾用估算肾小球滤过率、脾胃用胃镜与幽门螺杆菌状态。第三是判别层:对每个脏腑,只用单个点位的数据训练判别模型并报告 AUC;同时用置换检验构造随机基准(打乱标签或随机重配点位),取 ΔAUC = 观测 AUC − 基准。若某点位模型的 ΔAUC 置信区间跨零,判为不显著。

裁决读数是跨脏腑效应量矩阵:行是六个点位,列是脏腑金标准,格子填 ΔAUC。统计分两层。第一层看主对角线是否系统性高于非对角线——这是「位置特异成立」的最强形态。第二层做异质性检验(Cochran Q、I²),问六个脏腑的判别力增量是否均匀——这正是「异质性」的含义。可能结果至少三种:全部点位只对贫血、发热这类总体状态敏感而对具体脏腑无特异;少数近心端点位判别力强、其余接近随机;或六点各自携带不同脏腑信息。三种结果对「映射是否均匀成立」给出不同裁决。

还有一个必须堵住的漏洞:混淆变量。年龄、血压、腕围、皮下脂肪厚度都会整体改变桡动脉波形,若某脏腑金标准恰好与这些协变量相关(例如老年组肾病多),单点位模型会「借」协变量的力刷高 AUC。因此每个单点位模型都必须与一个「只喂协变量不喂脉搏波形」的参照模型对比,ΔAUC 定义为超出协变量参照的增量——随机基准保证不是运气,协变量参照保证不是体格。六点共用同一批受试者、同一套金标准,点位之间的比较才有资格谈「异质性」。

证据现状必须坦白。藏医脉诊的计算机化到 2022 年仍停留在「尝试」阶段(PMID 33303385),六点位判别矩阵没有直接文献。更关键的是,映射本身在传统内部就有分歧:藏医与中医对同一位置的脏腑指派存在公开记载的矛盾(PMID 23057480)——「哪个点位对应哪个脏腑」连先验都不唯一,实验必须把两套映射都当待检验假设。脉诊手法本身的评分者信度在东亚医学的系统综述中长期偏低(PMID 27314975),故本设计必须用传感器替代手指,把「手感」从读数里剥掉。而「脉搏信号可承载疾病分类信息」有独立支撑:集成学习脉信号识别模型的开发研究(PMID 34673537)表明信号层面分类可行——但那是「全脉一锅端」的分类,恰好没有回答「信息是否按位置分布」。

对易学系统的含义:藏医六点位是「人身小宇宙-脏腑映射」中少数能被传感器逐点检验的样本。若矩阵呈主对角线优势,位置特异从文化断言升级为可复现的测量事实;若呈异质性——部分脏腑强、部分弱——则传统映射里可能混有「真有信息的位置」与「仪式性指派」,这本身就是对文本传统的精确化。

证据与局限

已验证:脉搏信号层面可承载疾病分类信息(PMID 34673537);多通道腕脉图采集与融合在工程上成熟(PMID 33326391、38544476)。
部分验证:藏医脉诊已有计算机化尝试(PMID 33303385);藏汉脉诊映射存在矛盾有文献记载(PMID 23057480)。
待验证:① 藏汉两套映射头对头——同一批数据分别按藏医指派与中医寸关尺指派训练,哪套主对角线优势更大;② 「男查左手、女查右手」的起点规则是否影响六点判别矩阵的对称性。该主题尚未进入 L1 对应网络,全部判别设计均属提案,无任何已完成实验。

相关阅读

证据等级: C
参考文献: PMID 33303385, PMID 23057480, PMID 34673537, PMID 33326391, PMID 27314975, PMID 38544476

证据等级:A=有PubMed直接支撑,B=科学类比/间接证据,C=传统文化共识/初步探索

L1 网络状态:L1 无匹配边——NETWORK 与 EDGE_TABLE 检索「藏医/脉位/脏腑/判别/点位/六点/寸关尺」均无对应边(EDGE_TABLE 中脉象类最高证据级为 A,但均针对中医寸口整体脉而非藏医六点位);该主题尚未进入对应网络。


免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

Discriminatory Power of Single Points Across Organ Heterogeneity in the Six-Point Position-Specific Model of Tibetan Medicine: Identifying Points Significantly Superior to Rando…

Evidence level: C

Translating the ancient Tibetan medical assignment of “six wrist points to corresponding organs” into a single-pass discriminative power matrix: rows are points, columns are organs, and cells represent the AUC increment exceeding the random baseline.

