藏医六点脉诊的判别增量检验:位置特异模型能赢过全局模型吗

Written by

in

Language

Evidence level: B

藏医认为左右手腕六个触诊点各自”候”不同的脏腑——判别增量检验把这句话翻译成一个可统计检验的问题:只用指定位点建模型,是否真的比”六个点全用”或”随便挑一个点”判别得更准?

一个能决定理论存亡的实验设计

藏医脉诊的核心断言不只是”脉搏波能反映疾病”,而是“位置-脏腑”的固定分配:左手冲位候心与小肠、甘位候脾与胃、恰位候左肾与生殖器;右手冲位候肺与大肠、甘位候肝与胆、恰位候右肾与膀胱。如果这个映射是真的,那么”用右甘位一个点的信号判别肝功能异常”应该显著优于”随便用哪个点”;如果映射只是文化约定,那么六点不过是对同一血流动力学状态的重复采样,位置特异模型不可能带来任何判别增量。

判别增量检验(incremental validity testing)正是为这种问题设计的统计框架:在同一批六点脉搏数据上,分别构建位置特异模型(只用藏医指派给目标脏腑的那个点位,如肝功能只用右甘位)与全局模型(六点全量特征),再用金标准(影像/生化)判定谁更准。关键统计量有两个:位置特异模型 vs “随机选一个点”基准线的 ΔAUC 检验;以及相对全局模型的净重分类指数(NRI)——即”换用位置特异模型后,有多少病人被正确重分类”。若位置特异模型跑赢随机基准线,藏医的脏腑分配表就拿到第一块直接证据;若六个点判别力无差异,则”位置映射”更像是代代相传的操作约定而非生理事实——这同样是科学上极有价值的结论,因为它把需要验证的问题精确圈定了出来。

位置不是摆设:三个物理基础证据

在动手检验之前,先看”位置差异”到底有没有物理基础。超声成像研究给出了最直接的答案:韩国学者对 44 名受试者用超声测量寸、关、尺三个位置及其邻近非诊脉段的动脉深度、直径和血流速度,结果显示三个诊脉位置之间的血流动力学和解剖参数存在显著差异(P<0.01),而这种差异在邻近非诊脉段并不明显(PMID 25763090)——桡动脉在 5 厘米范围内的解剖形态确实不是均匀的,诊脉位置恰好落在差异显著的区域。解剖学上同样支持这一点:对 16 具尸体的 32 侧上肢解剖显示,寸位的搏动血管可能是桡动脉也可能来自掌浅支,血管分布因人而异(PMID 31205476),这意味着不同人”按到的血管”可能本来就不一样。

频谱层面也有证据。2018 年一项研究用频谱分析仪测量中医诊脉各位置,发现寸、关、尺每位点呈现不同的谐波特征(PMID 30105076)。更直接的判别证据来自 2011 年一项 138 人的研究:研究者系统测量了左右腕六个诊脉位置的桡动脉压脉冲与心率变异性,发现 10-50 Hz 频段能量占比(EP 10-50Hz%)在六个位置间存在显著差异,且只有中医理论中”候心脏”的左寸位能够显著区分正常血压与高血压受试者(PMID 21978219)——这是”指定位点确实携带其他位点没有的判别信息”的罕见直接证据。

全局信号很强:ML 判别已多次验证

支撑”全局模型”的文献非常扎实。2024 年一项 608 人的研究用桡动脉脉搏波判别冠心病,Extra Trees 分类器测试集准确率 85.8%、AUC 0.936(PMID 39285363)。2025 年一项 58 名血液透析患者的研究用脉搏波判别冠状动脉钙化严重度(金标准为低剂量 CT 的 Agatston 评分),GBDT 模型五折交叉验证平均准确率 84.1%、宏 AUC 0.962(PMID 40883745)。2023 年的 PCOS 研究同样证实桡动脉脉搏波能高准确率判别多囊卵巢综合征(PMID 37957660)。这说明”整段桡动脉波形携带疾病信息”已是共识——而判别增量检验要回答的,正是藏医体系更激进的那个主张:不仅整段波形有用,特定位置的波形还有专属的脏腑对应。

技术底座:六点同步采集已不是障碍

藏医”三指同时按六点”的物理难点在于手指不可能完全同步、等压覆盖六个位置。2023 年发表于 ACS Nano 的 27 通道(3×9)高密度柔性脉搏感知阵列,可以一次提取脉搏的空间-时间分布信息,自动识别率高达 97.8%(PMID 36716225)。结合藏医脉诊计算机化的既有探索(PMID 33303385),”一块贴片覆盖整个诊脉区、左右各三点同时采集、每点记录按压深度”已是工程上可行的常规操作。

