脉搏谐波与脏腑频率映射:基于机器学习的C1-C10重验证

Written by

in

Language

Evidence level: B

导言

脉搏波是人体心血管系统运行状态的窗口。中医脉诊理论认为,不同脉位(寸、关、尺)与不同脏腑存在对应关系,而脉搏的形态特征(浮、沉、滑、涩)则反映脏腑功能的偏倾。这与现代医学中脉搏波形的频域分析——即将脉搏信号分解为基频与各次谐波分量——在方法论上具有深层呼应。

台湾生物力学专家王唯工提出,脉搏波的前十个谐波分量(C1至C10)分别对应不同脏腑的共振频率:C1对应心脏(心经),C2对应肝脏(肝经),C3对应脾脏(脾经),以此类推。这一”谐波-脏腑映射”假说若成立,则意味着脉搏谐波分析可以成为一种无创、实时的脏腑功能评估手段。

然而,王唯工的原始研究主要依赖动物实验和有限的临床样本,其方法学在当代生物信号处理标准下存在较大验证空间。本文系统梳理脉搏谐波与脏腑映射的科学证据,评估其机器学习重验证的研究现状,并提出标准化的验证框架。

脉搏谐波的信号处理基础

频域分析的生理学依据

脉搏波是由左心室周期性收缩产生的压力波在动脉系统中传播的结果。每一个心跳周期,动脉压力传感器记录到的波形并非单一频率的正弦波,而是由基频(心率,通常1-2 Hz)及其整数倍谐波组成的多频复合信号。

通过快速傅里叶变换(FFT),可将时域脉搏波信号分解为:

$$P(t) = \sum_{n=1}^{10} A_n \sin(2\pi n f_0 t + \phi_n)$$

其中 $f_0$ 为心率基频,$A_n$ 为第 $n$ 次谐波的幅值,$\phi_n$ 为相位。各谐波的幅值比(Harmonic Ratio Index, HRI)反映了心血管系统对不同频率成分的选择性放大或衰减。

为什么关注C1-C10?

脉搏波的能量主要集中于前10个谐波分量(C1-C10)。高频谐波(C7-C10)对应血管树的远端特性(动脉硬化、外周阻力),而低频谐波(C1-C3)主要受心脏收缩力和大动脉顺应性影响。中医五脏六腑分布于胸腔、腹腔的不同解剖深度,与动脉树的不同节段存在物理距离上的对应关系——这为”谐波-脏腑”假说提供了生物力学上的逻辑起点。

机器学习在脉搏谐波映射中的验证应用

已有证据:脉搏频谱分析的多疾病分类

近年来,多项研究将脉搏波频域特征与机器学习结合,用于疾病分类和功能评估,其中部分研究隐含地验证了谐波与生理状态之间的映射关系:

PMID 33903610(Lin SK et al., Sci Rep 2021):采集桡动脉血压波(radial BPW)和指尖光电容积波(PPG)信号,提取频谱特征后输入多层感知器(MLP)神经网络,用于区分阿尔茨海默病患者与正常对照。结论显示脉搏频谱特征可有效区分认知功能障碍患者(分类准确率AUC>0.85),提示脉搏谐波分布模式与脑部血流动力学状态存在可检测的关联——这与中医”心主神明”理论形成跨学科呼应。

PMID 36294440(Chen CC et al., J Clin Med 2022):结合非侵入性脉搏波测量和机器学习,分析辉瑞-BioNTech新冠疫苗(BNT162b2)不同副作用受试者的脉搏波形变异性。结果发现不同心血管副作用组的脉搏谐波特征存在显著差异,验证了脉搏频谱特征对心血管功能微小扰动的敏感性。

PMID 42305919(Lin CN et al., Front Physiol 2026):开发基于光电容积脉搏波(PPG)的可解释机器学习模型,实现收缩压的非侵入性评估。模型利用PPG波形的时域和频域特征,通过特征重要性分析揭示了低次谐波(C1-C3)对血压预测的贡献权重最高,与大动脉弹性功能的生理学预期一致。

PMID 36509825(Wu LW et al., Sci Rep 2022):将脉搏波频域分析与机器学习相结合,用于区分肌少症和代谢综合征患者。研究表明不同病理状态下的脉搏谐波能量分布存在系统性差异,提示脉搏频谱特征可作为多器官功能状态的”指纹”。

PMID 35161551(Hsiu H et al., Sensors 2022):采集桡动脉血压波信号,提取频谱指数后用机器学习分析,用于区分社区受试者的不同认知功能状态。结果证明脉搏谐波特征与认知功能之间存在统计显著的映射关系,间接支持了”脉搏频谱携带多系统信息”的假设。