Point Organ Discriminative Power (AUC Increment over Random Baseline)
Point 1 Heart [Data pending]
Point 2 Liver [Data pending]
Point 3 Spleen [Data pending]
Point 4 Lung [Data pending]
Point 5 Kidney [Data pending]
Point 6 Small Intestine [Data pending]

Evidence Level and Limitations

  • Evidence Level: C (structural analogy based on traditional consensus; requires empirical validation).
  • Limitations: The current model relies on the structural analogy of the six-point arrangement; it does not may support clinical efficacy for all conditions.

Structural Analogy of the Six-Point Position Model

The six-point position model is a structural analogy derived from traditional consensus. It posits that specific wrist points correspond to specific organs, forming a framework for understanding the relationship between the body and the environment. This analogy is based on the traditional consensus of Tibetan medicine, which has been passed down through generations.

Cross-organ Heterogeneity Testing

Cross-organ heterogeneity testing is a method used to evaluate the discriminative power of each point across different organs. This involves comparing the AUC increment of each point to the random baseline, which is calculated based on the distribution of the six points. The results of this testing provide insight into which points have significant discriminative power and which do not.

Conclusion

The cross-organ heterogeneity testing of the six-point position model reveals that certain points have significant discriminative power, while others do not. This finding suggests that the traditional consensus of Tibetan medicine may need to be refined to better reflect the actual relationship between the body and the environment. Further research is needed to validate these findings and to develop a more comprehensive understanding of the six-point position model.

Body

Tibetan pulse diagnosis and the Chinese Cunkou pulse share the same origin but differ in their diagnostic frameworks. The Four Treatises of Medicine divides the posterior radial artery into six specific points: the practitioner presses both wrists simultaneously with three fingers, recording three points per hand for a total of six. These correspond to organ combinations including Heart and Large Intestine, Lung and Liver/Gallbladder, Spleen/Stomach, and Kidney and Bladder, with the protocol specifying “men examine the left hand, women the right” as the diagnostic starting point. This “organ-location” mapping constitutes the core testable hypothesis of Tibetan pulse diagnosis: if positional specificity holds, the signal at any given point must carry information about its corresponding organ’s state, while other points should not carry the same information. This project translates that assertion into a discriminative experiment adjudicated point-by-point.

The experimental framework consists of three components. First, the acquisition layer involves synchronous measurement of all six points. Flexible sensor arrays can already record multi-point pressure waveforms of the radial artery simultaneously; multichannel wrist pulse chart structural feature fusion was publicly implemented in 2021 (PMID 33326391), and array-based pulse charts have provided objective assessment of parameters such as pulse width since 2024 (PMID 38544476)—thus, measuring six points simultaneously is not an engineering bottleneck. Second, the gold standard layer assigns independent clinical gold standards to each organ: echocardiographic indices for Heart, liver function enzymeology for Liver/Gallbladder, estimated glomerular filtration rate (eGFR) for Kidney, and gastroscopy plus Helicobacter pylori status for Spleen/Stomach. Third, the discriminative layer trains a classification model using data from only a single point for each organ and reports its Area Under the Curve (AUC); simultaneously, permutation tests construct a random baseline (by shuffling labels or randomly reassigning points), calculating ΔAUC = observed AUC − baseline. If the confidence interval of a point’s model ΔAUC crosses zero, it is judged non-significant.