判别增量检验的三步落地设计

把证据链闭合成一个可执行实验:第一步,用独立客观指标定义脏腑功能障碍金标准(肝功能取 ALT/AST 或弹性成像、肾功能取 eGFR、心功能取超声心动),判定者不知脉诊结果,且金标准与脉诊不共享任何测量通道;第二步,同一批六点数据上建三套模型——位置特异模型(仅用指派点位)、全局模型(六点全量)、随机点位模型(随机选一个点作基准线),比较 ΔAUC 与 NRI;第三步,控住关键混杂:按压深度必须记录为协变量(藏医冲轻按、恰重按,深度本身改变波形谱,PMID 30105076),左右手分开建模,灰色地带样本单独处理。

证据与局限

已验证: 桡动脉脉搏波经机器学习可判别冠心病(AUC 0.936)、肾病钙化(AUC 0.962)、PCOS 等疾病;寸关尺三位置的解剖与血流参数存在显著差异(超声);六位点中左寸位特异判别高血压;每位点谐波特征不同;多通道阵列可同步采集六点以上信号。

部分验证: “谐波-脏腑”对应有系统综述支持但方法学异质性强;中医体系(寸关尺)的位置特异判别有零星直接证据,藏医体系(冲甘恰)的脏腑分配尚未被任何判别增量研究检验过。

待验证: 藏医六点”位置-脏腑”映射能否转化为判别增量(位置特异模型 vs 全局模型 vs 随机基准线的 ΔAUC/NRI);按压深度梯度作为脏腑判别独立特征的增量价值;六点位置特异模型的多中心泛化稳定性。

相关阅读

证据等级: B(”脉搏波+ML判别疾病”与”位置存在物理差异”有直接证据,但”藏医六点位置-脏腑映射的判别增量”整体尚无直接实证,属实验设计提案)
参考文献: PMID 21978219, PMID 25763090, PMID 30105076, PMID 31205476, PMID 39285363, PMID 40883745, PMID 37957660, PMID 33303385, PMID 36716225


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。

Incremental Validation of Six-Point Pulse Diagnosis in Tibetan Medicine: Can Location-Specific Models Outperform Global Models?

Evidence level: B

Tibetan medicine posits that each of the six palpation points on the wrists “corresponds” to specific organs. Incremental validity testing translates this assertion into a statistically testable question: Does building a model using only designated points yield significantly better discrimination than using “all six points” or “randomly selecting one point”?

An Experimental Design Capable of Determining Theoretical Viability

The core assertion of Tibetan pulse diagnosis is not merely that “pulse waves reflect disease,” but the fixed allocation between location and organ: The left hand’s Chong position corresponds to the heart and small intestine, Gan to the spleen and stomach, and Cha to the left kidney and reproductive organs; the right hand’s Chong corresponds to the lung and large intestine, Gan to the liver and gallbladder, and Cha to the right kidney and bladder. If this mapping is physiologically valid, then “using the signal from a single point at the right Gan position to discriminate liver dysfunction” should significantly outperform “using any arbitrary point.” Conversely, if the mapping is merely a cultural convention, the six points represent redundant sampling of the same hemodynamic state, and a location-specific model would yield no incremental discriminative power.

Incremental validity testing is the statistical framework designed precisely for such questions. It constructs a location-specific model (using only the point assigned to the target organ per Tibetan prescription, e.g., right Gan for liver function) and a global model (using all six points as features) on the same set of six-point pulse data, then judges accuracy against a gold standard (imaging/biochemistry). Two key statistics are employed: the ΔAUC test comparing the location-specific model against a “randomly selected point” baseline; and the Net Reclassification Index (NRI) relative to the global model—specifically, how many patients are correctly reclassified when switching to the location-specific model. If the location-specific model surpasses the random baseline, the Tibetan organ allocation table receives its first direct empirical evidence. If discriminative power shows no difference across the six points, then “location mapping” appears more like an inherited operational convention than a physiological fact—a conclusion equally valuable scientifically as it precisely delineates the problem requiring verification.

Location Is Not Ornamental: Three Physical Basis Evidence

Before conducting the test, one must examine whether “location differences” possess a physical basis. Ultrasound imaging studies provide the most direct answer: Korean researchers measured arterial depth, diameter, and blood flow velocity at the Cun, Guan, and Chi positions versus adjacent non-diagnostic segments in 44 subjects. Results showed significant hemodynamic and anatomical parameter differences between diagnostic locations (P<0.01), whereas these differences were not apparent in adjacent non-diagnostic segments (PMID 25763090)—the anatomical morphology of the radial artery within a 5 cm range is indeed non-uniform, and diagnostic points fall precisely into regions with significant variation. Anatomical evidence similarly supports this: dissection of 32 upper limbs from 16 cadavers revealed that the pulsating vessel at Cun could be the radial artery or the superficial palmar branch, with distribution varying by individual (PMID 31205476), implying that the vessel “felt” by different individuals may inherently differ.