从疾病分类到脏腑映射:关键缺口

以上研究的共同局限在于:它们验证的是”谐波特征与某种特定疾病/状态之间存在关联”,但并未直接验证”某次谐波对应某个特定脏腑”。从机器学习角度看,这相当于只训练了”整体脉搏谱图→疾病标签”的分类器,而没有验证”某次谐波的独立变异是否反映对应脏腑的独立功能变化”。

王唯工C1-C10假说要求的是特定谐波与特定脏腑之间的一一对应。要验证这一假设,需要:
1. 同时采集多脏器功能指标(金标准)
2. 检验特定谐波分量与对应脏器指标之间的独立相关性(控制其他变量后)
3. 在独立队列中复现该映射关系

C1-C10脏腑映射的科学假说与验证框架

假说内容

谐波 频率范围(估算) 中医对应 生理学对应
C1 ~1-2 Hz 心 心脏收缩力
C2 ~2-4 Hz 肝 肝血流量/门脉系统
C3 ~3-6 Hz 脾 消化系统血液分配
C4 ~4-8 Hz 肺 肺循环/气体交换
C5 ~5-10 Hz 肾 肾血流/水液代谢
C6-C10 ~6-20 Hz 六腑等 远端血管阻力/动脉硬化

注:实际频率范围因个体心率差异而变化,表中数值为以60 BPM心率为基准的估算值

验证方法论要求

信号采集标准化:采用同步三导联桡动脉压力波(radial APW)+指尖PPG,采样率≥200 Hz,时长≥3分钟/样本,确保频域分辨率≤0.1 Hz。

金标准脏器功能指标:需同步采集各脏腑功能金标准(如超声心动图测心输出量、肝功能酶谱、肾功能肌酐/尿素氮、肺功能FEV1等),将连续变量作为回归目标,而非仅做疾病有无的二分类。

统计方法:多变量线性回归或梯度提升树(XGBoost),以HRI向量(C1-C10的幅值比)为特征向量,各脏器功能指标为因变量,检验各次谐波对特定脏器的独立解释力(偏决定系数)。

交叉验证:留出法(hold-out)+ 5折交叉验证,防止过拟合,报告预测误差的置信区间。

证据等级与局限性

本课题的总体证据等级为 B级(观察性研究支持,存在初步机制研究,但尚缺乏严格控制的多中心验证)。

支持证据:多项独立研究证明脉搏谐波特征可区分不同生理/病理状态,且这种区分具有统计显著性和生物学合理性,间接支持谐波携带多维器官信息的假设。

关键局限:
1. 一一对应未验证:现有研究均为”整体谐波谱→单病种”的分类模式,未检验”谐波n独立变异→脏腑n独立功能”的直接对应关系
2. 中医脏腑的复杂性:中医”脏腑”是功能集合概念,与现代医学的器官并非简单一一对应,”脾”不等于解剖学上的脾脏,”肝”不等于单纯的肝脏——这增加了验证的操作化难度
3. 个体差异与标准化难题:不同年龄、性别、体型的受试者谐波基准值差异显著,建立通用基准图谱需要大样本、多中心数据
4. 信号采集标准化不足:现有研究设备各异、采样参数不统一,限制了跨研究比较

相关阅读

  • Wang WQ. The Research of Blood Flow Frequency Distribution and Resonance Theory of Internal Organs. Taiwan University Press, 1984.(王唯工脏腑共振理论原始文献)
  • Lin SK et al. (2021) Classification of patients with Alzheimer’s disease using the arterial pulse spectrum and a multilayer-perceptron analysis. Sci Rep 11:8882. PMID: 33903610
  • Hsiu H et al. (2022) Discrimination of the Cognitive Function of Community Subjects Using the Arterial Pulse Spectrum and Machine-Learning Analysis. Sensors 22(3):806. PMID: 35161551
  • Wu LW et al. (2022) Discrimination between possible sarcopenia and metabolic syndrome using the arterial pulse spectrum and machine-learning analysis. Sci Rep 12:21452. PMID: 36509825

本文属于池子③(生物频率密码)勘探系列 | 课题#6 C1-C10脏腑谐波映射机器学习重验证 | 证据等级:B | YY(分析官)2026-10-03


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。

Pulse Harmonics and Organ Frequency Mapping: Revalidation of C1-C10 Based on Machine Learning

Evidence level: B

Introduction

The pulse wave serves as a window into the operational state of the cardiovascular system. Traditional Yijing Studies theory posits that specific pulse positions (cun, guan, chi) correspond to distinct organs, while pulse morphology characteristics (floating, deep, slippery, choppy) reflect deviations in organ function. This aligns methodologically with modern medical frequency-domain analysis of pulse waves, which decomposes pulse signals into a fundamental frequency and its harmonic components.

