Evidence level: B
藏医”冲轻按、甘中按、恰重按”的按压深度梯度可能不只是手感习惯——现代脉诊量化研究已证明,施加压力会系统性改变脉波谐波特征。问题只在于:这些深度依赖的变化,是否真的携带了位置之外独立的脏腑判别信息?
一个被忽略的独立变量
判别增量检验问的是”位置”维度。但脉诊还有第二个维度——按压深度:藏医冲位轻按(候浅层)、甘位中按、恰位重按(候深层),与现代中医”浮取、中取、沉取”一脉相承。如果深度只是操作习惯,无论按多重,同一位置的脉波谱都应一样;如果深度真携带信息,”同一位置在不同压力下波形的变化轨迹”就是一组全新判别特征。
科学表述是:按压深度是否具有独立于位置的判别增量——固定位置采集时,深度梯度(压力-波形关系曲线)能否显著提升对特定脏腑状态的判别准确率。这与位置判别增量正交:位置问”按哪里”,深度问”按多重”。
深度不是玄学:物理证据
最直接的证据来自按脉深度标准化研究:2015 年一项研究用神经网络,以接触压力和传感器位移为输入估算桡动脉宽度来标准化浮中沉三候,平均预测误差仅 1.19%,比直接用位移模型误差降低 71.62%(PMID 25522334)——”浮中沉”对应稳定可测量的物理量,不是模糊的主观手感。
第二层证据是压力确实改变脉波本身。2011 年一项 45 人研究,对手指施加 0-200 mmHg 接触压力梯度,同步测量指端光电容积脉搏波(PPG)与桡动脉血压波(BP),频域分析显示不同压力下 BP-PPG 传递函数谐波参数显著不同,60 mmHg 处一致性最优(PMID 22034741)——按多重,测到的谐波谱就是不一样的。
第三层是机制层面:2017 年心血管模拟器研究用硅胶主动脉-桡动脉模型独立控制平均压力、心率、反射波叠加与脉压,成功复现浮脉(低压力下振幅最大)与沉脉(高压力下振幅最大)(PMID 28246538)。这从流体力学确认了”浮主表、沉主里”的机制:当血管内压与外加压力相等时脉搏振幅达峰(P-H 曲线顶点),浮沉本质是峰点在压力轴上的位置,反映血管壁张力与血压的力学状态。
深度信息能用吗:机器学习已试过
如果深度梯度真的改变波形,”不同压力下的波形差异”就是天然可分类的信号。2022 年一项研究给健康年轻人施加”合适压力(67.80±1.55 mmHg)”与”更高压力(151.80±3.19 mmHg)”两种接触压力,采集桡动脉血压波,提取 40 个谐波指标,用八种机器学习算法判别波形来自哪种压力条件,随机森林留一受试者交叉验证 AUC 达 0.96(PMID 36433203)。虽该研究目的是筛选波形质量,但证明:不同压力下采集的脉波,谐波差异大到机器学习可近完美区分。
解剖层面还有细节:2019 年 16 具尸体 32 侧上肢解剖显示,寸位搏动血管可能是桡动脉也可能来自掌浅支,个体差异大,按压时指腹接触面变宽(PMID 31205476)——”按多重”还改变实际触到的血管段。桡动脉压平眼压测量综述则提供工程侧支撑:轻度压力部分压平动脉可稳定获取桡动脉压波并反演中心主动脉压(PMID 20435839)。
判别增量检验的实验设计
闭合为可执行实验:第一步,同一批受试者、固定传感器与位置,逐级扫描压力(30-180 mmHg,每 15 mmHg 一档),每档记录波形并提取谐波特征;第二步,构建三组特征——单一固定压力、深度轨迹(P-H 曲线峰位/峰宽/斜率)、固定压力+深度轨迹叠加,分别判别目标脏腑状态(金标准用影像/生化);第三步,比较三者 ΔAUC 与净重分类指数(NRI):若叠加模型显著优于仅固定压力,则深度梯度携带独立判别信息;若不显著,则深度只是位置的冗余编码。关键混杂:BMI 与皮下脂肪厚度(胖人脉位深,需按体重校准,25522334)、血管个体差异(31205476)、左右手分开建模。
证据与局限
已验证: 浮中沉三候可由接触压力+位移精确量化(预测误差 1.19%);不同接触压力系统性改变脉波谐波特征(45 人 BP-PPG 研究);压力差异大到机器学习可近完美判别(AUC 0.96);心血管模拟器可复现浮沉脉力学机制。
部分验证: “深度-脏腑”映射有机制链支持,但尚无研究直接用深度梯度特征做脏腑判别增量检验;P-H 曲线峰位与血压/血管壁张力对应较清晰,与具体脏腑的关联仍是传统知识。
待验证: 深度梯度特征相对固定压力特征的判别增量(ΔAUC/NRI);深度轨迹特征作为新特征源的稳定性与跨中心泛化;深度与位置信息的交互(藏医冲轻甘中恰重是否各有独立最优深度);BMI/年龄/血管个体差异对深度判别的混杂量化。
相关阅读
- 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
- 藏医六点脉诊的盲法判别模型:让”按哪里候什么”接受统计检验
- 藏医六点脉诊的信号差异与脏器归属
- 脉证相应的盲法判别与增量价值检验
- 研究池子:生物频率密码
证据等级: B(”按脉深度可量化”与”压力改变波形特征”有直接证据,但”深度梯度携带独立脏腑判别信息”尚无直接实证,属实验设计提案)
参考文献: PMID 25522334, PMID 22034741, PMID 28246538, PMID 36433203, PMID 31205476, PMID 20435839
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。
Is Pulse Depth Merely a Formality: Can the Floating, Medium, and Deep Levels Carry Independent Organ Discrimination Information
Evidence level: B
The pressure gradient in Tibetan medicine—”chong” (light), “gan” (medium), and “qia” (heavy)—may not merely reflect tactile habit. Modern quantification studies of pulse diagnosis have demonstrated that applied pressure systematically alters harmonic features of the pulse wave. The question is whether these depth-dependent variations truly carry organ discrimination information independent of location.
An Overlooked Independent Variable