Evidence also exists at the spectral level. A 2018 study using a spectrum analyzer measured pulse points across TCM diagnostic locations, finding distinct harmonic characteristics at each point (PMID 30105076). More direct discriminative evidence comes from a 2011 study of 138 individuals: researchers systematically measured radial artery pressure pulses and heart rate variability at six diagnostic locations on both wrists, finding significant differences in energy proportion within the 10–50 Hz band (EP 10-50Hz%) across locations, and notably, only the left Cun position, which TCM theory assigns to the “heart,” could significantly distinguish between normotensive and hypertensive subjects (PMID 21978219)—a rare direct evidence that “designated points carry discriminative information absent at other locations.”

Strong Global Signal: ML Discrimination Already Verified Multiple Times

Literature supporting the “global model” is robust. A 2024 study with 608 participants used radial artery pulse waves to discriminate coronary heart disease, achieving a test set accuracy of 85.8% and an AUC of 0.936 with an Extra Trees classifier (PMID 39285363). A 2025 study of 58 hemodialysis patients used pulse waves to discriminate coronary artery calcification severity (gold standard: Agatston score from low-dose CT), with a GBDT model showing 84.1% average accuracy via five-fold cross-validation and a macro AUC of 0.962 (PMID 40883745). A 2023 PCOS study similarly confirmed that radial artery pulse waves can discriminate polycystic ovary syndrome with high accuracy (PMID 37957660). This indicates that “the entire radial artery waveform carries disease information” is now a consensus. The incremental validity testing aims to answer the more radical claim of the Tibetan system: not only is the entire waveform useful, but waveforms at specific locations have exclusive organ correspondences.

Technical Foundation: Six-Point Synchronous Acquisition Is No Longer an Obstacle

The physical challenge in Tibetan pulse diagnosis—”pressing six points simultaneously with three fingers”—lies in the impossibility of perfect synchronization and equal pressure coverage across six locations. A 2023 study published in ACS Nano introduced a 27-channel (3×9) high-density flexible pulse sensing array capable of extracting spatiotemporal distribution information of pulses in a single extraction, with an automatic recognition rate up to 97.8% (PMID 36716225). Combined with existing explorations into computerizing Tibetan pulse diagnosis (PMID 33303385), “using a single patch to cover the entire diagnostic area, collecting three points on each side simultaneously, and recording pressure depth for each point” is now a routine engineering feasibility.

Three-Step Implementation Design for Incremental Validity Testing

Closing the evidence chain into an executable experiment: Step 1: Define gold standards for organ dysfunction using independent objective indicators (ALT/AST or elastography for liver function, eGFR for kidney function, echocardiography for heart function); ensure the adjudicator is blind to pulse diagnosis results, and that the gold standard shares no measurement channels with pulse diagnosis. Step 2: Construct three models on the same set of six-point data: a location-specific model (using only assigned points), a global model (all six points), and a random point model (randomly selecting one point as the baseline); compare ΔAUC and NRI. Step 3: Control key confounders: record pressure depth as a covariate (Tibetan Chong is light, Cha is heavy; depth itself alters waveform spectra, PMID 30105076); model left and right hands separately; handle gray-zone samples independently.

Evidence and Limitations

Verified: Radial artery pulse waves can discriminate coronary heart disease (AUC 0.936), kidney calcification (AUC 0.962), PCOS, etc., via machine learning; significant differences exist in anatomical and hemodynamic parameters among the three Cun-Guan-Chi positions (ultrasound); the left Cun among six points shows specific discriminative power for hypertension; each point exhibits distinct harmonic features; multi-channel arrays can synchronously capture signals from six or more points.

Partially Verified: Systematic reviews support “harmonic-organ” correspondences, though methodological heterogeneity is strong; the Tibetan system (Chong-Gan-Cha) has not yet been subjected to any incremental validity study regarding its organ allocation, despite some sporadic direct evidence for location-specific discrimination in the TCM system (Cun-Guan-Chi).

To Be Verified: Whether the Tibetan six-point “location-organ” mapping can be converted into incremental validity (comparing location-specific vs. global vs. random baseline via ΔAUC/NRI); the incremental value of pressure depth gradients as independent discriminative features for organs; the multi-center generalization stability of the six-point location-specific model.

Related Reading

Evidence Level: B (“Pulse waves + ML discrimination of disease” and “existence of physical differences in location” have direct evidence, but the “incremental validity of Tibetan six-point location-organ mapping” lacks overall direct empirical support and remains an experimental design proposal).
References: PMID 21978219, PMID 25763090, PMID 30105076, PMID 31205476, PMID 39285363, PMID 40883745, PMID 37957660, PMID 33303385, PMID 36716225


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 *