Taiwanese biomechanics expert Wang Weigong proposed that the first ten harmonic components of the pulse wave (C1 through C10) correspond to the resonant frequencies of different organs: C1 corresponds to the heart (Heart Meridian), C2 to the liver (Liver Meridian), C3 to the spleen (Spleen Meridian), and so on. If this “harmonic-organ mapping” hypothesis holds true, pulse harmonic analysis could serve as a non-invasive, real-time assessment tool for organ function.

However, Wang Weigong’s original research relied primarily on animal experiments and limited clinical samples, leaving significant validation space under contemporary bio-signal processing standards. This paper systematically reviews the scientific evidence regarding pulse harmonics and organ mapping, evaluates the current status of machine learning re-validation studies, and proposes a standardized validation framework.

Signal Processing Foundations of Pulse Harmonics

Physiological Basis for Frequency Domain Analysis

The pulse wave is the result of pressure waves generated by the periodic contraction of the left ventricle propagating through the arterial system. Each cardiac cycle, the waveform recorded by an arterial pressure sensor is not a single-frequency sine wave but a multi-frequency composite signal composed of a fundamental frequency (heart rate, typically 1–2 Hz) and its integer multiple harmonics.

Through Fast Fourier Transform (FFT), the time-domain pulse wave signal can be decomposed as follows:

$$P(t) = \sum_{n=1}^{10} A_n \sin(2\pi n f_0 t + \phi_n)$$

Where $f_0$ is the heart rate fundamental frequency, $A_n$ is the amplitude of the $n$-th harmonic, and $\phi_n$ is the phase. The ratio of harmonic amplitudes (Harmonic Ratio Index, HRI) reflects the cardiovascular system’s selective amplification or attenuation of specific frequency components.

Why Focus on C1-C10?

The energy of the pulse wave is concentrated primarily in the first ten harmonic components (C1-C10). High-frequency harmonics (C7-C10) correspond to distal characteristics of the vascular tree (arteriosclerosis, peripheral resistance), while low-frequency harmonics (C1-C3) are mainly influenced by cardiac contractility and large artery compliance. The five zang-fu organs of Traditional Chinese Medicine are distributed at different anatomical depths within the thoracic and abdominal cavities, corresponding physically to different segments of the arterial tree. This provides a biomechanical logical starting point for the “harmonic-organ” hypothesis.

Validation Applications of Machine Learning in Pulse Harmonic Mapping

Existing Evidence: Multi-Disease Classification via Pulse Spectral Analysis

In recent years, numerous studies have combined pulse wave frequency-domain features with machine learning for disease classification and functional assessment. Some of these studies implicitly validate the mapping between harmonics and physiological states:

PMID 33903610 (Lin SK et al., Sci Rep 2021): Radial blood pressure waves (radial BPW) and photoplethysmographic (PPG) signals from the fingertip were collected. Spectral features were extracted and input into a Multi-Layer Perceptron (MLP) neural network to distinguish patients with Alzheimer’s disease from normal controls. Results indicated that pulse spectral features could effectively distinguish patients with cognitive dysfunction (classification accuracy AUC > 0.85), suggesting a detectable association between pulse harmonic distribution patterns and cerebral hemodynamic status—a cross-disciplinary resonance with the TCM theory of “Heart governs the Spirit.”

PMID 36294440 (Chen CC et al., J Clin Med 2022): Combining non-invasive pulse wave measurement with machine learning, the study analyzed pulse waveform variability among subjects experiencing different side effects of the Pfizer-BioNTech COVID-19 vaccine (BNT162b2). Results found significant differences in pulse harmonic features across different cardiovascular side-effect groups, validating the sensitivity of pulse spectral features to minor perturbations in cardiovascular function.

PMID 42305919 (Lin CN et al., Front Physiol 2026): An interpretable machine learning model based on Photoplethysmographic Pulse Waves (PPG) was developed for non-invasive systolic blood pressure assessment. The model utilized time-domain and frequency-domain features of the PPG waveform; feature importance analysis revealed that lower-order harmonics (C1-C3) had the highest contribution weight for blood pressure prediction, consistent with physiological expectations regarding large artery elasticity.

PMID 36509825 (Wu LW et al., Sci Rep 2022): Combining pulse wave frequency-domain analysis with machine learning, the study distinguished between sarcopenia and metabolic syndrome patients. Research showed systematic differences in pulse harmonic energy distribution under different pathological states, suggesting that pulse spectral features can serve as a “fingerprint” for multi-organ functional status.

PMID 35161551 (Hsiu H et al., Sensors 2022): Radial arterial blood pressure wave signals were collected, and spectral indices were extracted for machine learning analysis to distinguish different cognitive function states among community subjects. Results proved a statistically significant mapping relationship between pulse harmonic features and cognitive function, indirectly supporting the hypothesis that “pulse spectra carry multi-system information.”

From Disease Classification to Organ Mapping: Critical Gaps

The common limitation of the above studies is that they validate an “association between harmonic features and a specific disease/state,” but do not directly verify that a specific harmonic corresponds to a specific organ. From a machine learning perspective, this is akin to training only a classifier mapping “overall pulse spectrum → disease label,” without verifying whether independent variations in a specific harmonic reflect independent functional changes in the corresponding organ.

Wang Weigong’s C1-C10 hypothesis requires a one-to-one correspondence between specific harmonics and specific organs. To verify this hypothesis, the following are needed:
1. Simultaneous collection of multi-organ functional indicators (gold standard).
2. Verification of independent correlation between specific harmonic components and corresponding organ indicators (controlling for other variables).
3. Replication of this mapping relationship in independent cohorts.

Scientific Hypothesis and Validation Framework for C1-C10 Organ Mapping

Hypothesis Content

Harmonic Estimated Frequency Range TCM Correspondence Physiological Correspondence
C1 ~1-2 Hz Heart Cardiac Contractility
C2 ~2-4 Hz Liver Hepatic Blood Flow/Portal System
C3 ~3-6 Hz Spleen Digestive Blood Allocation
C4 ~4-8 Hz Lung Pulmonary Circulation/Gas Exchange
C5 ~5-10 Hz Kidney Renal Blood Flow/Water Metabolism
C6-C10 ~6-20 Hz Six Fu Organs etc. Distal Vascular Resistance/Arteriosclerosis

Note: Actual frequency ranges vary due to individual heart rate differences; values in the table are estimates based on a 60 BPM heart rate.

Methodological Requirements for Validation

Standardized Signal Acquisition: Use synchronized tri-lead radial arterial pressure waves (radial APW) + fingertip PPG, with sampling rates ≥200 Hz and duration ≥3 minutes/sample to ensure frequency domain resolution ≤0.1 Hz.

Gold Standard Organ Functional Indicators: Simultaneous collection of gold-standard indicators for each organ function (e.g., echocardiography for cardiac output, liver function enzyme profiles, renal function creatinine/BUN, lung function FEV1), treating continuous variables as regression targets rather than performing binary classification of disease presence/absence.

Statistical Methods: Multi-variable linear regression or Gradient Boosting Trees (XGBoost), using the HRI vector (amplitude ratios of C1-C10) as feature vectors and each organ functional indicator as the dependent variable, to test the independent explanatory power (partial $R^2$) of each harmonic for a specific organ.

Cross-Validation: Hold-out method + 5-fold cross-validation to prevent overfitting; report confidence intervals for prediction errors.

Evidence Level and Limitations

The overall evidence level for this topic is Level B (observational study support, preliminary mechanistic research exists, but lacks strictly controlled multi-center verification).

Supporting Evidence: Multiple independent studies demonstrate that pulse harmonic features can distinguish different physiological/pathological states, and this distinction possesses statistical significance and biological plausibility, indirectly supporting the hypothesis that harmonics carry multi-dimensional organ information.

Key Limitations:
1. One-to-One Correspondence Not Verified: Existing studies follow a “whole harmonic spectrum → single disease type” classification mode, failing to test the direct correspondence relationship of “independent variation of harmonic $n$ → independent function of organ $n$.”
2. Complexity of TCM Organs: TCM “organs” are functional collection concepts and do not map simply to modern medical anatomical organs; the “Spleen” is not equivalent to the anatomical spleen, and the “Liver” is not equivalent to the liver alone—this increases operational difficulty in verification.
3. Individual Differences and Standardization Challenges: Harmonic baseline values vary significantly among subjects of different ages, genders, and body types; establishing a universal baseline atlas requires large-sample, multi-center data.
4. Insufficient Standardization in Signal Acquisition: Existing studies utilize diverse equipment with non-uniform sampling parameters, limiting cross-study comparison.

Further Reading

  • Wang WQ. The Research of Blood Flow Frequency Distribution and Resonance Theory of Internal Organs. Taiwan University Press, 1984. (Original literature on Wang Weigong’s organ resonance theory)
  • Lin SK et al. (2021) Classification of patients with Alzheimer’s disease using the arterial pulse spectrum and a multilayer-perceptron analysis. Sci Rep 11:8882. PMID: 33903610
  • Hsiu H et al. (2022) Discrimination of the Cognitive Function of Community Subjects Using the Arterial Pulse Spectrum and Machine-Learning Analysis. Sensors 22(3):806. PMID: 35161551
  • Wu LW et al. (2022) Discrimination between possible sarcopenia and metabolic syndrome using the arterial pulse spectrum and machine-learning analysis. Sci Rep 12:21452. PMID: 36509825

This article belongs to Pool ③ (Bio-frequency Codes) Exploration Series | Project #6 C1-C10 Organ Harmonic Mapping Machine Learning Re-validation | Evidence Level: B | YY (Analyst) 2026-10-03


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 *