Incremental discrimination testing addresses the “location” dimension. However, pulse diagnosis possesses a second dimension—pressing depth: Tibetan medicine distinguishes between chong (light pressing for superficial layers), gan (medium pressing), and qia (heavy pressing for deep layers), a lineage shared with modern Traditional Chinese Medicine’s “floating, medium, and sinking grasping.” If depth is merely an operational habit, the pulse spectrum at a single position should remain identical regardless of pressure intensity; conversely, if depth carries information, the “trajectory of waveform changes at the same position under varying pressures” constitutes a novel set of discriminative features.
The scientific formulation is: Does pressing depth provide an incremental discriminative gain independent of location? When data is collected at a fixed position, can the depth gradient (pressure-waveform relationship curve) significantly improve discrimination accuracy for specific organ states? This is orthogonal to location discrimination: location asks “where to press,” while depth asks “how much pressure.”
Depth Is Not Metaphysics: Physical Evidence
The most direct evidence comes from studies standardizing pulse depth. A 2015 study utilized neural networks, using contact pressure and sensor displacement as inputs to estimate radial artery width for standardizing the three depths. The average prediction error was only 1.19%, a 71.62% reduction compared to direct displacement models (PMID 25522334)—indicating that “floating, middle, and sinking” correspond to stable, measurable physical quantities rather than vague subjective sensations.
A second layer of evidence confirms that pressure alters the pulse wave itself. A 2011 study involving 45 subjects applied a contact pressure gradient of 0–200 mmHg to the finger while simultaneously measuring photoplethysmographic (PPG) waves at the fingertip and blood pressure (BP) waves in the radial artery. Frequency domain analysis revealed that harmonic parameters of the BP-PPG transfer function differed significantly under different pressures, with optimal consistency observed at 60 mmHg (PMID 22034741)—meaning the harmonic spectrum measured under different pressures is distinct.
The third layer addresses mechanism: a 2017 cardiovascular simulator study used a silicone aorta-radial artery model to independently control mean pressure, heart rate, reflected wave superposition, and pulse pressure. It successfully reproduced the “floating pulse” (maximum amplitude under low pressure) and the “sinking pulse” (maximum amplitude under high pressure) (PMID 28246538). This confirms from fluid dynamics the mechanism of “floating indicates exterior, sinking indicates interior”: when intravascular pressure equals applied external pressure, pulse amplitude peaks at the vertex of the P-H curve. The essence of floating and sinking is the position of this peak on the pressure axis, reflecting the mechanical state of vascular wall tension and blood pressure.
Can Depth Information Be Utilized? Machine Learning Has Tried It
If the depth gradient truly alters the waveform, “differences in waveforms under different pressures” constitute naturally classifiable signals. A 2022 study applied two contact pressures to healthy young adults: “appropriate pressure” (67.80±1.55 mmHg) and “higher pressure” (151.80±3.19 mmHg), collected radial artery BP waves, extracted 40 harmonic indicators, and used eight machine learning algorithms to discriminate whether the waveform originated from which pressure condition. Random forest leave-one-subject-out cross-validation yielded an AUC of 0.96 (PMID 36433203). Although the study aimed to filter waveform quality, it demonstrated that harmonic differences between pulses collected under different pressures are large enough for machine learning to distinguish them with near-perfect accuracy.
Anatomical details also exist: a 2019 anatomical study of 32 sides from 16 cadavers showed that the pulsating vessel at the cun position could be the radial artery or the superficial palmar branch, with significant individual variation. During pressing, the contact area of the fingertip widens (PMID 31205476)—meaning “pressing deeper” also alters the specific vascular segment actually contacted. A review on radial artery pressure flattening for intraocular pressure measurement provides engineering support: mild pressure partially flattens the artery, stabilizing the acquisition of radial artery BP waves and enabling central aortic pressure inversion (PMID 20435839).
Experimental Design for Incremental Discrimination Testing
The closed experimental design proceeds as follows: Step 1: Using the same cohort of subjects, fixed sensors, and fixed positions, scan pressure levels sequentially (30–180 mmHg, every 15 mmHg), recording the waveform and extracting harmonic features at each level; Step 2: Construct three feature groups—single fixed pressure, depth trajectory (peak position/width/slope of the P-H curve), and the superposition of fixed pressure plus depth trajectory—each separately discriminating target organ states (using imaging/biochemistry as the gold standard); Step 3: Compare the ΔAUC and Net Reclassification Index (NRI) among the three models: if the superposition model significantly outperforms the fixed-pressure-only model, the depth gradient carries independent discriminative information; if not significant, depth is merely redundant encoding of location. Key confounders include BMI and subcutaneous fat thickness (deeper pulse locations in heavier individuals require weight calibration, PMID 25522334), individual vascular variations (PMID 31205476), and separate modeling for left and right hands.
Evidence and Limitations
Verified: The three depths of floating, middle, and sinking can be precisely quantified via contact pressure and displacement (prediction error 1.19%); different contact pressures systematically alter pulse wave harmonic features (45-subject BP-PPG study); pressure differences are large enough for machine learning to discriminate with near-perfect accuracy (AUC 0.96); and cardiovascular simulators can reproduce the hemodynamic mechanisms of floating and sinking pulses.
Partially Verified: There is mechanistic support for a “depth-organ” mapping chain, but no study has yet directly tested depth gradient features for incremental organ discrimination; the correspondence between P-H curve peak positions and blood pressure/vascular wall tension is clear, but the link to specific organs remains traditional knowledge.
To Be Verified: The incremental discriminative gain of depth gradient features relative to fixed pressure features (ΔAUC/NRI); the stability and cross-center generalization of depth trajectory features as a new feature source; the interaction between depth and location information (whether Tibetan medicine’s chong-light, gan-medium, qia-heavy each has an independent optimal depth); and the quantification of confounding effects of BMI, age, and individual vascular variations on depth discrimination.
Related Reading
- Incremental Discrimination Testing of Tibetan Six-Point Pulse Diagnosis: Can Position-Specific Models Outperform Global Models?
- Blinded Discrimination Model for Tibetan Six-Point Pulse Diagnosis: Subjecting “Where to Press” to Statistical Testing
- Signal Differences and Organ Attribution in Tibetan Six-Point Pulse Diagnosis
- Blinded Discrimination and Incremental Value Testing of Pulse-Syndrome Correspondence
- Research Pool: Biological Frequency Codes
Evidence Level: B (“Pressing depth can be quantified” and “pressure alters waveform features” have direct evidence, but “depth gradient carries independent organ discrimination information” lacks direct empirical validation; this is an experimental design proposal)
References: PMID 25522334, PMID 22034741, PMID 28246538, PMID 36433203, PMID 31205476, PMID 20435839
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.
Leave a Reply