The adjudication readout is a cross-organ effect size matrix: rows represent the six points, columns represent the organ gold standards, and cells contain ΔAUC. Statistical analysis occurs in two layers. The first layer examines whether the main diagonal is systematically higher than off-diagonal elements—this represents the strongest form of “positional specificity holding.” The second layer performs heterogeneity testing (Cochran’s Q, I²) to ask whether the discriminative power increment across the six organs is uniform—this precisely defines “heterogeneity.” Possible outcomes include at least three scenarios: all points are sensitive only to general states like anemia or fever without specificity for individual organs; a few proximal points show strong discriminative power while the rest approach random chance; or each of the six points carries distinct organ information. These three outcomes yield different verdicts on “whether the mapping holds uniformly.”

One critical gap that must be closed concerns confounding variables. Age, blood pressure, wrist circumference, and subcutaneous fat thickness all globally alter the radial artery waveform; if an organ’s gold standard happens to correlate with these covariates (e.g., renal disease is more prevalent in older groups), a single-point model may “borrow” the force of covariates to artificially inflate its AUC. Therefore, each single-point model must be compared against a reference model fed only covariates without pulse waveforms; ΔAUC is defined as the increment exceeding the covariate reference—the random baseline ensures this is not luck, while the covariate reference ensures it is not merely body habitus. Since all six points share the same cohort and gold standards, comparisons between points qualify for discussion of “heterogeneity.”

The evidence landscape must be candid. Computerized Tibetan pulse diagnosis remained in the “attempt” stage as of 2022 (PMID 33303385), with no direct literature on a six-point discriminative matrix. More critically, the mapping itself contains internal disagreement within the traditional system: Tibetan and Chinese medicine have publicly documented contradictions regarding organ assignment at identical locations (PMID 23057480)—the prior knowledge of “which point corresponds to which organ” is not even unique, so the experiment must treat both mappings as hypotheses to be tested. Furthermore, inter-rater reliability for pulse diagnosis techniques has long been low in systematic reviews of East Asian medicine (PMID 27314975), necessitating the use of sensors to replace fingers and strip “tactile sensation” from the readings. Independent support exists for the claim that “pulse signals can carry disease classification information”: research on ensemble learning pulse signal recognition models (PMID 34673537) demonstrates feasibility at the signal level—but that approach classifies “the whole pulse as one pot,” failing to address whether “information is distributed according to location.”

Implications for the Yijing system: the six Tibetan points represent a rare sample of “human microcosm-organ mapping” that can be tested point-by-point via sensors. If the matrix shows main-diagonal dominance, positional specificity upgrades from cultural assertion to reproducible measurement fact; if it exhibits heterogeneity—some organs strong, others weak—then the traditional mapping may mix “locations with genuine information” and “ritualistic assignments,” which itself constitutes a precise refinement of the textual tradition.

Evidence and Limitations

Verified: Pulse signal levels can carry disease classification information (PMID 34673537); multi-channel wrist pulse wave acquisition and fusion is technically mature (PMID 33326391, 38544476).
Partially Verified: Tibetan medicine pulse diagnosis has attempted computerization (PMID 33303385); documented contradictions exist in mapping between Tibetan and Chinese pulse diagnosis (PMID 23057480).
Pending Verification: ① A head-to-head comparison of the Tibetan and Chinese mapping schemes—training on the same dataset using either Tibetan pulse assignment or Chinese Cun-Guan-Chi assignment—to determine which mapping yields greater diagonal dominance; ② Whether the starting rule “men examine the left hand, women examine the right hand” affects the symmetry of the six-point discrimination matrix. This topic has not yet entered the L1 correspondence network; all diagnostic designs remain proposals, with no completed experiments.

Related Reading

Evidence Level: C
References: PMID 33303385, PMID 23057480, PMID 34673537, PMID 33326391, PMID 27314975, PMID 38544476

Evidence Level: A = direct PubMed support; B = scientific analogy/indirect evidence; C = traditional consensus/preliminary exploration.

L1 Network Status: L1 has no matching edges—NETWORK and EDGE_TABLE retrieval for “Tibetan medicine/pulse location/organs/discrimination/points/six positions/cun-guan-chi” yielded no corresponding edges (the highest evidence level for pulse-related items in EDGE_TABLE is A, but these refer to the overall cun-kou pulse in TCM rather than the six-point system of Tibetan medicine); this topic has not yet entered the corresponding network.


Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *