Blog

  • 静息与练功之间那道缝,是越练越窄还是越练越宽:两态差值的年限斜率裁决

    Evidence level: C

    单看”修炼者静息时更腹式”证明不了什么——那可能只是他躺上测量床就自动开始用功。真正能分辨”练进默认”和”留着一手技巧”的,是同一批人身上两个状态的差值,随修炼年限往哪个方向走。

    一、把问题从”水平”改成”斜率”

    丹田呼吸的长期效果,常被表述成一句水平比较:老手静息时的腹区占比(%AB)是否高于新手、胸腹相位角(φ)是否低于新手。但这里有个绕不开的漏洞——人知道自己正在被测量。呼吸是少数”意识一秒就能接管”的自主系统,一个练了十年的人躺下,很可能不自觉就把功架摆出来。这样测到的”静息”,其实是”隐性练功”。

    所以干净的量不是水平,是差值。记练功态 P(有指令、刻意做),静息态 R(无呼吸提示),定义 Δ = P − R(%AB 维)与 Δφ = φ(P) − φ(R)。Δ 衡量”这个人要靠多少刻意努力,才到得了他的最好状态”。功夫越深,距离按理越短;若只是一个随时能调用的技巧,距离会一直保留,甚至越练越大——他能把 P 推得更极致,而 R 纹丝不动。

    二、两个假说给出相反的符号,这是全部价值所在

    内化默认假说:长期练习把膈肌主导重写成呼吸中枢的默认输出,R 随年限向 P 靠拢、P 基本不变,Δ 单调收窄趋零。机制不空想——瑜伽老手静息潮气量确实更高(0.7 vs 0.5 L,PMID 27055287),哈他瑜伽资深者静息 5.5 次/分、1.03 L,对照 13.4 次/分、0.56 L(PMID 6798002),”不练的时候也不一样”有依据。

    可调用技巧假说:练习提升的是”能调多高”,不是”平时多高”。P 随年限继续远离 R,R 不动,Δ 反而放大。

    两个假说在水平比较上完全无法区分,但在 Δ 的斜率符号上互斥。这就是这道题值得做的原因。

    三、先排掉三个会伪造斜率的坑

    回归均值。 Δ 由两个带误差的量相减,R 又在负号位上,Δ 天然与 R 负相关;只要年限与基线 R 有关,Δ 的年限斜率就会凭空冒出。正确处理是把 P 对 R 做基线校正后再回归年限(PMID 15333621)。

    测量反应性。 只要在测,注意力就落在呼吸上;注意呼吸与刻意控制呼吸动用两套不同但重叠的脑网络(PMID 28954895),对呼吸的内感受注意还会调制脑-心耦合(PMID 35964864)。”R 态”从来不是纯净默认态。解法是加第三态:无提示 + 认知分心任务下的呼吸,那才接近无人看管时的输出。

    共同驱动。 年龄、体能、体重会同时推高 P 和 R,此时 Δ 可能完全不变而两态都在漂——最容易被误读成”没效果”。必须把”Δ 不变 + 两态同向漂移”单列为一种独立结局。

    四、测量侧:φ 的噪声底必须先量出来

    相位角的个体内变异极大——早产儿安静睡眠中 φ 的个体内变异系数达 11%–123%,且作者明确指出这部分变异来自胸腹相位关系本身的不稳,不是记录或算法造成的(PMID 24716708)。算法选择则已清楚:互相关与最大线性相关最稳健,10% 噪声下误差小于 5 度,传统 Lissajous 环分析误差可超 30 度(PMID 12422344)。胸腹分区贡献在成人运动中重测信度很高(ICC 0.87–0.93,PMID 35226168),但那是运动态;2026 年范围综述明确,绝对容积一致性随校准与体位变化,信度指标报告极不统一(PMID 42862511)。

    没算出最小可检测变化(MDC)之前,任何 Δ 的年限斜率都不能解释——否则测到的可能全是噪声。

    五、短训证据其实站在”技巧”那一边

    值得正视的反对证据:呼吸模式学习实验中,只有带反馈训练的一组在 24 小时后仍保持成绩(PMID 3733594);八个训练 session 之后,呼吸控制依然需要注意力(PMID 2062903)。在周/月尺度上并无自动化证据。另一端,两个月瑜伽呼吸干预确实提高了静息腹区贡献(PMID 34441810),但那是组间水平变化,不是”缝变窄”。

    六、可执行设计

    四格设计:年限分层(新手 / 1–5 年 / 5–15 年 / 15 年以上)× 两态(P / R),同批受试者、同体位同倾角、顺序随机平衡,先做肺功能筛查排除气流阻塞,φ 一律用互相关法。每人隔日重复三次,先算 MDC 与 ICC,再算 Δ 的年限斜率。判据:

    • Δ 显著收窄且 R 向 P 单向靠拢 → 内化默认成立
    • Δ 显著放大且 R 不变 → 可调用技巧成立
    • Δ 不变、两态同向漂移 → 第三方共同驱动,需换协变量
    • Δ 先放大后收窄 → 技能先精进、后沉淀的两段式

    L1 网络状态:该主题尚未进入 L1 对应网络——CORRESPONDENCE_NETWORK.md 中丹田、气功、胸腹相位角均无对应边,仅把「呼吸率 / 呼吸模式」列为可穿戴观测项、「胃肠蠕动」列为「气沉丹田」的候选物;EDGE_TABLE_science-yixue.md 仅有「丹田 51 篇 / 气功 48 篇」的文章计数,无该边状态判定。故本文证据等级封顶 C。

    证据与局限

    已验证:胸腹相位角可用互相关法稳健测量(PMID 12422344);胸腹分区贡献在成人中重测信度高(PMID 35226168);瑜伽长期练习者静息潮气量显著高于对照(PMID 27055287、6798002);回归均值可由基线校正处理(PMID 15333621)。

    部分验证:刻意呼吸与注意呼吸动用不同脑网络(PMID 28954895);两个月呼吸干预可提高静息腹区贡献(PMID 34441810)。

    待验证:Δ 的年限斜率方向本身——全世界尚无直接研究;呼吸技能在年尺度上是否自动化,现有证据相反(PMID 3733594、2062903);Δ 在第三态下的形态完全未知。

    相关阅读

    证据等级: C
    参考文献: PMID 15333621, PMID 24716708, PMID 35226168, PMID 42862511, PMID 27055287, PMID 6798002, PMID 34441810, PMID 28954895, PMID 3733594, PMID 2062903, PMID 35964864, PMID 12422344


    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。

    The Gap Between Rest and Practice: Is It Narrowing or Widening Over Time? A Judgment Based on the Annual Slope of State Differences

    Evidence level: C

    Observing that “practitioners exhibit more abdominal breathing at rest” does not prove anything—it may simply indicate they begin practicing automatically upon lying on the measurement bed. The only way to distinguish between “defaulting into practice” and “retaining a reserve of technique” is to observe how the differential between two states on the same individuals shifts over years of practice.

    I. Reframing the Question from “Level” to “Slope”

    The long-term effects of Dantian breathing are often described as a horizontal comparison: do experts have a higher %AB at rest compared to novices, or a lower thoraco-abdominal phase angle ($\phi$)? However, there is an unavoidable flaw here—subjects know they are being measured. Respiration is one of the few autonomic systems that can be consciously intercepted within a second; a practitioner with ten years of training likely unconsciously sets up their posture when lying down. Thus, the measured “rest” is actually “implicit practice.”

    Therefore, the clean metric is not the level, but the differential. Let $P$ denote the practice state (with instruction, deliberate effort) and $R$ denote the rest state (no breathing cue). Define $\Delta = P – R$ (%AB dimension) and $\Delta\phi = \phi(P) – \phi(R)$. $\Delta$ measures “how much deliberate effort this person requires to reach their best state.” According to theory, as skill deepens, the distance should shorten; if it is merely a retrievable technique, the distance remains constant or even widens with practice—they can push $P$ to extremes while $R$ remains unchanged.

    II. Two Hypotheses Yield Opposite Signs, Which Is the Entire Value

    Internalization Default Hypothesis: Long-term practice rewrites diaphragm dominance as the central respiratory default output. $R$ converges toward $P$ over years, while $P$ remains stable; $\Delta$ monotonically narrows toward zero. The mechanism is not empty speculation—long-term yoga practitioners indeed have higher resting tidal volumes (0.7 vs. 0.5 L, PMID 27055287), and senior Hatha yogis show resting rates of 5.5 bpm/1.03 L compared to novices at 13.4 bpm/0.56 L (PMID 6798002); “not being different even when not practicing” has empirical basis.

    Retrievable Technique Hypothesis: Practice improves “how high one can reach,” not “how high one is normally.” $P$ continues to diverge from $R$ over years, $R$ remains static, and $\Delta$ amplifies.

    These two hypotheses are indistinguishable in horizontal comparisons but are mutually exclusive regarding the sign of $\Delta$’s slope. This is why this question is worth investigating.

    III. Eliminating Three Pitfalls That Fabricate Slopes

    Regression to the Mean. $\Delta$ is derived by subtracting two noisy quantities, with $R$ in the negative position; $\Delta$ is naturally negatively correlated with $R$. If years of practice correlate with baseline $R$, a spurious slope for $\Delta$ will emerge. The correct approach is to baseline-correct $P$ against $R$ before regressing against years (PMID 15333621).

    Measurement Reactivity. As long as one is being measured, attention focuses on breathing; attention to breathing and deliberate control of breathing engage two distinct but overlapping brain networks (PMID 28954895), and interoceptive attention to breathing also modulates brain-heart coupling (PMID 35964864). The “R state” is never a pure default state. The solution is to add a third state: breathing under no cue + cognitive distraction, which approximates output when unobserved.

    Common Drivers. Age, fitness, and weight can simultaneously elevate both $P$ and $R$, potentially leaving $\Delta$ unchanged while both states drift—a scenario most easily misread as “no effect.” This must be listed as a distinct outcome: “$\Delta$ unchanged + bidirectional drift of both states.”

    IV. Measurement Side: The Noise Floor for $\phi$ Must Be Quantified First

    Intra-individual variability in phase angle is extremely high—coefficients of variation for $\phi$ during quiet sleep in preterm infants range from 11% to 123%, with authors explicitly stating this variability stems from instability in the thoraco-abdominal relationship itself, not recording or algorithmic error (PMID 24716708). Algorithm selection is clear: cross-correlation and maximum linear correlation are most robust, with errors under 5 degrees at 10% noise, whereas traditional Lissajous loop analysis can exceed 30 degrees (PMID 12422344). Thoraco-abdominal partition contributions show high retest reliability in adult exercise (ICC 0.87–0.93, PMID 35226168), but that is an exercise state; a 2026 systematic review confirms absolute volume consistency varies with calibration and posture, and reliability metrics are reported inconsistently (PMID 42862511).

    Without calculating the Minimum Detectable Change (MDC), no slope for $\Delta$ over years can be interpreted—otherwise, one might only be measuring noise.

    V. Short-Term Evidence Actually Supports the “Retrievable Technique” Side

    Notable counter-evidence: In breathing pattern learning experiments, only the group with feedback training maintained performance after 24 hours (PMID 3733594); after eight training sessions, breath control still requires attention (PMID 2062903). No evidence of automation exists at weekly or monthly scales. Conversely, a two-month yoga breathing intervention did increase resting abdominal contribution (PMID 34441810), but this represents between-group level changes, not a “narrowing of the gap.”

    VI. Executable Design

    Four-cell design: Stratify by years (novice / 1–5 years / 5–15 years / >15 years) × two states ($P$ / $R$), same cohort, same posture and inclination, randomized balanced order, screen for airflow obstruction via pulmonary function tests first, use cross-correlation for all $\phi$. Repeat three times per person on alternate days. First calculate MDC and ICC, then the slope of $\Delta$ over years. Criteria:

    • Significant narrowing of $\Delta$ with $R$ converging unidirectionally toward $P$ → Internalization default confirmed
    • Significant amplification of $\Delta$ with $R$ unchanged → Retrievable technique confirmed
    • Unchanged $\Delta$ with bidirectional drift of both states → Third-party common driver requires new covariates
    • Amplification followed by narrowing → Two-phase trajectory: skill refinement followed by sedimentation

    L1 Network Status: This topic has not yet entered the L1 corresponding network—CORRESPONDENCE_NETWORK.md lists no edges for Dantian, Qigong, or thoraco-abdominal phase angle; only “breathing rate / pattern” is listed as a wearable observation item and “intestinal peristalsis” as a candidate for “Qi sinking to Dantian.” EDGE_TABLE_science-yixue.md contains only article counts (“51 articles on Dantian,” “48 on Qigong”), with no state determination for this edge. Thus, the evidence level for this paper is capped at C.

    Evidence and Limitations

    Verified: Thoraco-abdominal phase angle can be robustly measured via cross-correlation (PMID 12422344); thoraco-abdominal partition contributions show high retest reliability in adults (PMID 35226168); long-term yoga practitioners have significantly higher resting tidal volumes than controls (PMID 27055287, 6798002); regression to the mean can be handled via baseline correction (PMID 15333621).

    Partially Verified: Deliberate breathing and attention to breathing engage different brain networks (PMID 28954895); a two-month breathing intervention can increase resting abdominal contribution (PMID 34441810).

    To Be Verified: The direction of $\Delta$’s slope over years itself—no direct study exists worldwide; whether breathing skills automate at the annual scale, current evidence suggests otherwise (PMID 3733594, 2062903); the morphology of $\Delta$ under a third state remains completely unknown.

    Related Reading

    Evidence Level: C
    References: PMID 15333621, PMID 24716708, PMID 35226168, PMID 42862511, PMID 27055287, PMID 6798002, PMID 34441810, PMID 28954895, PMID 3733594, PMID 2062903, PMID 35964864, PMID 12422344


    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.

  • σ 区间跨档审计:31 行里没有一行是无条件安全的

    Evidence level: C

    池④ · 关联 #51(P1)|配套治理卡 08-ops/experiments/pool4-sigma-interval-tier-governance.md(P4-AUDIT-006)
    证据等级 C|状态 draft|本轮对既有正文写入 = 0

    0. 一句话结论

    上一轮(#46a)手工判出「钢筋混凝土」「黏土/页岩」两行跨档,并说这是 #37 漏掉的一整类问题。

    本轮把这一类工具化(sigma-interval-tier-audit.py,自检 15/15 通过)后全库扫描,结论比预期硬得多:

    本库 31 个带 σ 区间的表格行里,没有任何一行的区间比窄到「无论厚度多少都必然同档」。

    原因是有一个恒等式:A_max/A_min = √(σmax/σmin),与厚度 z、频率 f 都无关。所以「跨不跨档」只由区间比 q 与 z 共同决定;而 q ≤ (13/7)² = 3.449 时才算无条件安全。实测 31 行里 q 最小值就是 10 倍 —— 一行都没进安全区。


    1. 为什么这个问题必须用恒等式判,不能靠逐行试算

    趋肤深度与衰减:

    δ(σ) = 1/√(π f μ σ)              A_dB(z) = 8.686 · z · √(π f μ σ)
    

    两式相乘约掉 σ 的平方根关系,得到本文唯一真正需要的东西:

    A_max / A_min = √(σmax / σmin) = √q          ← 与 z、f 均无关
    

    第一个推论:一个 σ 区间的「A 展开倍数」是区间自身的常数。σ 跨 100 倍 → A 跨 10 倍;跨 10 倍 → A 跨 3.16 倍。

    第二个推论(关键):判定档 K2 带只有 7→13 dB,宽 1.857 倍。要让 A 的两端必落同一档,必须 √q ≤ 1.857,即

    q ≤ (13/7)² = 3.449        ← 无条件安全线
    

    q 超过 3.449,就一定存在某个厚度使 A 从 K3 穿到 K1。这不代表「这行一定错」,而代表「这行不能在不锁 z 的前提下给单一档位结论」。

    这把「区间跨档」从逐行试算变成了一个可批量的判据 —— 而且判据本身不含 z,所以连 z 都不知道的行也能判。


    2. 全库扫描结果:925 文件 / 117 行 / 31 区间行

    指标 数值
    扫描文件 925(03-sites/cn/topics/science-yixue + 08-ops/experiments)
    表格总数 866
    其中含 σ 列 31
    σ 数值行 117
    其中区间行 31
    区间行 q ≤ 3.449(无条件安全) 0
    区间行 q > 3.449(存在跨档 z) 31
    有 z 且实测跨档 5
    有 z 且实测同档 2
    无 z(彻底欠定) 24

    四层分级(本轮的实质产出):

    层 含义 行数 处置
    L0 无条件安全(√q ≤ 1.857,任意 z 同档) 0 —
    L1 条件安全(√q > 1.857 但锁定的 z 下两端同档) 2 保留,必须写出 z
    L2 跨档(锁定的 z 下两端不同档) 5 禁止唯一结论,改双列或加条件
    L3 欠定(连 z 都没写) 24 禁止任何档位结论,先补 z

    3. L2:5 行已确证的跨档行

    文件 行 σ 区间 q z A 两端 档位跳动
    mountain-sample-priority-sigma-pre-screening-sop.md 湿花岗岩(含水/含矿) 10⁻³–10⁻¹ 100 1 km 1.53→15.27 dB K3→K1
    同上 湿玄武岩 10⁻²–10⁰ 100 1 km 4.83→48.29 dB K3→K1
    同上 B 档(条件性测) 10⁻³–10⁻¹ 100 1 km 1.53→15.27 dB K3→K1
    schumann-station-candidate-tier-a-wet-basalt.md B 档 10⁻³–10⁻¹ 100 1 km 1.53→15.27 dB K3→K1
    elf-skin-depth-table-numeric-integrity-audit.md A′ 10⁻²–10⁻¹ 10 1 km 4.83→15.27 dB K3→K1

    最刺眼的是「B 档(条件性测)」:#22 SOP 给它写的是「1–10 dB、中、验证 H0/H1 边界」。但按它自己声明的 σ 区间 10⁻³–10⁻¹、1 km 复算,下端 1.53 dB、上端 15.27 dB——上端已经越过 K1 门槛,不是「中」而是「可测」。这行的建议(”A 档测完仍有预算再上”)建立在区间下端,而它的上端已经够格排在 A 档。

    反过来看,「A′」行(10⁻²–10⁻¹,标注「中(需 ≥2 km 山体或降低门槛到 5 dB)」)的下端 4.83 dB 其实是偏低的——它的建议方向没错,但同样只看了区间一端。

    共同点:这 5 行都在用区间的一端代表整行。 而 q=100 意味着另一端差 10 倍。


    4. L3:24 行的 z 根本没写,连跨不跨档都无法判

    24 行分布在 building-mountain-terrain-elf-attenuation-analogy.md(6)、schumann-station-sigma-literature-values.md(6)、skin-depth-reference-table.md §二(5)、pool4-sigma-fmeas-directional-judgment-sop.md(5)、schumann-station-sigma-database.md(2)。

    这些行只给了「介质 / σ / μr」,没有厚度列、没有 A(dB) 值。工具对它们的处理是给「歧义 z 带」而不是判档:

    以 σ 10⁻³–10⁻¹(q=100,√q=10)为例,7.83 Hz:

    门槛 歧义 z 带 读法
    7 dB(K3→K2) 458 m – 4.58 km z 小于 458 m,两端都在 K3;z 大于 4.58 km,两端都在 K2 以上
    13 dB(K2→K1) 851 m – 8.51 km 落在带内 = 无法判档

    这条带最宽 10 倍。 「1 km 山体」正好落在 7 dB 与 13 dB 两条带的交点区域——这就是为什么同一个「1 km、湿花岗岩」在不同文件里能被写成「透明」也能被写成「数十 dB」。

    这不是文件互相矛盾,是每一份都只取了区间的一个端点。


    5. 附带发现:#22 / #25 两张表的 δ 列系统性偏大 37.25 倍

    扫描中顺带核对 δ 列时发现一处不属区间问题、但更基础的事实。

    mountain-sample-priority-sigma-pre-screening-sop.md §一 与 schumann-station-candidate-tier-a-wet-basalt.md §一 共用一张三档表,其 δ 列声明:

    行 σ 区间 δ 声明 7.83 Hz 实算 判读
    普通花岗岩(干) 10⁻⁸–10⁻⁴ 数百米–数千米 18.0 km – 1800 km 数量级不符(差 2–3 个量级)
    湿花岗岩 10⁻³–10⁻¹ 几十米–几百米 0.569 km – 5.69 km 数量级不符(差 1–2 个量级)
    湿玄武岩 10⁻²–10⁰ 几米–几十米 0.180 km – 1.80 km 数量级不符(差 1–2 个量级)

    引用一张表的具体数字比逐行比对更硬:schumann-station-candidate-tier-a-wet-basalt.md 明写「A 档 σ ≥ 10⁻² → δ ≤ 67 km」。而 σ=10⁻² 在 7.83 Hz 的 δ 是 1.7986 km——

    67 km / 1.7986 km = 37.2508
    

    差 37.25 倍。 而 mountain-sample-priority-sigma-pre-screening-sop.md §二 的 A 档判据写「σ ≥ 10⁻² → ≥ 10 dB」,B 档写「δ 几十米–几百米」,两文件用的是同一套数字,37.25 这个比值在两文件上完全一致。

    反推:要让 σ=10⁻² 给出 67 km,等效频率必须是 0.00564 Hz,不是 7.83 Hz。

    这说明这两张表的 δ 列不是按 7.83 Hz 算的,而是从同一个上游数值整段继承来的(否则不会在两文件撞出同一个 37.25)。

    而这个错误 #37 扫描器抓不到:我对这两个文件单独跑 delta-table-scan.py,结果是「命中文件 0|断言总数 0」——完全没解析进去。原因是它的表头感知规则没把「δ(7.83 Hz)」这一列认成 δ 列,于是整张表被跳过。

    这与第 9 轮的教训同型:工具跑通了,但指标没覆盖到目标。区别是这次是覆盖率漏了,不是指标选错了。

    必须分清:A(dB) 列不是算错,是区间覆盖不全。

    行 σ 区间两端实算 A 表里写的 A 覆盖情况
    湿花岗岩(#22 §一) 1.53 – 15.27 dB 「1–10 dB 量级」 下端近似对,上端漏了 5.27 dB(越 K1)
    湿玄武岩(#22 §一) 4.83 – 48.29 dB 「数十 dB 量级」 上端对,下端 4.83 dB 不是「数十」

    所以不能说「整表错」——δ 列错、A 列对但覆盖不全,两类问题互相独立。


    6. 对本池既有产出的影响

    既有产出 影响
    #16 skin-depth-reference-table.md §三 δ 表 不在 37.25 倍问题内(该表 δ 与公式一致,实测 σ=10⁻⁴ → 17.99 km 对得上)。但 §二 参数表被工具认出 5 个区间行,q=10 或 100,属 L3 欠定(该表本来就不给 z,按参数表口径处置,见治理卡 §4)
    #22 mountain-sample-priority-sigma-pre-screening-sop.md δ 列需重算(37.25× 偏差);A 列可用;三档判定(A/B/C)建立在 A 列上,故分档结论本身不动,但「1–10 dB」「数十 dB」的区间写法应改
    #25 schumann-station-candidate-tier-a-wet-basalt.md 同上:δ 列需重算,「δ ≤ 67 km」应改 1.8 km;A 档判据(σ ≥ 10⁻²)不受影响
    #37 全量扫描(59 格 MISMATCH) 结论不变(那是表内自洽性);但需明确它的覆盖率不含这两文件,不能拿「0 命中」当「无问题」
    #46a / P4-AUDIT-005(f_meas 方向性) 本轮与之正交:f_meas 决定「σ 锚在哪个频率」,q 决定「区间内能差多少」。两者可叠加:真实不确定度 ≈ √q × f_meas 因子
    #14 H1a 门槛(≤3 dB) 维持 P4-AUDIT-005 的判断:反对下调

    7. 证据与局限

    已有证据(可核):

    • A_max/A_min = √(σmax/σmin) 与 q ≤ (13/7)² = 3.449 无条件安全线,均为解析恒等式,不依赖任何文献假设;
    • 全库扫描数字(925 文件 / 866 表 / 31 含 σ 表 / 117 σ 行 / 31 区间行 / 5 跨档 / 24 无 z)可由 sigma-interval-tier-audit.py scan 逐条复现,JSON 可导出;
    • 工具自检 15/15 通过,含「跨档判定」「同档判定」「上标科学计数法解析」「区间分隔符与上标负号区分」四类边界用例;
    • δ 列 37.25 倍偏差是对两个文件逐字符复算,且两处在不同 σ 上撞出同一比值,属文件层面可确证;
    • 岩石 σ 强频率依赖有 5 篇独立文献支撑(Mehran 1977、Jonscher 1977、Revil 2013、Laštovičková 1992、Hall 2004),说明「区间」不是纸面虚构而是真实的物理不确定度;
    • 3 个 PMID 经 E-utilities 回查、5 个 DOI 经 Crossref 回查,题名/首作者/期刊/年份全部一致。

    局限(必须与结论同时读):

    1. 37.25 倍偏差的成因本文不给结论。只确证「等效频率 0.00564 Hz」与「两处比值一致」,不猜测上游是手算失误、单位混淆还是抄自另一频率的表;不排除是该表作者有意使用了一个低频口径(但表头明写 7.83 Hz,故按表头判为不符);
    2. 工具只认 markdown 表格。正文里的裸数值断言(如「某山体衰减 3 dB」)不在覆盖范围内,与第 9 轮裸列表问题同型;
    3. 117 / 31 这两个数字是「被工具认出的」,不是「库里真实存在的」。未认出 ≠ 不存在;本次已因表头规则少认两张关键表(见 §5),故实际 σ 行数应 ≥ 117;
    4. 判定档门槛 13/7 dB 是池④ 自定,不是行业标准。若门槛改,3.449 这条线随之变(q ≤ (T_K1/T_K2)²);本文不主张该门槛经外部验证;
    5. 良导体近似、无限厚板、垂直入射三重简化仍在,A(dB) 绝对值不主张任何场地真值;
    6. q > 3.449 不等于「这行错了」。它等于「这行不能在不锁 z 的前提下给单一档位结论」——这是欠定,不是错误。

    8. 给下一轮的动作建议

    优先级 动作 依赖
    P1 L3 的 24 行补 z 列——这是解锁全部判定的唯一前置,纯本地 0 成本 无
    P1 L2 的 5 行改双列(区间下端 / 上端各一列 + 档位),或加「本行限于 σ 取 X 时」条件句 需 JOE 授权改既有正文
    P1 #22/#25 δ 列重算(37.25× 偏差)——纯算术,但改的是已发布文章正文,需 JOE 授权 需 JOE
    P2 delta-table-scan.py 补表头规则,覆盖「δ(7.83 Hz)」这类带单位的列名 本地可做
    P2 把 q ≤ (T₁/T₂)² 无条件安全线写进 P4-AUDIT-004,作为新写表的准入检查 本地可做
    P3 反向扫池③ OMOP 三轴归因表与池⑤ F1–F5 门槛是否也有同型欠定 待 P1 完成

    本文为 draft。 全部为解析复算 + 本库文件文本解析,无实地测量。
    不构成对任何具体山体/场地 ELF 衰减的真值声明。
    进入 779 终版前需走 intake;进入 782 前需去 AI 化。
    「31 行里 0 行无条件安全」是本轮最硬的数字,也是最不该被包装成「已修复」的数字——它只是把欠定说清楚了,一行都还没改。
    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。

    Sigma Interval Spanning Audit: No Row in the 31-Row Query Is Unconditionally Safe

    Evidence level: C

    Pool④ · Related #51 (P1) | Supporting Governance Card 08-ops/experiments/pool4-sigma-interval-tier-governance.md (P4-AUDIT-006)
    Evidence Level C | Status draft | Write count for existing body in this round = 0

    0. One-Sentence Conclusion

    In the previous iteration (#46a), manual analysis identified two rows spanning category boundaries (“reinforced concrete,” “clay/shale”) and flagged this as a whole new category of issues missed in #37.

    This iteration toolized the audit using sigma-interval-tier-audit.py (all 15 self-checks passed) and scanned the entire database. The conclusion is far more stringent than anticipated:

    Among the 31 table rows in this library containing σ intervals, not a single row has an interval narrow enough to may support categorization regardless of thickness.

    This stems from the identity: A_max/A_min = √(σmax/σmin), which is independent of thickness z or frequency f. Therefore, whether a row spans categories depends solely on the interval ratio $q$ and thickness $z$. Unconditional safety only occurs when $q \le (13/7)^2 \approx 3.449$. In the empirical data of 31 rows, the minimum $q$ is 10 times larger — no row enters the safe zone.

    1. Why an Identity Criterion is Required Rather Than Row-by-Row Trial Calculation

    Skin Depth and Attenuation:

    δ(σ) = 1/√(π f μ σ)              A_dB(z) = 8.686 · z · √(π f μ σ)
    

    Multiplying the two equations eliminates the square root relationship involving $\sigma$, yielding the sole necessary factor for this study:

    A_max / A_min = √(σmax / σmin) = √q          ← Independent of z and f
    

    First Inference: The “A expansion ratio” for a $\sigma$ interval is a constant intrinsic to that interval. A $\sigma$ span of 100-fold results in an A span of 10-fold; a 10-fold span results in an A span of 3.16-fold.

    Second Inference (Critical): The K2 band for judgment grade only spans 7→13 dB, a width of 1.857-fold. To ensure both ends of A fall within the same grade, it is required that $\sqrt{q} \le 1.857$, implying

    q ≤ (13/7)² = 3.449        ← Unconditional safety threshold
    

    If $q$ exceeds 3.449, there necessarily exists a specific thickness where A transitions from K3 to K1. This does not imply “this row is definitely incorrect,” but rather that “this row cannot yield a single-grade conclusion without locking $z$.”

    This transforms the “interval crossing grades” issue from a row-by-row trial calculation into a scalable criterion — and since the criterion itself contains no $z$, rows where $z$ is unknown can still be evaluated.

    2. Full-Database Scan Results: 925 Files / 117 Rows / 31 Interval Rows

    Metric Value
    Scanned Files 925 (03-sites/cn/topics/science-yixue + 08-ops/experiments)
    Total Tables 866
    Tables with σ column 31
    Rows with σ values 117
    Interval Rows 31
    Interval rows where q ≤ 3.449 (unconditional safety) 0
    Interval rows where q > 3.449 (existence of cross-tier z) 31
    With z and observed cross-tier 5
    With z and observed same tier 2
    No z (completely underdetermined) 24

    Four-Tier Classification (Substantive Output of This Round):

    | Tier | Meaning | Row Count | Action |
    |:–:|:–|–:|
    | L0 | Unconditional safety (√q ≤ 1.857, any z same tier) | 0 | — |
    | L1 | Conditional safety (√q > 1.857 but locked z shows same tier at both ends) | 2 | Retain, must specify z |
    | L2 | Cross-tier (locked z shows different tiers at both ends) | 5 | Prohibit unique conclusion; use dual columns or add conditions |
    | L3 | Underdetermined (no z specified) | 24 | Prohibit any tier conclusion; first supplement with z |


    3. L2: Verified Cross-Range Five Phases

    File Phase $\sigma$ Range $q$ $z$ A Ends Range Jump
    mountain-sample-priority-sigma-pre-screening-sop.md Wet Granite (water/mineral-bearing) $10^{-3}$–$10^{-1}$ 100 1 km 1.53→15.27 dB K3→K1
    Same Wet Basalt $10^{-2}$–$10^{0}$ 100 1 km 4.83→48.29 dB K3→K1
    Same B Phase (Conditional Test) $10^{-3}$–$10^{-1}$ 100 1 km 1.53→15.27 dB K3→K1
    schumann-station-candidate-tier-a-wet-basalt.md B Phase $10^{-3}$–$10^{-1}$ 100 1 km 1.53→15.27 dB K3→K1
    elf-skin-depth-table-numeric-integrity-audit.md A′ $10^{-2}$–$10^{-1}$ 10 1 km 4.83→15.27 dB K3→K1

    The most conspicuous issue is the “B Phase (Conditional Test)” row: #22 SOP describes it as “1–10 dB, medium, verifying H0/H1 boundaries.” However, recalculating based on its own declared $\sigma$ range ($10^{-3}$–$10^{-1}$) and 1 km distance yields a lower bound of 1.53 dB and an upper bound of 15.27 dB. The upper bound already exceeds the K1 threshold; thus, it is not “medium” but rather “measurable.” The recommendation for this row (“Proceed to A Phase only if budget remains after A Phase measurement”) relies on the lower end of the interval, while its upper end qualifies for inclusion in A Phase.

    Conversely, the “A′” row ($10^{-2}$–$10^{-1}$, labeled “medium (requires ≥2 km mountain mass or lower threshold to 5 dB)”) has a lower bound of 4.83 dB, which is relatively low. Its directional advice is correct but similarly addresses only one end of the interval.

    Commonality: All five rows represent the entire range using only one endpoint. Given $q=100$, the other endpoint differs by a factor of 10.

    4. L3: The z-values for the 24 rows are not explicitly stated; even determining whether they cross a threshold is impossible

    The 24 rows are distributed across building-mountain-terrain-elf-attenuation-analogy.md (6), schumann-station-sigma-literature-values.md (6), skin-depth-reference-table.md §2 (5), pool4-sigma-fmeas-directional-judgment-sop.md (5), and schumann-station-sigma-database.md (2).

    These rows provide only “medium / σ / μr”; there is no thickness column and no A(dB) value. The tools process them by assigning a “ambiguous z-band” rather than determining a category:

    For σ 10⁻³–10⁻¹ (q=100, √q=10) at 7.83 Hz:

    Threshold Ambiguous z-band Interpretation
    7 dB (K3→K2) 458 m – 4.58 km If z < 458 m, both ends are in K3; if z > 4.58 km, both ends are above K2
    13 dB (K2→K1) 851 m – 8.51 km Falling within the band = cannot determine category

    This band spans a factor of 10. A “1 km mountain” falls precisely into the intersection region of the 7 dB and 13 dB bands—this explains why the same “1 km, wet granite” can be described as “transparent” in one document and “tens of dB” in another.

    This is not a contradiction between documents; each document only captures one endpoint of the interval.

    5. Incidental Findings: Systematic Overestimation of the δ Column in Tables #22 and #25 by a Factor of 37.25

    During the scan, while cross-verifying the δ column, a fact emerged that is not an interval issue but more fundamental.

    mountain-sample-priority-sigma-pre-screening-sop.md §1 and schumann-station-candidate-tier-a-wet-basalt.md §1 share a three-tier table. The δ column in this table declares:

    Row σ Interval δ Declaration Calculated at 7.83 Hz Interpretation
    Dry Granite 10⁻⁸–10⁻⁴ Hundreds of meters – Thousands of km 18.0 km – 1,800 km Order-of-magnitude mismatch (off by 2–3 orders)
    Wet Granite 10⁻³–10⁻¹ Tens of meters – Hundreds of meters 0.569 km – 5.69 km Order-of-magnitude mismatch (off by 1–2 orders)
    Wet Basalt 10⁻²–10⁰ Meters – Tens of meters 0.180 km – 1.80 km Order-of-magnitude mismatch (off by 1–2 orders)

    Citing specific figures from a single table is more rigorous than line-by-line comparison: schumann-station-candidate-tier-a-wet-basalt.md explicitly states “Tier A σ ≥ 10⁻² → δ ≤ 67 km“. However, at 7.83 Hz, the δ corresponding to σ=10⁻² is 1.7986 km——

    67 km / 1.7986 km = 37.2508
    

    The discrepancy is exactly 37.25 times. Meanwhile, mountain-sample-priority-sigma-pre-screening-sop.md §2 defines Tier A criteria as “σ ≥ 10⁻² → ≥ 10 dB” and Tier B as “δ tens of meters – hundreds of meters”. Both files utilize the same numerical set, and the ratio of 37.25 is identical across both documents.

    Back-calculation reveals that to yield 67 km for σ=10⁻², the equivalent frequency must be 0.00564 Hz, not 7.83 Hz.

    This indicates that the δ column in these two tables was not calculated at 7.83 Hz, but rather inherited from an upstream value across the entire section (otherwise, the identical 37.25 ratio would not appear in both files).

    This error (#37) is undetectable by the scanner: I ran delta-table-scan.py against these two files individually, resulting in “Matched files: 0 | Total assertions: 0″——the parser completely failed to recognize the table. The root cause is that its header perception rule fails to identify the column labeled “δ(7.83 Hz)” as a δ column, causing the entire table to be skipped.

    This mirrors the lesson from Round 9: The tool executes successfully, but the metric coverage does not reach the target. The distinction here is that this is a coverage gap, not a metric selection error.

    It must be distinguished clearly: The A(dB) column is not calculated incorrectly; rather, the interval coverage is incomplete.

    Row Calculated A at Interval Ends Stated A in Table Coverage Status
    Wet Granite (#22 §1) 1.53 – 15.27 dB “Order of 1–10 dB” Lower end approximates correctly; upper end misses by 5.27 dB (exceeds K1)
    Wet Basalt (#22 §1) 4.83 – 48.29 dB “Order of tens of dB” Upper end correct; lower end 4.83 dB is not “tens”

    Therefore, one cannot claim the “entire table is wrong”—the δ column is erroneous, while the A column is correct but lacks full interval coverage. These are two distinct issues operating independently.

    6. Impact on Existing Outputs from the Current Repository

    Existing Output Impact
    #16 skin-depth-reference-table.md §3 δ Table Not within the 37.25× scope (the table’s δ aligns with the formula; measured σ=10⁻⁴ yields ~17.99 km). However, §2 parameter table was flagged by the tool as 5 interval rows with q=10 or 100, constituting an L3 underdetermined state (the table originally lacks z; handled per parameter table guidelines, see Governance Card §4)
    #22 mountain-sample-priority-sigma-pre-screening-sop.md δ column requires recalculation (due to 37.25× deviation); Column A remains usable. Since the three-tier classification (A/B/C) relies on Column A, the categorical conclusions remain unchanged, but interval notations such as “1–10 dB” and “tens of dB” should be revised
    #25 schumann-station-candidate-tier-a-wet-basalt.md Same as above: δ column requires recalculation; “δ ≤ 67 km” should be corrected to 1.8 km; the Tier A criterion (σ ≥ 10⁻²) remains unaffected
    #37 Full Scan (59 Grid MISMATCH) Conclusions remain unchanged (reflecting internal consistency within the table); however, it must be explicitly clarified that its coverage does not include these two files, and “zero hits” cannot be interpreted as “no issues”
    #46a / P4-AUDIT-005 (f_meas Directionality) Orthogonal to this round: f_meas determines “which frequency σ anchors to,” while q determines “the magnitude of difference within the interval.” They can be combined: true uncertainty ≈ √q × f_meas factor
    #14 H1a Threshold (≤3 dB) Maintain the judgment from P4-AUDIT-005: Oppose lowering the threshold

    7. Evidence and Limitations

    Verified Evidence (Reproducible):

    • The equations $A_{max}/A_{min} = \sqrt{\sigma_{max}/\sigma_{min}}$ and $q \le (13/7)^2 = 3.449$ represent unconditional safety thresholds derived as analytical identities, independent of any literature assumptions;
    • The full-library scan counts (925 files / 866 tables / 31 tables containing $\sigma$ / 117 $\sigma$ rows / 31 interval rows / 5 cross-tier cases / 24 without $z$) can be reproduced line-by-line via sigma-interval-tier-audit.py scan, with JSON export capability;
    • Tool self-checks passed 15/15, covering four boundary cases: “cross-tier determination,” “same-tier determination,” “scientific notation parsing for superscripts,” and “distinguishing interval delimiters from negative superscript signs”;
    • The 37.25x deviation in the $\delta$ column was confirmed by recalculating two files character-by-character; the identical ratio appearing at different $\sigma$ values across both files is verifiable at the file level;
    • Strong frequency dependence in rock $\sigma$ is supported by five independent literature sources (Mehran 1977, Jonscher 1977, Revil 2013, Laštovičková 1992, Hall 2004), indicating that “intervals” reflect genuine physical uncertainty rather than fictional constructs;
    • Three PMID citations were verified via E-utilities, and five DOIs via Crossref; titles, first authors, journals, and publication years remain consistent across all checks.

    Limitations (Must Read Concurrently with Conclusions):

    1. The cause of the 37.25x deviation is not addressed in this text. Only the equivalence of “frequency 0.00564 Hz” and the consistency of ratios at two locations are confirmed; no speculation is made regarding upstream causes such as manual calculation errors, unit confusion, or copying from a table based on another frequency. It cannot be excluded that the table’s author intentionally employed a low-frequency scope, though the header explicitly states 7.83 Hz (thus deemed inconsistent with the header);
    2. The tool recognizes only Markdown tables. Naked numerical assertions in the main text (e.g., “a certain mountain attenuates by 3 dB”) fall outside the scope, similar to issues raised in Round 9 regarding naked lists;
    3. The figures 117 and 31 represent “tool-recognized” counts, not necessarily “truthfully existing” counts in the library. Unrecognized does not imply non-existence; given current table header rules, two critical tables were missed in recognition (see §5), so the actual number of $\sigma$ rows should be $\ge 117$;
    4. The tier threshold of 13/7 dB is self-defined by Pool ④, not an industry standard. If the threshold changes, the 3.449 line shifts accordingly ($q \le (T_{K1}/T_{K2})^2$). This paper does not advocate for external validation of this threshold;
    5. The approximations of good conductors, infinite thickness plates, and normal incidence remain in effect; absolute values of $A(\text{dB})$ are not claimed to represent true field values for any specific site;
    6. $q > 3.449$ does not equal “this row is wrong.” It equals “this row cannot yield a single-tier conclusion without locking $z$”—this represents an underdetermined system, not an error.

    8. Action Recommendations for the Next Iteration

    Priority Action Dependency
    P1 Complete the 24 rows in column z for L3—this is the sole prerequisite to unlock all determinations, costing nothing locally None
    P1 Modify the dual columns for the 5 rows in L2 (one column at the interval lower bound, one at the upper bound + tier), or add a conditional clause stating “limited to σ taking X for this row” Requires JOE authorization to modify existing text
    P1 Recalculate columns δ for #22/#25 (37.25× deviation)—pure arithmetic, but modifies published article text, requires JOE authorization Requires JOE
    P2 Add header rules to delta-table-scan.py to cover column names with units, such as “δ(7.83 Hz)” Can be done locally
    P2 Incorporate the unconditional safety threshold q ≤ (T₁/T₂)² into P4-AUDIT-004 as an admission check for new table generation Can be done locally
    P3 Backward scan pool ③ OMOM three-axis attribution table and pool ⑤ F1–F5 thresholds for similar under-determination Pending completion of P1

    This document is a draft. All content consists of analytical recalculation and text parsing from this repository; no field measurements were conducted.
    This does not constitute a truth claim regarding ELF attenuation for any specific mountain or site.
    Before entering version 779 final, an intake process is required; before entering version 782, the text must be de-AI’d.
    “Zero rows with unconditional safety among the 31 rows” is the hardest number in this round, and also the one least suitable to be packaged as “fixed”—it merely clarifies under-determination without modifying a single row.


    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.

  • 账本对不平的那几笔钱,是谁花的:高谐波残差的归因顺序与阶次指纹

    Evidence level: B

    方差分解告诉你”谁占几成”,但先查谁、后查谁,会改变谁占几成。残差归因真正难的不是算份额,是定顺序。

    正文

    上一篇把谐波预测误差拆成几何、弹性、边界三方份额,却留了个包袱:拆份额前得先决定”按什么顺序查账”。顺序不是技术细节——同一笔钱,先查甲的账甲就背,先查乙就乙背。这篇专门处理这被跳过的一步。

    先说清为什么顺序会出事。方差分解的前提是输入彼此独立,但生理上这三项天然共变:年纪大的人血管又粗又硬,几何与弹性同涨;交感张力上去,粘性和出口阻力一起变。输入相关时,经典主效应分解会把”两个变量共同造成的那块方差”记不清归属,甚至让它凭空消失。近年的方法学综述点得很直白:绝大多数已发表敏感性分析仍是单因子轮换,默认模型线性可加,而这些假设往往站不住(PMID 26934843)。更关键的一条——传统方差基方法假设输入独立,这在校准过的模型上几乎从不成立;要公平分配就得用 Shapley 效应,而它又要求能对相关联合分布做条件采样(PMID 41838759)。技术含义是:“三方各占几成”在方法选错时是个假数字,假的方式取决于查账顺序。

    第二件事,残差本身带指纹。参数误差随机散布,漏掉的结构项有方向性偏移,两者能分开看。非线性弹性管实验给了最干净的对照:同一根管、两种生理脉幅,线性方法在高阶散射明显增大,且随脉幅升高,三点法在低阶开始给出负的衰减估计(PMID 10468357)——负衰减物理上不成立,说明那部分”残差”根本不是血管的,是方法与非线性打架的产物。所以第一个筛子不是”谁占几成”,而是”这笔残差跨重复测量能不能复现”。测量链这条漏账线也不是虚的:用指尖光电容积波重建中心波形时,研究者得准备三套传递函数、按输入信号的谐波轮廓挑其中一套,才把形状误差压到 1.8%(PMID 39479582)——误差大小取决于这一拍的谐波谱长什么样,本身就是谐波依赖的系统偏移。不可复现的残差属于测量链,归因无意义。

    第三件事,三条非几何候选的证据成色并不一样,混着用会互相顶账。

    弹性这条最像”已结案”,其实没结。体外 37 条动脉全参数实测、零拟合下,纯弹性壁会在末梢产生欠阻尼高频振荡,加入粘弹性后第 5-20 谐波压力误差从 0.7% 降到 0.5%、流量侧 7.0% 降到 3.3%(PMID 21724188)。数字干净,但前提是”弹性已知”;一旦从实测变成反解,账就还回去了。更要紧的是弹性不是物性常数——人体实验里抑制一氧化氮合成,桡动脉壁相对粘性从 34.9% 掉到 28.9%;阻断钙激活钾通道则从 31.7% 升到 35.7%,两者叠加升到 41.5%(PMID 29158357)。同一根血管、同一个下午,只改内皮信号,粘性就动七八个百分点。

    边界这条影响确实大,但大不等于可分离。换成多孔弹性管代表微循环后,作者明确说边界参数对压力与流量分布有”深刻影响”(PMID 30900050)。问题在于,”深刻影响”与”能唯一反解”是两件事,前者成立恰恰让后者更难。

    交感这条最容易被漏账,因为它在模型里长得像边界参数。综述汇总微神经电图证据,结论是交感能在独立于当前血流动力学与血管运动张力的条件下调节大动脉硬度(PMID 33035441)。也就是说它既能改弹性也能改边界,是个跨两栏的项。跨栏项被硬塞进某一栏,另一栏就会显得解释力不足。

    第四件事,也是最要命的一件:共线已在实测层面被看见了。谐波失真这个指标被系统检验过——24 小时里变异很大,随收缩压升高线性下降,同时与切向模量、脉搏波速度线性相关(PMID 35433650);传输线模型的敏感性分析则显示,增强指数、脉搏波速度、谐波失真三者主要对结构硬度敏感,而谐波失真与增强指数同时与几何参数强相关(PMID 40097550)。同一张谱,既随硬度走,也随几何走,还随当下血压走。任何”先固定几何、再扫弹性、再扫边界”的单向顺序,后扫者都会把前面所有的账吸进自己身上——这叫参数补偿,不叫归因。

    所以方法序必须改。建议五步:第零步先做残差可归因性体检,看跨重复测量是否复现、是否与压力心率相关;第一步改用 Shapley 效应分配;第二步按压力与心率分层,把幅值依赖的非线性单拎出来(PMID 10468357 的负衰减即此层证据);第三步用可辨识性筛掉反解不出来的床;第四步报份额,并同时报”相关”与”独立”两种输入设定下的差值——差值本身就是共线强度。

    判据提一个新的:“小修还是大修”的分界线不是份额大小,而是残差能不能被已有参数吸收。若某个非几何参数扫过一遍残差就掉到噪声底,共振腔假说只需小修——把该参数个体化即可;若单参数都吸收不了、必须几个参数同时调且互相补偿才勉强对上,说明模型形式错了,缺的是结构性维度(如压力依赖、频率依赖的粘性),这就不是调参能救的。

    证据与局限

    已验证: 输入相关时经典方差分解的独立性假设不成立,须改用 Shapley 效应并做条件采样(PMID 41838759);已发表敏感性分析多数仍用单因子轮换并默认线性可加(PMID 26934843);管壁粘弹性显著压低第 5-20 谐波误差(PMID 21724188);内皮 NO 与钙激活钾通道可急性改变人桡动脉壁相对粘性(PMID 29158357);交感可独立于血流动力学调节大动脉硬度(PMID 33035441);非线性弹性管中线性方法在高阶散射增大、随脉幅出现负衰减(PMID 10468357);谐波失真随收缩压线性下降且与硬度、几何同时相关(PMID 35433650、PMID 40097550);边界模型参数对压力流量分布影响深刻(PMID 30900050)。

    部分验证: “残差可分噪声型与结构型”目前是从非线性管实验与敏感性方法学两头推出来的,尚未在”同个体 + 逐阶谐波输出”上直接做过判别检验;三条候选的相对份额(弹性/边界/交感)从未在相关输入设定下同时估计;”参数补偿导致后扫者背账”是机制推论,缺逐阶实证。

    待验证: ①同批受试者上”独立输入 vs 相关输入”两种设定下三方份额的差值(共线强度的直接度量);②残差跨重复测量的可复现性分层(可复现/不可复现各占多少阶次);③残差与瞬时血压、心率的逐阶相关结构;④单参数扫描能否把残差压到噪声底的”可吸收性”测试——这是小修/大修判据的唯一直接检验。

    相关阅读

    证据等级: B
    参考文献: PMID 10468357, PMID 21724188, PMID 26934843, PMID 29158357, PMID 30900050, PMID 33035441, PMID 35433650, PMID 40097550, PMID 41838759, PMID 39479582


    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。

    Who Spent the Discrepancies in the Ledger: Attribution Sequence and Order Fingerprints of High Harmonic Residuals

    Evidence level: B

    Variance decomposition tells you “what share each factor occupies,” but the order in which you audit determines that share. The true difficulty of residual attribution is not calculating shares but establishing the sequence.

    Main Text

    The previous article decomposed harmonic prediction errors into three shares: geometry, elasticity, and boundary. However, it left an unresolved issue: before assigning shares, one must decide “in what order to audit the accounts.” Order is not a technical detail—for the same debt, if you audit Account A first, A takes responsibility; if you audit B first, B takes responsibility. This article addresses this skipped step.

    First, clarify why order causes errors. Variance decomposition assumes inputs are independent, but physiologically these three factors naturally covary: older individuals have thicker, stiffer vessels where geometry and elasticity rise together; increased sympathetic tension causes viscosity and outlet resistance to change simultaneously. When inputs are correlated, classical main-effect decomposition fails to correctly attribute the variance caused jointly by two variables, sometimes even making it vanish. Recent methodological reviews state plainly that most published sensitivity analyses still use single-factor rotation, assuming linearity and additivity—assumptions that often fail (PMID 26934843). A more critical point: traditional variance-based methods assume independent inputs, which almost never holds for calibrated models; fair attribution requires Shapley effects, which in turn require conditional sampling over correlated joint distributions (PMID 41838759). The technical implication is: “What share each party occupies” becomes a false number when the method is wrong, and the nature of the falsehood depends on the audit order.

    Second, residuals themselves carry fingerprints. Parameter errors scatter randomly, while missed structural terms exhibit directional bias; these can be distinguished. Nonlinear elastic tube experiments provided the cleanest control: with the same tube and two physiological pulse amplitudes, linear methods showed significantly increased higher-order scattering, and as pulse amplitude rose, the three-point method began yielding negative attenuation estimates at lower orders (PMID 10468357)—negative attenuation is physically impossible, indicating that portion of the “residual” was not from the vessel but a product of conflict between method and nonlinearity. Thus, the first filter is not “what share,” but “can this residual be reproduced across repeated measurements.” The leak in the measurement chain is also not fictitious: when researchers reconstruct central waveforms using fingertip photoplethysmography, they must prepare three transfer functions and select one based on the harmonic profile of the input signal to reduce shape error to 1.8% (PMID 39479582)—error magnitude depends on the harmonic spectrum of this specific beat, constituting a harmonic-dependent systematic offset. Irreproducible residuals belong to the measurement chain; attribution is meaningless.

    Third, the evidentiary quality of the three non-geometric candidates varies; mixing them causes mutual offsetting.

    The elasticity line appears “closed,” but is not. Under zero-fitting conditions with full-parameter measurement of 37 in-vitro arteries, pure elastic walls produce under-damped high-frequency oscillations at the periphery; adding viscoelasticity reduced pressure error at harmonics 5–20 from 0.7% to 0.5% and flow-side error from 7.0% to 3.3% (PMID 21724188). The numbers are clean, but only under the premise that “elasticity is known.” Once moving from measurement to inversion, the ledger returns. More importantly, elasticity is not a constant property—in human experiments, inhibiting nitric oxide synthesis dropped relative wall viscosity from 34.9% to 28.9%; blocking calcium-activated potassium channels raised it from 31.7% to 35.7%, with both mechanisms combined raising it to 41.5% (PMID 29158357). In the same vessel, same afternoon, altering only endothelial signals shifts viscosity by nearly ten percentage points.

    The boundary line has a significant impact, but significance does not imply separability. When replaced by porous elastic tubes representing microcirculation, the authors explicitly state that boundary parameters have a “profound influence” on pressure and flow distributions (PMID 30900050). The issue is that “profound influence” and “unique invertibility” are two different matters; the former holds precisely makes the latter harder.

    The sympathetic line is most easily overlooked because it resembles boundary parameters within the model. A review summarizing micro-neurographic evidence concludes that sympathetic activity can regulate large artery stiffness independently of current hemodynamics and vascular motion tension (PMID 33035441). That is, it can alter both elasticity and boundary conditions, acting as an item spanning two columns. Forcing a cross-column item into a single column causes the other column to appear insufficiently explanatory.

    Fourth, and most critical: collinearity has been observed at the empirical level. The harmonic distortion metric has undergone systematic testing—variation is high over 24 hours, decreases linearly with rising systolic pressure, and correlates linearly with tangential modulus and pulse wave velocity (PMID 35433650); sensitivity analysis of transmission line models shows that enhancement index, pulse wave velocity, and harmonic distortion are primarily sensitive to structural stiffness, while both harmonic distortion and enhancement index strongly correlate with geometric parameters (PMID 40097550). The same spectrum moves with stiffness, geometry, and current blood pressure simultaneously. Any unidirectional sequence such as “fix geometry first, then sweep elasticity, then sweep boundary” causes the latter to absorb all preceding accounts—this is parameter compensation, not attribution.

    Therefore, the methodological sequence must change. We propose five steps: Step Zero performs a residual attribution viability check, verifying reproducibility across repeated measurements and correlation with pressure and heart rate; Step One adopts Shapley effect allocation; Step Two stratifies by pressure and heart rate to isolate amplitude-dependent nonlinearity (the negative attenuation in PMID 10468357 serves as evidence for this layer); Step Three uses identifiability screening to eliminate beds that cannot be inverted; Step Four reports shares while simultaneously reporting the difference under “correlated” and “independent” input settings—the difference itself quantifies collinearity strength.

    We propose a new criterion: The dividing line between “minor repair” and “major overhaul” is not share size, but whether existing parameters can absorb the residual. If a non-geometric parameter reduces the residual to noise floor upon a single sweep, the resonance cavity hypothesis requires minor repair—individualizing that parameter suffices. If no single parameter absorbs it, and multiple parameters must be adjusted simultaneously with mutual compensation to barely match, the model form is incorrect; it lacks structural dimensions (such as pressure-dependent or frequency-dependent viscosity), which cannot be saved by parameter tuning alone.

    Evidence and Limitations

    Verified: Independence assumptions in classical variance decomposition fail under correlated inputs; Shapley effects with conditional sampling are required (PMID 41838759); most published sensitivity analyses still use single-factor rotation assuming linearity and additivity (PMID 26934843); wall viscoelasticity significantly suppresses errors at harmonics 5–20 (PMID 21724188); endothelial NO and calcium-activated potassium channels can acutely alter relative wall viscosity in human radial arteries (PMID 29158357); sympathetic activity can regulate large artery stiffness independently of hemodynamics (PMID 33035441); linear methods show increased higher-order scattering and negative attenuation with rising pulse amplitude in nonlinear elastic tubes (PMID 10468357); harmonic distortion decreases linearly with systolic pressure and correlates with both stiffness and geometry (PMID 35433650, PMID 40097550); boundary model parameters profoundly influence pressure and flow distributions (PMID 30900050).

    Partially Verified: The distinction between “noise-type” and “structural-type” residuals is currently inferred from nonlinear tube experiments and sensitivity methodology, lacking direct discriminant testing in “same individual + stepwise harmonic output”; relative shares of the three candidates (elasticity/boundary/sympathetic) have never been estimated simultaneously under correlated input settings; “parameter compensation causing posterior absorption of accounts” is a mechanistic inference lacking stepwise empirical validation.

    To Be Verified: ① Differences in three-party shares under “independent input” vs. “correlated input” settings within the same cohort (direct measure of collinearity strength); ② Stratification of residual reproducibility across repeated measurements (proportion of reproducible vs. irreproducible orders); ③ Stepwise correlation structures between residuals and instantaneous blood pressure and heart rate; ④ “Absorbability” testing via single-parameter scanning to reduce residuals to noise floor—the only direct test for the minor/major repair criterion.

    Related Reading

    Evidence Level: B
    References: PMID 10468357, PMID 21724188, PMID 26934843, PMID 29158357, PMID 30900050, PMID 33035441, PMID 35433650, PMID 40097550, PMID 41838759, PMID 39479582


    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.

  • 同一张谐波表,五个器官不一定都认:把 C1-C10 对应表拆成逐器官裁决

    Evidence level: B

    “一脏一谐波”这张表如果真有物理基础,它应该过得了很朴素的一关:五个器官的门槛高度本来不一样,检出率就该不一样;如果模型连五个床都分不开,实测谱更不可能一一对上。

    正文

    上一篇(organ-arterial-geometry-harmonic-mapping)把问题架到模型层:用器官供血动脉几何喂 1-D 模型,看谐波能量分配能不能对上 C1-C10。那是整体一次比对,颗粒度太粗。真正能决定这张表命运的,是拆开看——五个器官挨个过一遍,谁命中、谁偏离、往哪偏。

    为什么必须拆开?因为五个床的物理门槛本来就不同。肾和肝是两条性质相反的血管床:肾动脉高阻低流、波形尖锐,早期收缩峰是它的标志特征,而这个峰的出现率与血管顺应性直接相关,年轻健康人反而常看不到(PMID 7644627);肝动脉低阻高舒张、波形圆钝,阻力指数单次测量标准差就有 0.08、收缩加速时间 39 毫秒,个体内变异占总方差 76%(PMID 8756922)。一个”尖”一个”钝”,一个重复性好一个差。若谐波映射为真,这两类床在谐波域的表现必然不同步,而不是整齐划一”各占一阶”。

    第一层:结扎实验给的是”能分开”,不是”能对上”。 大鼠实验里结扎肾、胃、脾、肠系膜上动脉,尾巴动脉压力波的谐波谱以不同模式改变、且近似线性可加(PMID 1529726)——器官床在频率域确实像分频滤波器。用肾与肠系膜的不同结扎比例组合,还能分别模拟血管加压素与血管紧张素 II 的谱变化(PMID 16210450)——”从谱反推哪个床阻力变了”原理可行。但两条证据的共同点是:都靠”干预到极端”的差分。 结扎把阻力推到无穷,药物推到极值,这种条件下任何滤波器都会显出差。C1-C10 表要的却是正常人静息、无干预、五个床都在生理范围内时,能量分配能不能一一对应。从”差分可分辨”跨到”绝对可定量”,这一大步从没人走过。

    第二层:模型层早知道”出口参数”比几何更难定。 1-D 模型的出口边界是最大不确定源:用多孔介质管替代 RCR 描述微循环阻力,须重新标定才能复现实测波形(PMID 30900050);67 段、16 个血管床的手臂模型上,16 个末端阻力得靠解非线性方程组反解,还要做敏感性分析确认反解站得住(PMID 22771032);冠脉的不确定性量化则直接给出排序——几何影响最大,边界阻力次之,黏度再次(PMID 26803339)。翻成白话:你想用几何预测谐波分配,可谐波分配对出口阻力的敏感度仅次于几何,而出口阻力恰是五个器官差异最大的那个参数。

    第三层:五个器官的血流占比从来不是固定值。 用 99mTc 微球测人心输出量分配,脑、心、肾的占比都与心指数负相关(r = −0.70、−0.67、−0.47),低心排时脑心优先保住、肾次之(PMID 3947835)。餐后更剧烈:标准餐后肠系膜上动脉血流上升、乳糜干血管舒张,而 Fontan 病人的 celiac 舒张反应直接消失(PMID 32857602)。这些重分配发生在分钟到小时级。 而 C1-C10 是静态查表,它假设”第几阶归哪个脏”恒定。若占比本身动态调制,那”某阶能量主要来自某器官”在饭前饭后就该给出不同答案——这张表要么自认是”空腹静息特例”,要么得加一个时间变量。

    第四层:具体怎么裁。 ≥60 名健康受试者,空腹 12 小时、固定体位、固定测量时刻(避开昼夜窗口)。每人做两件事:①超声或 MRI 重建肝、双肾、脾、胃(左胃动脉)、肠系膜上动脉五条供血动脉的入口几何(管径、分支角、主干长度);②同步记录桡动脉连续波形 ≥60 拍做 FFT,取 C1-C10 相对能量。再用同一个已验证的 1-D 模型(个体化几何 + 人群均值弹性 + 个体化出口阻力,阻力由静息血流反解,按 PMID 22771032 的方法),输出五个分支口各自的逐阶能量预测。裁决分两层:器官内一致性——某器官的预测峰值阶次在受试者间是否稳定落在同一阶;器官间可分性——五个器官的预测剖面能否被聚类或判别分析重新分回五个器官(留一交叉验证混淆矩阵)。第二层才是关键:若连模型自己都分不开这五个床,实测谱就更不可能一一对上。

    三种结局都是知识。五个剖面聚类后能正确归位、峰值阶次又与 C1-C10 吻合 ⇒ “一脏一谐波”首次获得”逐器官可计算”的支持,经验查表可升级成定量映射。若只有肾、肝两床可分(它们阻力性质差异最大),脾、胃、肠系膜互相混淆 ⇒ 这张表实际只支持两三个大类,应按器官合并修订——不是否定传统,是把过细的表修成站得住的表。若五床在模型层完全不可分、峰值阶次与表的相关接近零 ⇒ “一脏一谐波”的物理基础不成立,谐波谱读不出”哪个脏”,只能读”整体血管状态”;共振腔假说须从”器官分频”退到”整体模态”,而脉象-脏腑对应得另找通路(例如自主神经支配差异而非共振频率)。

    证据与局限

    已验证: 器官血管床在频率域有可分辨的滤波特性,结扎不同器官动脉引发脉搏谱特定模式变化且近似线性可加(PMID 1529726);结扎诱导的谐波变化可反推区域性血管阻力升高(PMID 16210450);1-D 模型出口阻力须反解标定且对波形有实质影响(PMID 22771032、PMID 30900050);不确定性排序中几何最大、边界阻力次之(PMID 26803339);器官血流占比随心指数(PMID 3947835)与进食(PMID 32857602)显著重分配;肾动脉波形尖锐度受血管顺应性直接调制(PMID 7644627);肝动脉阻力指数与收缩加速时间个体内变异占总方差 76%(PMID 8756922)。

    部分验证: 五个器官在谐波域的剖面能否互相区分,目前只有动物极端干预数据,缺少正常人静息状态下逐器官的模型输出对比;”一脏一谐波”的峰值阶次归属从未在逐器官粒度被检验。

    待验证: ① 五器官分支口预测剖面能否按器官正确聚类(留一交叉验证混淆矩阵);② 各器官峰值阶次与 C1-C10 的吻合是全面、部分还是随机;③ 考虑餐后重分配后,映射是否需引入时间变量(空腹表与餐后表是否不同);④ 出口阻力反解在五个床上的可辨识性是否一致——若某床反解不唯一,该床的比对结论即不成立。

    相关阅读

    证据等级: B
    参考文献: PMID 1529726, PMID 16210450, PMID 22771032, PMID 30900050, PMID 26803339, PMID 3947835, PMID 32857602, PMID 7644627, PMID 8756922


    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。

    One Harmonic Chart, Five Organs Not Always Agree: Splitting the C1-C10 Reference Table into Organ-Specific Judgments

    Evidence level: B

    “One organ, one harmonic” If this table possesses a physical basis, it should pass a very basic test: the threshold height for detection varies inherently among the five organs; therefore, detection rates should differ. If a model cannot distinguish between five beds, the measured spectra certainly cannot align one-to-one with each organ.

    Main Text

    The previous article (organ-arterial-geometry-harmonic-mapping) framed the issue at the model level: feeding organ arterial geometry into a 1-D model to see if harmonic energy distribution aligns with C1-C10. That was an overall single comparison with coarse granularity. What truly determines the fate of this table is the dissection approach—examining each of the five organs individually: who hits the mark, who deviates, and in which direction.

    Why must we dissect? Because the physical thresholds for the five beds differ inherently. The renal and hepatic vascular beds are two vessels with opposite properties: renal arteries exhibit high resistance and low flow with sharp waveforms; the early systolic peak is a hallmark feature, and its occurrence rate correlates directly with vascular compliance, often absent in young healthy individuals (PMID 7644627). Hepatic arteries show low resistance and high diastolic flow with blunt waveforms; the resistance index has a single-measurement standard deviation of 0.08, and acceleration time is 39 ms, with intra-individual variation accounting for 76% of total variance (PMID 8756922). One is “sharp,” the other “blunt”; one has high reproducibility, the other low. If harmonic mapping were true, these two beds would inevitably exhibit different behaviors in the harmonic domain rather than uniformly occupying one order each.

    First Layer: Ligation experiments provide “separability,” not “alignment.” In rat experiments, ligation of renal, gastric, splenic, and superior mesenteric arteries caused the tail artery pressure wave harmonic spectra to change in different patterns that were approximately linearly additive (PMID 1529726)—vascular beds indeed act like bandpass filters in the frequency domain. Using different combinations of renal and mesenteric ligation ratios can also simulate spectral changes induced by vasopressin and angiotensin II (PMID 16210450)—the principle of “inferring which bed’s resistance changed from the spectrum” is feasible. However, the common point of both evidences is reliance on “extreme intervention” differentiation. Ligation drives resistance to infinity, drugs push values to extremes; under such conditions, any filter will display differences. The C1-C10 table requires determining whether energy distribution corresponds one-to-one under normal human resting conditions with no intervention and all five beds within physiological ranges. Crossing from “differentiable via differentiation” to “absolutely quantifiable” is a leap no one has taken.

    Second Layer: Model layers already know “exit parameters” are harder to define than geometry. The exit boundary of the 1-D model is the source of maximum uncertainty: replacing RCR with porous media tubes to describe microcirculatory resistance requires re-calibration to replicate measured waveforms (PMID 30900050); on a 67-segment, 16-bed arm model, 16 terminal resistances must be solved by inverting nonlinear equations, requiring sensitivity analysis to confirm the inversion is robust (PMID 22771032); coronary uncertainty quantification directly provides a ranking—geometry has the largest impact, followed by boundary resistance, then viscosity (PMID 26803339). Put plainly: You want to predict harmonic distribution using geometry, but harmonic distribution is second only to geometry in sensitivity to exit resistance, and exit resistance is precisely the parameter with the greatest variation among the five organs.

    Third Layer: Blood flow proportions among the five organs are never fixed values. Using 99mTc microspheres to measure cardiac output distribution, the proportions for brain, heart, and kidney are all negatively correlated with cardiac index (r = −0.70, −0.67, −0.47); during low cardiac output, brain and heart are prioritized for preservation, followed by kidney (PMID 3947835). Postprandial changes are more intense: superior mesenteric artery flow increases and celiac trunk vessels dilate after a standard meal, while the celiac dilation response disappears directly in Fontan patients (PMID 32857602). These reallocations occur over minutes to hours. Yet C1-C10 is a static lookup table assuming “which order belongs to which organ” is constant. If proportions are dynamically modulated, then “energy from a specific order primarily originates from a specific organ” should yield different answers before and after meals—this table must either admit itself as an “empty-stomach resting exception” or incorporate a time variable.

    Fourth Layer: How to adjudicate. With ≥60 healthy subjects, fasting for 12 hours, fixed posture, and fixed measurement time (avoiding circadian windows). Each person performs two tasks: ① Reconstruct entry geometry of five supplying arteries (liver, bilateral kidneys, spleen, left gastric artery, superior mesenteric artery) via ultrasound or MRI (tube diameter, branch angles, main trunk length); ② Simultaneously record continuous radial artery waveforms ≥60 beats for FFT, taking relative energy at C1-C10. Then use the same verified 1-D model (individualized geometry + population mean elasticity + individualized exit resistance, derived by inverting resting flow per PMID 22771032) to output predicted step-by-step energy for each of the five branch outlets. Adjudication has two layers: Organal Internal Consistency—whether the predicted peak order for a specific organ remains stable across subjects; Organal Inter-Comparability—whether the predicted profiles for five organs can be re-clustered or discriminated back into five organs via leave-one-out cross-validation confusion matrices. The second layer is critical: If the model itself cannot distinguish these five beds, the measured spectra certainly cannot align one-to-one.

    Three outcomes are all knowledge. If clustering of five profiles correctly repositions them and peak orders match C1-C10 ⇒ “One organ, one harmonic” gains initial support for “organ-specific calculability,” and empirical lookup tables can upgrade to quantitative mappings. If only renal and hepatic beds are distinguishable (due to maximal resistance property differences), while spleen, stomach, and mesentery confuse each other ⇒ the table actually supports only two or three broad categories and should be revised by merging organs—not denying tradition, but refining an overly detailed table into a standable one. If all five beds are completely indistinguishable at the model level, with peak orders correlating near zero to the table ⇒ “One organ, one harmonic” lacks a physical basis; harmonic spectra cannot read “which organ,” only “overall vascular state”; the resonance cavity hypothesis must retreat from “organ frequency division” to “global modes,” while pulse-diagnosis organ correspondence needs alternative pathways (e.g., autonomic innervation differences rather than resonance frequencies).

    Evidence and Limitations

    Verified: Vascular beds exhibit distinguishable filtering characteristics in the frequency domain; ligation of different organ arteries induces specific pattern changes in pulse spectra that are approximately linearly additive (PMID 1529726); harmonic changes induced by ligation can infer regional vascular resistance elevation (PMID 16210450); exit resistance in 1-D models must be inverted for calibration and substantially affects waveforms (PMID 22771032, PMID 30900050); uncertainty ranking shows geometry largest, boundary resistance second (PMID 26803339); organ blood flow proportions significantly reallocate with cardiac index (PMID 3947835) and feeding (PMID 32857602); renal artery waveform sharpness is directly modulated by vascular compliance (PMID 7644627); hepatic artery resistance index and acceleration time show intra-individual variation accounting for 76% of total variance (PMID 8756922).

    Partially Verified: Whether profiles of five organs can be distinguished in the harmonic domain currently relies only on animal extreme intervention data, lacking model output comparisons under normal human resting states for organ-specific granularity; peak order attribution of “one organ, one harmonic” has never been tested at the organ-specific level.

    To Be Verified: ① Can predicted profiles of five organ branch outlets be correctly clustered by organ (leave-one-out cross-validation confusion matrix); ② Is the alignment of peak orders with C1-C10 comprehensive, partial, or random; ③ Considering postprandial reallocation, does mapping require introducing a time variable (different between fasting and postprandial tables); ④ Is the identifiability of exit resistance inversion consistent across the five beds—if inversion is non-unique for a bed, that bed’s comparison conclusion fails.

    Related Reading

    Evidence Level: B
    References: PMID 1529726, PMID 16210450, PMID 22771032, PMID 30900050, PMID 26803339, PMID 3947835, PMID 32857602, PMID 7644627, PMID 8756922


    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.

  • 扣掉节律之后更像你自己了吗:用识别力 AUC 裁决校正谱到底买到了什么

    Evidence level: B

    稳定性回答的是”同一个人两次测像不像”,识别力回答的是”两个人之间分不分得开”。扣掉节律能改善前者,但对后者可能一点用都没有,甚至帮倒忙。

    正文

    前面两篇一直在算一致性:扣掉昼夜与季节节律之后,脉搏谐波谱的重测 ICC 会涨多少、涨在哪几阶。但一致性高不等于这个人可辨认。这是两个数学上完全不同的问题,而脉象文献把两者混为一谈已经很久了。

    先看差别在哪。ICC 是一个比值:个体间方差除以(个体间方差 + 个体内方差)。AUC 问的是另一件事:随便抽两次测量,来自同一个人的那对的相似度,超过来自不同人的那对的概率是多少。前者只看”波动占多大比例”,后者看”人和人之间的间隔,相对每个人的抖动有多大”。一个人可以极其稳定(ICC 0.95),但如果他和邻居本来就长得一样(间隔窄),AUC 依然逼近 0.5。反过来,抖动大但个体间隔更宽,AUC 照样能上 0.9。

    这个区分不是文字游戏,它决定了节律校正到底该不该做。

    校正有两种性质,ICC 分不出来,AUC 能。 如果把所有人按同一时刻的平均节律曲线扣掉——这是最常见也最容易实现的做法——本质上是把节律当成共同噪声。此时个体内方差下降、个体间方差不动,ICC 必然上涨。但 AUC 涨不涨,取决于节律里有没有个体成分。如果节律是全民共享的(所有人同相、同幅),扣它就是纯降噪,AUC 跟着涨;如果每个人有自己的相位和幅度——有人是早起型、有人是夜猫型,有人季节反应强、有人几乎不动——那么扣掉群体平均曲线,等于把每个人身上最独特的那部分一起削掉了,AUC 反而会跌。

    所以 AUC 是唯一能裁决”节律是噪声还是身份”的指标,而 ICC 对此完全瞎。 前面那篇逐阶 ΔICC 的剖面,无论节律是哪种性质,都会画出上升曲线。

    现有文献正好把这条路走了一半。脉搏波做生物识别已经相当成熟:40 人的小样本上用二维 Gram 矩阵加深度网络能到 99% 准确率(PMID 39796830);孪生网络做单周期匹配,AUC 0.988、准确率 0.972(PMID 37430548);跨 ECG/EEG/PPG 的统一框架里,多数数据集识别准确率超过 96%,且用 SHAP 拆出最有判别力的特征是小波分解的第 1 层与第 6 层系数以及收缩期形态(PMID 41812367)。这些数字看着漂亮,但有一篇复现研究把话说明白了:多数工作建立在受控环境、小样本、短期评估、且没有”未见用户”测试之上,一旦进入真实场景、面对信号变异与训练测试之间的时间间隔,性能会大幅下滑(PMID 41471580)。他们给出的整改清单第一条就是时间切分,而不是随机切分。

    最接近”节律校正用于识别”的一项工作,做法是把心电与脉搏建模成变节律的循环随机过程,先用估计出的节律函数对齐心动周期再提特征,运动中 PPG 准确率 0.962、HTER 0.039(PMID 42385745)。但作者自己写得很清楚:该数据集每人每个活动只有一个记录会话,所以这是会话内可分性,不能当作长期身份恒定性的证据。换句话说,”先对齐节律再识别”这条路只在同一次坐下来的几十分钟里被验证过,跨时辰、跨季节一次都没测。

    还有两个设计上的坑必须先填。第一是样本量本身会压低 AUC。用个性化心搏做 ECG 识别,参考库从 10 人扩到 230 人,12 导联准确率从 98.4% 掉到 87.4%(PMID 29382064)。所以”40 人 AUC 0.99″和”200 人 AUC 0.85″不是同一件事,跨研究比 AUC 毫无意义,这也是一致性文献里那句老话的翻版:相关系数类指标高度依赖测量值的分布范围,不能跨样本比较(PMID 9820922),选哪种 ICC 形式也必须显式交代(PMID 27330520)。第二是识别的可能不是人,是年龄。脉搏谱被反复证明能判血管老化(PMID 34508787)、能分代谢综合征(PMID 27067750,C1、C2 显著变小而 C4 至 C9 变大),这些恰好是”个体间差异”的主要来源。如果冒充者池里年龄、血压、体型都不匹配,模型很可能只是在猜年龄段。必须做年龄性别血压匹配的冒充者池。

    可执行的裁决设计:同一批不少于 120 人,每人跨两个季节各测 6 次以上、覆盖不同时辰。三臂头对头——原始谱、群体节律校正谱、以及不做任何建模只用 24 小时均值(24 小时平均的增强指数与脉搏波速度 ICC 已证实大于 0.8,PMID 24063914,这是最笨但最不会过拟合的对照)。训练用季节 A、测试用季节 B,冒充者池按年龄性别血压匹配。主终点是开集识别 AUC 的差值及 bootstrap 置信区间,同时把 ΔAUC 分解成个体间方差变化与个体内方差变化两部分。

    三种结局都有意义。校正后 AUC 显著高于原始谱——节律是共同噪声,扣掉确实更可辨人,中医”脉有常”这一侧拿到硬支撑。校正后 AUC 与原始谱无差异、但 ICC 明显上升——那说明前面几篇算到的稳定性增益只是把抖动压小了,身份信息一分没多,”更稳”和”更像你”被彻底拆开,这是对整套校正思路最不客气但最诚实的结果。校正后 AUC 反而下降——最有意思的一种:节律里带着个体成分,每个人有自己的时辰曲线,扣掉群体节律等于削掉了签名的一部分。若真如此,”脉有常”与”脉应时辰”不必再靠分配谐波阶次来调和,因为节律本身就是身份的一部分。

    证据与局限

    已验证: 脉搏波/PPG 具备生物识别能力,小样本受控条件下准确率可达 96% 至 99%(PMID 39796830、PMID 37430548、PMID 41812367);但复现研究显示进入真实场景、存在训练测试时间间隔后性能显著下降,现有文献普遍缺时间切分与未见用户测试(PMID 41471580);节律自适应对齐后会话内 PPG 识别准确率 0.962(PMID 42385745);识别准确率随参考库人数增加而单调下降,10 人 98.4%、230 人 87.4%(PMID 29382064);脉搏谱可判别血管老化与代谢综合征状态(PMID 34508787、PMID 27067750);24 小时平均的波反射与硬化指标 ICC 大于 0.8(PMID 24063914);节律性指标季节变异远大于血压,AVI 的 CV 为 12.6%、血压仅 5.1%(PMID 30409046);相关系数类信度指标高度依赖样本取值范围、不可跨样本比较(PMID 9820922),ICC 有十种形式须显式声明(PMID 27330520)。

    部分验证: “节律校正提高 ICC”已有间接支持,但”提高 ICC 会同步提高 AUC”从未被检验——两者在数学上不必然同向,现有文献没有任何一篇在同一批人身上同时报告校正前后的一致性与判别力。SHAP 显示小波第 1、6 层与收缩期特征最具判别力(PMID 41812367),提示判别信息偏向高频细节与形态,而非低频谐波幅度,但该结论建立在单库、单次采集之上。

    待验证: 群体节律扣除后个体识别 AUC 的净变化方向;节律的个体间相位与幅度差异占总节律方差的比例;跨季节外部验证下 AUC 的衰减率是否与 ICC 的衰减率同阶;以及”扣掉多少比例的群体方差”这一校正强度参数是否存在最优值(扣得越狠 ICC 越高,但 AUC 可能先升后降)。

    相关阅读

    证据等级: B
    参考文献: PMID 41471580, PMID 39796830, PMID 37430548, PMID 41812367, PMID 42385745, PMID 29382064, PMID 27330520, PMID 9820922, PMID 34508787, PMID 27067750, PMID 30409046, PMID 24063914

    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


    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。

    Without Rhythm, Do You Look More Like Yourself? What We Actually Bought When Using AUC to Arbitrate Corrected Spectra

    Evidence level: B

    Stability answers “Does this person look the same in two measurements?”, while identification power answers “Can we distinguish between two different people?”. Removing rhythm improves the former, but may be useless for the latter, or even counterproductive.

    Main Text

    The previous two articles have been calculating consistency: by removing circadian and seasonal rhythms, how much does the retest ICC of the pulse harmonic spectrum increase, and at which orders? But high consistency does not equal recognizability. These are two mathematically distinct problems, yet pulse literature has conflated them for a long time.

    First, let’s clarify the difference. ICC is a ratio: between-subject variance divided by (between-subject variance + within-subject variance). AUC asks a different question: given two random measurements, what is the probability that a pair from the same person is more similar than a pair from different people? The former looks only at “what proportion of the variance is fluctuation,” while the latter looks at “how wide the gap is between people relative to each person’s jitter.” One can be extremely stable (ICC 0.95), but if they naturally resemble their neighbor (narrow gap), AUC still approaches 0.5. Conversely, if jitter is large but the inter-individual gap is wider, AUC can still reach 0.9.

    This distinction is not semantic; it determines whether rhythm correction should be performed at all.

    Correction has two natures that ICC cannot distinguish, but AUC can. If everyone’s average rhythm curve at the same time is subtracted—which is the most common and easiest implementation—rhythm is essentially treated as common noise. In this case, within-subject variance decreases while between-subject variance remains unchanged, so ICC must increase. However, whether AUC increases depends on whether the rhythm contains an individual component. If the rhythm is universally shared (same phase and amplitude for all), subtracting it is pure noise reduction, and AUC will rise. If everyone has their own phase and amplitude—some are early birds, some night owls, some have strong seasonal responses, others almost none—then subtracting the group average curve effectively chops off the most unique part of each individual, causing AUC to drop.

    Therefore, AUC is the only metric capable of adjudicating whether rhythm is noise or identity, while ICC is completely blind to this. The previous profile of stepwise ΔICC will show an upward curve regardless of the nature of the rhythm.

    Existing literature has walked half this path. Pulse wave biometric identification is already mature: a 40-person small sample achieved 99% accuracy using a 2D Gram matrix plus deep networks (PMID 39796830); twin networks for single-period matching achieved an AUC of 0.988 and accuracy of 0.972 (PMID 37430548); within a unified framework across ECG/EEG/PPG, most datasets exceed 96% identification accuracy, with SHAP analysis identifying the most discriminative features as coefficients from the 1st and 6th layers of wavelet decomposition and systolic morphology (PMID 41812367). These numbers look impressive, but one replication study clarified the situation: most works rely on controlled environments, small samples, short-term assessments, and lack “unseen user” testing; once moved to real-world scenarios facing signal variability and time intervals between training and testing, performance drops significantly (PMID 41471580). Their first recommendation for rectification is time splitting, not random splitting.

    The work closest to “rhythm correction for identification” models ECG and pulse as cycle-random processes with varying rhythms, aligns cardiac cycles using estimated rhythm functions before feature extraction, achieving 0.962 accuracy and 0.039 HTER in motion PPG (PMID 42385745). However, the authors explicitly state: each participant has only one recording session per activity in this dataset; thus, this is within-session separability, not evidence of long-term identity constancy. In other words, the “align rhythm then identify” path has only been validated within a sitting session of tens of minutes; it has never been tested across hours or seasons.

    Two design pitfalls must also be addressed first. First, sample size itself suppresses AUC. Using personalized heartbeats for ECG identification, expanding the reference library from 10 to 230 people caused 12-lead accuracy to drop from 98.4% to 87.4% (PMID 29382064). Therefore, “40-person AUC 0.99” and “200-person AUC 0.85” are not the same thing; cross-study comparison of AUC is meaningless, which is a reiteration of an old saying in consistency literature: reliability metrics like correlation coefficients heavily depend on the range of measurement values and cannot be compared across samples (PMID 9820922); which ICC form must also be explicitly stated (PMID 27330520). Second, identification may not be of the person, but of age. Pulse spectra have been repeatedly proven to judge vascular aging (PMID 34508787) and distinguish metabolic syndrome (PMID 27067750, where C1 and C2 significantly decrease while C4 to C9 increase)—these are exactly the main sources of “between-subject differences.” If the impostor pool does not match in age, blood pressure, or body type, the model might simply be guessing age groups. An impostor pool matched by age, gender, and blood pressure is required.

    Executable adjudication design: A single batch of at least 120 people, each measured more than 6 times across two seasons covering different times of day. Three arms head-to-head: raw spectrum, rhythm-corrected spectrum, and no modeling at all using only the 24-hour mean (the enhancement index of the 24-hour average and pulse wave velocity ICC are already confirmed >0.8, PMID 24063914; this is the dumbest but least likely to overfit control). Training uses Season A, testing uses Season B; the impostor pool is matched by age, gender, and blood pressure. The primary endpoint is the difference in open-set identification AUC and its bootstrap confidence interval, simultaneously decomposing ΔAUC into changes in between-subject variance and within-subject variance.

    All three outcomes are meaningful. If AUC is significantly higher after correction than the raw spectrum—rhythm is common noise, removing it makes people more distinguishable, providing hard support for the TCM concept “pulse has constants.” If AUC is unchanged but ICC rises significantly—then the stability gains calculated in previous articles merely suppressed jitter without adding any identity information; “more stable” and “more like you” are completely separated, which is the most unkind yet honest result for the entire correction framework. If AUC actually drops after correction—the most interesting case: rhythm contains individual components, everyone has their own time-specific curve, and subtracting the group rhythm chops off part of the signature. If true, there is no need to reconcile “pulse has constants” and “pulse responds to time” by assigning harmonic orders; because rhythm itself is part of identity.

    Evidence and Limitations

    Verified: Pulse wave/PPG possesses biometric identification capabilities; under small-sample controlled conditions, accuracy can reach 96% to 99% (PMID 39796830, PMID 37430548, PMID 41812367); however, replication studies show performance drops significantly upon entering real-world scenarios with time intervals between training and testing, and existing literature generally lacks time splitting and unseen user testing (PMID 41471580); after rhythm adaptive alignment, within-session PPG identification accuracy is 0.962 (PMID 42385745); identification accuracy monotonically decreases as the reference library size increases (98.4% for 10 people, 87.4% for 230 people) (PMID 29382064); pulse spectra can distinguish vascular aging and metabolic syndrome states (PMID 34508787, PMID 27067750); the ICC for wave reflection and stiffening indices of the 24-hour average is greater than 0.8 (PMID 24063914); rhythmic indices show seasonal variation far greater than blood pressure, with AVI CV at 12.6% vs. 5.1% for blood pressure (PMID 30409046); reliability metrics like correlation coefficients heavily depend on sample value ranges and cannot be compared across samples (PMID 9820922); there are ten forms of ICC requiring explicit declaration (PMID 27330520).

    Partially Verified: “Rhythm correction improves ICC” has indirect support, but “improving ICC simultaneously improves AUC” has never been tested—mathematically they do not necessarily move in the same direction, and no existing literature reports both consistency and discriminative power before and after correction on the same batch of people. SHAP analysis indicates that wavelet layers 1 and 6 and systolic features are most discriminative (PMID 41812367), suggesting discriminative information leans toward high-frequency details and morphology rather than low-frequency harmonic amplitudes, but this conclusion is based on a single library and single acquisition.

    To Be Verified: The net direction of change in individual identification AUC after subtracting group rhythm; the proportion of total rhythm variance accounted for by between-subject phase and amplitude differences; whether the attenuation rate of AUC under cross-season external validation is of the same order as the ICC attenuation rate; and whether there exists an optimal value for this correction intensity parameter (“how much group variance to subtract”—more subtraction yields higher ICC, but AUC might first rise then fall).

    Related Reading

    Evidence Level: B
    References: PMID 41471580, PMID 39796830, PMID 37430548, PMID 41812367, PMID 42385745, PMID 29382064, PMID 27330520, PMID 9820922, PMID 34508787, PMID 27067750, PMID 30409046, PMID 24063914

    Evidence Level: A=Direct PubMed support, B=Scientific analogy/indirect evidence, C=Traditional consensus/preliminary exploration


    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.

  • 扣掉节律之后,哪几阶谐波才真变稳:逐阶 ΔICC 剖面的裁决

    Evidence level: B

    节律校正从来不是”全谱一刀切扣”。如果它真能提高个体签名的稳定性,增益必然不均匀地落在某几个谐波阶次上;而”落在哪几阶”这件事,同时裁决了”低阶是身份层”和”校正有必要”这两个前提。

    正文

    上一篇论证的是”扣不扣”:既然昼夜节律能解释脉搏频谱总变异四成以上,而短期重测一致性又足够高,那么把节律参数化扣除,剩下的”节律校正谱”理应比原始谱更稳。但那个推论证的是一个整体 ICC,把 C1 到 C10 打成一个包在比——这在统计上等于假设节律对各阶次影响均匀,而该假设从没被检验过,也几乎没有理由成立。

    先确认测量层是干净的,免得把仪器误差当成生理节律。人工脉搏发生器那项做得最彻底:用已知波形的机械源喂仪器,11 个谐波成分重测信度 ICC 全部大于 0.95、变异系数全部小于 3%(PMID 25693606);TD01C 标准化协议那项,22 名健康人观察者内 ICC 大于 0.9、观察者间 0.83 至 0.96(PMID 27170904)。但请注意它报的是什么:报的是”前 11 个谐波”整体落在这个区间,没有逐阶列出。我们至今只知道这串谐波”总体很好”,并不知 C1 是 0.96 还是 C10 是 0.83。这个看似琐碎的缺口,正是整条论证的入口。

    阶次分层在横向病理证据里其实已经很硬。最漂亮的一份来自临终关怀:21 例末期癌症病人、31 名健康人、47 名门诊病人,测谐波幅度变异系数 HCV。健康组六个 HCV 全部低于 5%,门诊组低于 8%;癌症组第三至第六阶全部超过 15%;而临终前一天,连第一、第二阶也超过 15%(PMID 15472498)。它同时给出方向和梯度:高次谐波先失稳,低次谐波后失稳。高阶天生是”易变层”、低阶天生是”稳固层”,这个分层不依赖任何节律假设,是从生死边缘量出来的。

    但变异并不随阶次单调上升,它有选择性。卒中病人双侧对比:患侧 CV1 至 CV7 全部显著大于健侧,可一旦与健康对照组比,显著的只剩 CV1、CV3、CV4、CV5、CV7——CV2 和 CV6 掉队了(PMID 22156035)。若是纯粹的”阶次越高越不稳”,不该出现这种断点。谱上存在几个特定的”易感阶次”,各对应具体的血流或阻力改变,不是一条光滑斜坡。

    这里撞上了一个真矛盾。系统综述汇总 50 篇临床报告,反复出现的结论是:剧烈运动或交感兴奋后低阶谐波升高,静息或副交感占优时低阶下降;心血管病人 C1 升高对应肝经,糖尿病人 C3 减弱对应脾经(PMID 38481553)。而自主神经恰恰是昼夜节律的主要执行者。若低阶被自主张力调制得最厉害,校正节律时受益最大的就该是低阶而非高阶——这与”低阶稳、高阶变”正好相反。两个都站得住的事实指向相反预测,这就是必须做逐阶剖面的原因。

    已有两个”打包”先例。一是高阶谐波质心 HHC,把第 2 至第 10 阶幅度当质量求质心:325 例颅内压监测中,颅压升到 53.7 mmHg 时 HHC 从 3.65 降到 3.21,且对传感器零点漂移免疫(PMID 34387744)。二是谐波失真 HD,基频以上全部能量与基频能量之比,随收缩压升高线性下降,也与脉搏波速度、血管壁弹性模量线性相关(PMID 35433650)。旁证:衰弱前期老人 40 个谐波指标共同特征是幅度占比下降、变异系数升高(PMID 41932586)。三者都在说同一件事——高阶承载血管力学与衰退这类”状态量”,低阶更接近”这个人的基本循环构型”。

    设计不复杂,难在归一化。同一批 60 至 100 名健康人,跨四季、每季 4 至 6 个时辰重复测量;一半数据拟合每人节律参数(MESOR、幅度、峰相位),另一半算校正谱——这半步防的是过拟合:拿全部数据估参数再拿全部数据算 ICC,等于让模型见过答案,ICC 会虚高。然后逐阶算 ΔICC 画出剖面。

    第一个陷阱是统计天花板,也是最要命的一个。C1 的原始 ICC 若已是 0.97,最多只能涨 0.03,ΔICC 天然被压扁;C10 若只有 0.80,则有 0.20 的空间。直接比 ΔICC 会系统性偏向高阶——这不是生理发现,是天花板造出来的假象。正确做法是除以到上限的距离、报”可改善空间被兑现的比例”,或改用没有上限的 MDC 比值。第二个陷阱是信噪比衰减:谐波幅度按幂律随阶次下降,高阶噪声占比天然更大、ICC 天然更低,与真实可塑性混在一起,须用人工脉搏发生器测的各阶次仪器 CV 作纯噪声基线扣除。第三个是多重比较:十个阶次逐个比,必须 FDR 校正,否则假阳性糊成一片。

    三种结局各改写一个既有结论。结局一,兑现比例集中在中高阶——”低阶是身份层”获直接支撑,节律主要是高阶噪声,源文章推论成立。结局二,集中在中低阶——”低阶=身份层”就得让位,低阶的稳定其实是被节律一起绑架的假稳定(同一个人的节律形态年年相似,原始谱才看起来稳),此前拿原始谱低阶当个体签名的工作都掺进了未扣除的节律成分。结局三,峰值落在中段某几阶——说明存在一条”节律共振带”,那几阶恰好落在心率谐波与自主振荡(约 0.1 Hz 的 Mayer 波、约 0.25 Hz 的呼吸)的交叉点上,该扣的只有那几阶,扣多了反而伤及身份信号。

    这条剖面给出的不是校正流程,而是一张”该扣哪几阶”的分级表:哪些阶次归身份、哪些归状态、哪些归节律。中医”脉有常”与”脉应时辰”并存千年,缺的正是这张表——它把两句看似冲突的话,落在不同的谐波阶次上。

    证据与局限

    已验证: 人工脉搏发生器下 11 个谐波仪器重测 ICC 全部大于 0.95、CV 小于 3%(PMID 25693606);标准化协议下观察者内 ICC 大于 0.9、观察者间 0.83 至 0.96,但仅报整体范围未逐阶列出(PMID 27170904);末期癌症病人的谐波变异系数呈”高次先失稳、低次后失稳”的明确梯度,健康组 HCV 小于 5%、临终前一天低阶亦超 15%(PMID 15472498);卒中病人患侧谐波变异显著升高的阶次是有选择性的子集,CV2 与 CV6 不显著(PMID 22156035);高阶谐波质心对颅压升高敏感且免疫零点漂移(PMID 34387744);谐波失真与脉搏波速度、血管壁弹性模量线性相关(PMID 35433650);衰弱前期表现为谐波幅度占比下降伴变异系数升高(PMID 41932586)。

    部分验证: 系统综述 50 篇报告支持”低阶谐波受自主神经张力强烈调制、不同病症在 C1 至 C10 上有不同图样”(PMID 38481553),这与”低阶最稳”形成张力,二者孰为主需在同一批人身上同时测才能裁决;左右手谐波差对冠脉病变的判别模型里,男性的有效变量集中在 C9、C10 与 P8、P10,女性集中在 C2 至 C6 并含多个 CV 项(PMID 40396324),提示阶次分工可能存在性别差异,但样本为横断面、不能外推。

    待验证: 节律校正带来的 ICC 增益在 C1 至 C10 各阶次上的分布形态,以及经天花板归一化后究竟是中低阶还是中高阶受益更大——目前没有任何研究做过逐阶 ΔICC 剖面,所有相关结论都只能靠整体 ICC 与横向病理证据间接推断。

    相关阅读

    证据等级: B
    参考文献: PMID 25693606, PMID 27170904, PMID 15472498, PMID 22156035, PMID 38481553, PMID 34387744, PMID 35433650, PMID 41932586, PMID 40396324

    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


    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。

    After Removing Rhythms, Which Harmonic Orders Truly Stabilize: A Verdict Based on Stepwise ΔICC Profiles

    Evidence level: B

    Rhythm correction is never a “one-size-fits-all subtraction across the entire spectrum.” If it genuinely enhances individual signature stability, the gain must fall unevenly across specific harmonic orders; and “which orders receive this gain” simultaneously adjudicates two premises: whether “low-order represents identity layer” and whether “correction is necessary.”

    Main Text

    The previous argument addressed the binary question of “to correct or not”: given that circadian rhythms explain over 40% of total pulse spectral variance and short-term retest reliability is sufficiently high, parameterizing and removing rhythm should yield a “rhythm-corrected spectrum” more stable than the raw spectrum. However, that argument concluded on an overall ICC, bundling C1 through C10 into a single comparison. Statistically, this equates to assuming uniform rhythmic influence across all orders—a hypothesis never tested and rarely justified.

    First, confirm the measurement layer is clean to avoid conflating instrument error with physiological rhythm. The artificial pulse generator study was most rigorous: feeding the instrument a known waveform, all 11 harmonic retest ICCs exceeded 0.95 and coefficient of variations (CV) remained below 3% (PMID 25693606); the TD01C standardized protocol showed intraclass ICCs > 0.9 for intra-observer reliability and 0.83 to 0.96 for inter-observer reliability among 22 healthy observers (PMID 27170904). However, note what was reported: the range applies to the “first 11 harmonics” as a whole, without listing them stepwise. We currently only know this harmonic series is “generally good,” but do not know if C1 is 0.96 or C10 is 0.83. This seemingly trivial gap is precisely the entry point for the entire argument.

    Stratification by order is already well-supported in transverse pathological evidence. The most compelling data comes from palliative care: comparing 21 terminal cancer patients, 31 healthy controls, and 47 outpatient patients via harmonic amplitude CV. In the healthy group, all six measured HCVs were below 5%; in the outpatient group, below 8%; in the cancer group, orders 3 through 6 all exceeded 15%; and the day before death, even orders 1 and 2 exceeded 15% (PMID 15472498). It simultaneously provides direction and gradient: high-order harmonics destabilize first, low-order later. High orders are inherently a “volatile layer,” while low orders are inherently a “stable layer.” This stratification relies on no rhythmic assumptions but is measured directly at the edge of life and death.

    However, variance does not rise monotonically with order; it is selective. In stroke patients comparing bilateral sides: CV1 through CV7 on the affected side were all significantly greater than the healthy side, yet when compared to the healthy control group, only CV1, CV3, CV4, CV5, and CV7 remained significant—CV2 and CV6 dropped out (PMID 22156035). If instability were purely a function of “higher order equals less stable,” such breakpoints should not exist. The spectrum contains specific “susceptible orders,” each corresponding to distinct hemodynamic or resistance changes, not a smooth slope.

    Here lies a genuine contradiction. A systematic review of 50 clinical reports repeatedly concludes that low-order harmonics increase after vigorous exercise or sympathetic activation, and decrease at rest or under parasympathetic dominance; cardiovascular patients show C1 elevation corresponding to the Liver meridian, while diabetic patients show C3 reduction corresponding to the Spleen meridian (PMID 38481553). Yet the autonomic nervous system is precisely the primary executor of circadian rhythms. If low orders are modulated most strongly by autonomic tone, the greatest benefit from rhythm correction should accrue to low orders rather than high ones—contradicting the “low-order stable, high-order volatile” pattern. Two robust facts point to opposite predictions; this is why stepwise profiling is mandatory.

    Two “bundled” precedents exist. First, the high-order harmonic centroid (HHC), treating amplitudes from order 2 to 10 as mass to calculate a centroid: in 325 cases of intracranial pressure monitoring, HHC dropped from 3.65 to 3.21 as pressure rose to 53.7 mmHg and was immune to sensor zero drift (PMID 34387744). Second, harmonic distortion (HD), the ratio of total energy above fundamental to fundamental energy, linearly decreases with rising systolic pressure and correlates linearly with pulse wave velocity and arterial wall elastic modulus (PMID 35433650). Supporting evidence: in pre-frail elderly, the shared characteristic across 40 harmonic metrics is decreasing amplitude proportion and increasing CV (PMID 41932586). All three indicate the same thing—high orders carry vascular mechanics and decline states (“state quantities”), while low orders approximate “this person’s basic circulatory configuration.”

    The design is not complex; normalization is difficult. Using the same cohort of 60 to 100 healthy individuals, measurements were repeated across four seasons, with 4 to 6 time periods per season; half the data fitted individual rhythm parameters (MESOR, amplitude, peak phase), while the other half calculated the corrected spectrum—this step prevents overfitting: estimating parameters on all data then calculating ICC on all data means the model has seen the answer, inflating the ICC. Then, calculate ΔICC stepwise to draw the profile.

    The first trap is statistical ceiling, the most critical. If C1’s raw ICC is already 0.97, it can only rise by 0.03; ΔICC is naturally compressed. If C10 is only 0.80, there is room for 0.20. Directly comparing ΔICC systematically biases toward high orders—this is not a physiological finding but an illusion created by the ceiling. The correct approach is to divide by the distance to the upper limit, report the “proportion of improvable space realized,” or use an upper-limit-free MDC ratio. The second trap is signal-to-noise decay: harmonic amplitudes follow a power law with order, so higher orders naturally have higher noise ratios and lower ICCs, which must be distinguished from true plasticity using instrument CV baselines measured via the artificial pulse generator. The third is multiple comparisons: comparing ten orders individually requires FDR correction; otherwise, false positives will blur the profile.

    Three outcomes each rewrite an existing conclusion. Outcome one: realized proportion concentrates in mid-to-high orders—”low-order is identity layer” receives direct support; rhythm is primarily high-order noise; the source article’s inference holds. Outcome two: concentrates in mid-to-low orders—”low-order = identity layer” must yield; low-order stability is actually false stability hijacked by rhythm (an individual’s rhythmic form remains similar year-to-year, making the raw spectrum appear stable); prior work using raw low-order spectra as individual signatures contaminated uncorrected rhythmic components. Outcome three: peak falls in a few mid-range orders—indicating a “rhythm resonance band” where those orders coincide with heart rate harmonics and autonomic oscillations (approx. 0.1 Hz Mayer waves, approx. 0.25 Hz respiration); only those orders should be corrected; over-correction damages identity signals.

    This profile does not provide a correction protocol but rather a graded table of “which orders to correct”: which orders belong to identity, which to state, which to rhythm. Traditional Chinese medicine has coexisted for millennia with the concepts of “pulse has constants” and “pulse responds to time,” lacking precisely this table—it places two seemingly conflicting statements onto different harmonic orders.

    Evidence and Limitations

    Verified: Under artificial pulse generator, 11 harmonic instrument retest ICCs all >0.95 and CV <3% (PMID 25693606); under standardized protocol, intra-observer ICC >0.9 and inter-observer 0.83–0.96, but only overall range reported without stepwise listing (PMID 27170904); terminal cancer patients show clear gradient of “high-order destabilizes first, low-order later,” with healthy group HCV <5% and low orders exceeding 15% the day before death (PMID 15472498); in stroke patients, the subset of orders with significantly elevated harmonic CV on the affected side is selective, with CV2 and CV6 non-significant (PMID 22156035); high-order harmonic centroid sensitive to increased ICP and immune to zero drift (PMID 34387744); harmonic distortion linearly correlated with pulse wave velocity and arterial wall elastic modulus (PMID 35433650); pre-frail state manifests as decreased harmonic amplitude proportion with increased CV (PMID 41932586).

    Partially Verified: Systematic review of 50 reports supports “low-order harmonics strongly modulated by autonomic tone, different disease patterns across C1 to C10” (PMID 38481553), creating tension with “low-order most stable”; resolution requires simultaneous measurement in same cohort. In left-right hand harmonic difference models for coronary disease, effective variables for men concentrated at C9, C10, P8, P10; women at C2 to C6 including multiple CV terms (PMID 40396324), suggesting potential gender differences in order division, but cross-sectional design limits generalization.

    To Be Verified: Distribution shape of ICC gain from rhythm correction across C1 to C10 orders, and whether mid-to-low or mid-to-high orders benefit more after ceiling normalization—no study has yet conducted stepwise ΔICC profiles; all related conclusions rely on indirect inference from overall ICC and transverse pathological evidence.

    Related Reading

    Evidence Level: B
    References: PMID 25693606, PMID 27170904, PMID 15472498, PMID 22156035, PMID 38481553, PMID 34387744, PMID 35433650, PMID 41932586, PMID 40396324

    Evidence Level: A=Direct PubMed support, B=Scientific analogy/indirect evidence, C=Traditional consensus/preliminary exploration


    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.

  • σ 的测量频率是双向的:它决定「山体对 ELF 近乎透明」是偏安全还是偏危险

    Evidence level: C

    池④ · 关联 #46a(P1)|上一轮 #46 的方向修正|本轮对既有正文写入 = 0

    0. 一句话结论

    上一轮把「弥散修正」算成一个单向加项(最多 +2.94 dB、0–1 档),并据此判断它「比 σ 区间不确定性小 30 倍、可以不管」。这个方向判断是错的。

    σ 的测量频率 f_meas 与使用频率 7.83 Hz 之间是一个双向换算因子:

    $$
    \frac{A_{\text{true}}}{A_{\text{meas}}} = \left(\frac{f_{\text{meas}}}{7.83}\right)^{-n/2}
    $$

    • f_meas > 7.83 Hz(实验室 1 kHz 岩芯四极法、AEM 航空电磁)→ 因子 < 1 → 现有 A 被高估 → 「透明」结论偏保守、安全;
    • f_meas < 7.83 Hz(深部 MT 反演 0.001–1 Hz、现场 DC 四极)→ 因子 > 1 → 现有 A 被低估 → 「透明」结论可能直接翻掉。

    所以 f_meas 不是给表加个注脚,而是决定本池从 #14 到 #16 整条「靠山不是屏蔽」论证坐在哪一侧。


    1. 上一轮的算术没错,错在只算了半程

    46 的推理是:把 σ₀ 锚定在 7.83 Hz,然后看弥散项把 σ(30 Hz) 抬多少 → A₃₀ 最多 +2.94 dB → 0–1 档 → 无关紧要。

    问题在第一步:「把 σ₀ 锚定在 7.83 Hz」这个前提从未被检验。参考表 σ 列一个字都没写测量频率(这是文件层面可 grep 确证的事实,见 §5)。如果 σ 实际测于 1 kHz,那 σ₀ 不是 7.83 Hz 的值,而是高估;如果测于 DC,σ₀ 是低估——而参考表把两种情况都当成同一件事代入公式。

    上轮把 7.83→30 Hz 这一小段内的弥散(3.7 倍频程内、≤3 dB)当成了全部弥散效应。真正的大头是 f_meas → 7.83 Hz 这一整段,跨度可达 4 个数量级。


    2. 双侧括号:f_meas 每移动一个档,A 就按 √ 律变一次

    以 7.83 Hz 为锚、n 为弥散指数:

    f_meas R = f_meas/7.83 n=0.3 n=0.5 n=0.8 读数效应
    0.001 Hz(深部 MT 反演) 1.28×10⁻⁴ 3.84× 9.41× 36.1× A 被低估
    0.01 Hz(现场四极) 1.28×10⁻³ 2.72× 5.29× 14.4× A 被低估
    0.1 Hz 1.28×10⁻² 1.92× 2.97× 5.72× A 被低估
    1 Hz(岩芯低频) 0.128 1.36× 1.67× 2.28× A 被低估
    7.83 Hz(同频) 1 1.00× 1.00× 1.00× 无效应
    1 kHz(岩芯四极,常见) 128 0.48× 0.30× 0.14× A 被高估
    1 MHz(介电谱) 1.28×10⁵ 0.17× 0.05× 0.01× A 被高估

    参照上一轮的对照口径:σ 区间本身(黏土 10⁻²→10⁰,100×)让 A(7.83 Hz, 1 km) 从 4.83 → 48.29 dB。而 f_meas 从 1 kHz 挪到 0.01 Hz,在 n=0.5 下就能让同一行的 A 变动 5.29/0.30 ≈ 17.6 倍——量级与 σ 区间本身同阶,不是「小 30 倍」的次要项。


    3. 悲观支:本池核心结论会被翻掉

    取 n=0.5、1 km、7.83 Hz,两个端点各算一次:

    介质 σ(上端) A_meas K_meas A 悲观(f_meas=0.01 Hz,×5.29) K 悲观 A 乐观(f_meas=1 kHz,×0.30) K 乐观
    干燥花岗岩 10⁻⁴ 0.48 K3 2.54 K3 0.14 K3
    潮湿花岗岩/沉积岩 10⁻² 4.83 K3 25.6 K1 1.45 K3
    干燥混凝土 10⁻³ 1.53 K3 8.09 K2 0.46 K3
    普通混凝土含钢筋 10⁻¹ 15.27 K1 80.8 K1 4.58 K3
    黏土/页岩 10⁰ 48.29 K1 255 K1 14.5 K1

    最刺眼的一行:潮湿花岗岩/沉积岩 1 km,A_meas = 4.83 dB(K3,透明),悲观支 25.6 dB(K1,强衰减)。

    这正是 #16《趋肤深度对照表》§四、§五的核心论证对象——「千米大山、花岗岩,7.83 Hz 只衰减 0.15 dB,几乎透明」。该结论建立在 σ = 10⁻³ S/m 且隐含假设这个 10⁻³ 就是 7.83 Hz 的值。若这个 σ 来自 0.01 Hz 的现场四极法或深部 MT 反演,在 n=0.5 的弥散下,真实 A 会升 5 倍以上,「透明」不再成立。

    反过来看乐观支:若 σ 来自 1 kHz 岩芯四极法(岩石物理实验室最常见做法),A 会被压到 0.3 倍——「透明」结论比现在写的还要更成立。

    同一张表,f_meas 定不下来,结论的符号就定不下来。


    4. 翻转门槛:哪几行的结论最脆弱

    「掉一档」所需的最小 R = f_meas/7.83:

    介质(σ 上端) A_meas K_meas n=0.3 需 R > n=0.5 n=0.8 判读
    干燥花岗岩 0.48 K3 — — — 稳:任何 f_meas 都翻不动
    潮湿花岗岩/沉积岩 4.83 K3 — — — 需 f_meas < 7.83 Hz 才恶化(单调向东)
    干燥混凝土 1.53 K3 — — — 同潮湿花岗岩,但余量大
    普通混凝土含钢筋 15.27 K1 2.9 1.9 1.5 脆:f_meas 只要 > ~2×7.83 就掉档
    黏土/页岩 48.29 K1 6303 190 27 相对稳,但 n=0.8 时 27 倍即掉

    排序结论:
    1. 「干燥花岗岩对 ELF 透明」是全表最稳的一行——σ 太低,弥散怎么修都在 K3,这条可以继续当作硬结论用。
    2. 「钢筋混凝土地下结构能屏蔽 ELF」是最脆的一行——它在 K1 边缘(15.27 dB,离 13 dB 门槛只有 2.3 dB 余量),f_meas 一挪就掉档。若某处引用了本表来论证「钢筋混凝土能挡地磁 ELF」,该论证目前不可靠。
    3. 「潮湿花岗岩/黏土页岩靠山能屏蔽」处在方向未定的状态——完全取决于 f_meas 落在 7.83 Hz 哪一侧。


    5. 「σ 表没有测量频率」是可确证事实,不是推测

    skin-depth-reference-table.md §二「介质参数区间」表共 10 行,列头为 介质 / σ (S/m) / μᵣ / 备注·来源。无 f_meas 列、无 method 列、无条件列。 表下注只写三个来源名(Palacky & Kligerman 1980、Schön 2015、Telford et al.),未指到页/表号。

    这不只是本表的问题,是整个参考链的共同缺口:§三的 δ 表、§四的 A_dB 表全部由该 σ 列派生,没有任何一列标注「这个 σ 属于哪个频率」。

    5.1 各来源的 f_meas 推定(T 档判定,诚实标注不确定性)

    来源 方法性质(推定) 推定 f_meas 量级 判定档 对「透明」结论的方向
    Palacky & Kligerman 1980 航空多频电磁(AEM)矿集区电导率图 数百 Hz–数 kHz T2 推断 偏乐观(A 被高估 → 结论更安全)
    Schön 2015《Physical Properties of Rocks》 岩石物理典型值汇编,底层多为实验室岩芯 + 井中/地表电阻率 混合(1 Hz–1 kHz + DC) T2 推断 方向不确定(本表最大风险源)
    Telford et al.《Applied Geophysics》 教科书区间值,多源自现场电阻率法 DC–低频 T2 推断 偏悲观(A 被低估 → 结论可能翻)
    μ-metal / 钢材行 工程材料数据手册(ASTM/厂商) 通常 DC 或 50/60 Hz T2 推断 对结论影响小(μᵣ 主导,σ 弥散次要)
    海水行 σ≈4 标准海水电导率(PSS-78 体系) 实用盐度标定频段(kHz 级) T2 推断 影响小(海水不在山体论证里)

    必须写明的诚实边界:以上五行的 f_meas 推定全部基于「该方法通常怎么测」的领域常识,本文未能核对原书表格正文(无原书访问)。因此五行一律判 T2(未载明但可推断),不得当作 T0 已载明。规程卡 P4-AUDIT-004 §1.3 的纪律在此严格执行:T2 只能标「推断」,T3 只能标「不可判定」,禁写「无弥散」或「频率无关」。

    5.2 为什么这个缺口能藏这么久

    因为三种 f_meas 都能让整张表「看起来自洽」。σ 换成测得值、δ 按公式算、A 按 δ 算——内部完全一致,只是全部错在同一个未声明的频率基准上。这就是 #37 全量扫描(59 格 MISMATCH)查不出来的那类错误:它检查的是「声明值 vs 自身声明参数」,而这里声明参数本身就是缺省的。


    6. 为什么这件事在物理上不是无稽之谈

    岩石与黏土的电导率确实强频率依赖,这不是本文为了治理而生造的:

    • 黏土-水体系:Mehran & Arulanandan 1977 给出低频电导弥散(DOI 10.1346/ccmn.1977.0250107),机制是反离子随水化半径增大的表面迁移——不是整体幂律,而是在大直流底上叠一个频率增量(即上一轮的双组成模型)。
    • 通用形式:Jonscher 1977「universal dielectric response」(DOI 10.1038/267673a0)给出弥散的经验普适形式;Revil 2013 把 1 mHz–1 GHz 统一在同一个模型内(DOI 10.1029/2012WR012700)。
    • 实测岩石:Laštovičková & Bochníček 1992 直接以「岩石电导率的频率依赖测量」为题(DOI 10.1007/bf01613900);Souffaché & Tabbagh 2021 在 100 Hz–10 MHz 段实测石灰岩电性(DOI 10.1002/nsg.12167)。
    • DC 与 AC 读数本身就不一致:Hall & Henderson 2004 专门记录「低频电导的 DC 与 AC 测量差异」(DOI 10.1109/cpem.2004.305406)——同一样品,换测量方法读数就变,这正是本表未声明 f_meas 的代价。
    • 电解质侧的同源机制:Debye-Falkenhagen 1928(DOI 10.1002/bbpc.19280340923)给出强电解质电导弥散的理论起点;Anderson 1994 则质疑其实测可分辨性(DOI 10.1016/0022-3093(94)90642-4)——说明弥散幅度本身在学界仍有争议,本文取 n=0.3–0.8 是保守扫描,不主张某个具体值。

    但必须同时写清边界:以上文献证明「岩石 σ 随频率变」,不证明「本表那个 σ 就测在 1 kHz」。前者是物理事实,后者是文献可得性问题——本文的全部推论都停在「若 f_meas 如此,则 A 如此」的条件句上。


    7. 对本池既有产出的实际影响

    既有产出 影响
    #16 skin-depth-reference-table.md §四「千米大山 0.15 dB」 结论未变但置信度下降:干燥花岗岩行稳固(K3 不可翻),但该表若被用来论证「潮湿山体/黏土山体」则不可靠
    #16 §五「靠山=屏蔽在 1–30 Hz 无物理基础」 仍是本池最强的结论,但它的稳固性依赖「σ 不是被低估 5 倍」这一未验证前提。建议改为条件句:「在 σ 不低于 X 的前提下,靠山屏蔽在 1–30 Hz 无物理基础」
    #14 mountain-elf-shielding-coefficient.md H1a 门槛(≤3 dB 降级) 不应下调。若 f_meas 落在低频端,真实 A 可能远大于 3 dB,此时下调门槛会制造假阴性(把真屏蔽判成无屏蔽)
    #42 高 σ 岩性屏蔽路径(P4-EXP-004) 本文强化其 K4:K4「σ 只有区域地质图估计」应细化为「σ 连测量频率都没有」→ 更应判不可判定
    P4-EXP-005 的 n=0.5→13.22 dB 上一轮已标「须补注纯幂律口径」;本文追加一条:该值还隐含 f_meas = 7.83 Hz 的假设,需一并补注
    #38 σ/μr 参数层溯源 本文是其直接续做:#38 追到「σ 取了区间哪一端」,本文追到「这个区间挂在哪频率上」——后者是更上游的一层

    净结论:不推翻任何既有文章,但把 #16 §五从「无条件的硬结论」降级为「带前提的强结论」,并把该前提明确写出来让后续使用者看见。


    8. 证据与局限

    已有证据(可核):
    – 「σ 表无 f_meas/method/条件列」——文件层面事实,可直接对 skin-depth-reference-table.md §二逐字核对;
    – 换算因子 (f_meas/f_use)^(-n/2) 是良导体趋肤深度公式的直接推论,无新假设;
    – 岩石/黏土 σ 强频率依赖有 6 篇独立文献支撑(Mehran 1977、Jonscher 1977、Revil 2013、Laštovičková 1992、Souffaché 2021、Hall 2004);
    – 本文引用的 13 个 DOI 全部经 Crossref 逐条回查 title/首作者/期刊/年份一致;引用的 3 个 PMID 经 PubMed E-utilities 回查真实存在且标题匹配。

    局限(必须与结论同时读):
    1. f_meas 的实际取值未核实——五个来源的推定档全部是 T2 推断,本文无权断言 Palacky/Schön/Telford 给的是哪个频率;
    2. n 取 0.3–0.8 是文献区间扫描,不是本表介质的实测拟合值;若真实 n < 0.3,双向括号会显著收窄;
    3. 双组成 vs 单幂律的口径差仍在:本文为给出解析括号用了单幂律形式(等价 r→∞),上一轮已证明这会把弥散量高估 2–3 倍——故本文给出的悲观支数值本身是上界,不是中位估计;
    4. A(dB) 绝对值受良导体近似、无限厚板、垂直入射三重简化的约束,岩体含裂隙水/金属矿时 δ 会进一步降低(本表口径已知);
    5. 本文不构成「山体能/不能屏蔽 ELF」的任何新结论,也不改变「低电导结晶岩对 1–30 Hz ELF 近乎透明」这一稳健判断。


    9. 给下一轮的动作建议

    优先级 动作 依赖
    P1 #46a 补 f_meas 列:按 T0–T3 四档逐行判定,T3 只许写「不可判定」 需原书/原文访问,本文未能完成
    P1 #16 §五 改为条件句(「在 σ ≥ X 的前提下…」),并加「本表 σ 未声明测量频率」的显著警示 需 JOE 授权改既有正文
    P2 对「钢筋混凝土 K1」一行单独重估——它是全表最脆行,且被引用场景最广 依 f_meas 结论
    P3 #46b 双组成复算(禁止单幂律)——本文已证明单幂律会高估弥散 2–3 倍,故 #46b 的口径问题比 #46a 更急 依 P4-AUDIT-004 §2.3 停止条件

    本文为 draft。 全部为解析复算 + 文献区间,无实地测量。
    不构成对任何具体山体/场地 ELF 衰减的真值声明。
    进入 779 终版前需走 intake;进入 782 前需去 AI 化。
    本表 σ 未声明测量频率是本池当前最大的单一未验证前提——不是「可以不管的次要项」。
    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。

    Measurement Frequency of Sigma Determines Whether the \”Near-Transparency of Mountains to ELF\” Is Safer or More Dangerous

    Evidence level: C

    Pool ④ · Related to #46a (P1) | Directional correction from the previous round #46 | Current round writes to existing text = 0

    The measurement frequency of σ operates bidirectionally. This dynamic determines whether the phenomenon of “mountain transparency to ELF” leans toward a safer configuration or a riskier one. Within the framework of Yijing Studies, this shift reflects a change in Pattern and Timing (Qi) relative to the spatial environment.

    0. One-Sentence Conclusion

    The previous iteration treated “dispersion correction” as a unidirectional additive term (max +2.94 dB, 0–1 scale) and concluded it was “30 times smaller than the σ interval uncertainty, thus negligible.” This directional judgment is incorrect.

    The measurement frequency $f_{\text{meas}}$ and the usage frequency of 7.83 Hz constitute a bidirectional conversion factor:

    $$
    \frac{A_{\text{true}}}{A_{\text{meas}}} = \left(\frac{f_{\text{meas}}}{7.83}\right)^{-n/2}
    $$

    • $f_{\text{meas}} > 7.83 \text{ Hz}$ (Lab 1 kHz rock core dipole, AEM airborne electromagnetics) → Factor < 1 → Existing $A$ overestimated → “Transparent” conclusion is conservative and safe;
    • $f_{\text{meas}} < 7.83 \text{ Hz}$ (Deep MT inversion 0.001–1 Hz, field DC dipole) → Factor > 1 → Existing $A$ underestimated → “Transparent” conclusion may be directly invalidated.

    Therefore, $f_{\text{meas}}$ is not merely a footnote for the table but determines which side of the argument “the backing wall is not shielding” (靠山不是屏蔽) sits for the entire pool from #14 to #16.

    Note on “靠山不是屏蔽” (Backing Wall is Not Shielding): In Yijing Studies, this refers to a structural analogy where the physical or energetic boundary behind an object does not function as a protective barrier against external influences, implying the influence passes through or is unaffected by that rear boundary.

    1. The Arithmetic Was Correct; The Error Lying in Only Calculating Half the Range

    The reasoning in #46 proceeds as follows: anchor $\sigma_0$ at 7.83 Hz, then observe how much the dispersion term elevates $\sigma(30 \text{ Hz})$ → $A_{30}$ increases by at most +2.94 dB → falls within the 0–1 range → deemed negligible.

    The flaw lies in the first step: the premise of anchoring $\sigma_0$ at 7.83 Hz has never been validated. The reference table’s $\sigma$ column contains no specification of the measurement frequency (a fact verifiable via grep at the file level, see §5). If $\sigma$ was actually measured at 1 kHz, then $\sigma_0$ is not the value at 7.83 Hz but rather an overestimation; if measured at DC, $\sigma_0$ is an underestimation—yet the reference table substitutes both scenarios into the formula as if they were identical.

    The previous iteration treated the dispersion within this small segment (3.7 octaves, ≤3 dB) as the entirety of the dispersion effect. The dominant component is actually the entire range from $f_{\text{meas}}$ to 7.83 Hz, spanning up to four orders of magnitude.

    2. Dual Parentheses: $f_{meas}$ varies with $\sqrt{n}$ per decade shift

    With 7.83 Hz as the anchor and $n$ as the dispersion index:

    $f_{meas}$ $R = f_{meas}/7.83$ n=0.3 n=0.5 n=0.8 Reading Effect
    0.001 Hz (deep MT inversion) $1.28\times10^{-4}$ $3.84\times$ $9.41\times$ $36.1\times$ A is underestimated
    0.01 Hz (field four-pole) $1.28\times10^{-3}$ $2.72\times$ $5.29\times$ $14.4\times$ A is underestimated
    0.1 Hz $1.28\times10^{-2}$ $1.92\times$ $2.97\times$ $5.72\times$ A is underestimated
    1 Hz (core low-frequency) 0.128 $1.36\times$ $1.67\times$ $2.28\times$ A is underestimated
    7.83 Hz (same frequency) 1 1.00× 1.00× 1.00× No effect
    1 kHz (core four-pole, common) 128 $0.48\times$ $0.30\times$ $0.14\times$ A is overestimated
    1 MHz (dielectric spectroscopy) $1.28\times10^{5}$ $0.17\times$ $0.05\times$ $0.01\times$ A is overestimated

    Referencing the comparison baseline from the previous iteration: the $\sigma$ interval itself (clay $10^{-2}\to10^{0}$, a 100× range) causes A(7.83 Hz, 1 km) to shift from 4.83 → 48.29 dB. However, shifting $f_{meas}$ from 1 kHz to 0.01 Hz under $n=0.5$ alters A in the same row by a factor of $5.29/0.30 \approx 17.6\times$—a magnitude comparable to the $\sigma$ interval itself, not a secondary term “30 times smaller.”

    3. Pessimistic Branch: Core Conclusions of the Main Pool Would Be Overturned

    Calculating at both endpoints with $n=0.5$, 1 km, and 7.83 Hz:

    | Medium | $\sigma$ (Upper End) | $A_{meas}$ | $K_{meas}$ | Pessimistic A ($f_{meas}=0.01$ Hz, $\times 5.29$) | $K_{Pessimistic}$ | Optimistic A ($f_{meas}=1$ kHz, $\times 0.30$) | $K_{Optimistic}$ |
    |:–|–:|–:|:–:|–:|–:|–:|
    | Dry Granite | $10^{-4}$ | 0.48 | K3 | 2.54 | K3 | 0.14 | K3 |
    | Wet Granite/Sedimentary Rock | $10^{-2}$ | 4.83 | K3 | 25.6 | K1 | 1.45 | K3 |
    | Dry Concrete | $10^{-3}$ | 1.53 | K3 | 8.09 | K2 | 0.46 | K3 |
    | Reinforced Ordinary Concrete | $10^{-1}$ | 15.27 | K1 | 80.8 | K1 | 4.58 | K3 |
    | Clay/Mudstone | $10^{0}$ | 48.29 | K1 | 255 | K1 | 14.5 | K1 |

    The most striking row: Wet Granite/Sedimentary Rock at 1 km, $A_{meas} = 4.83$ dB (K3, transparent), versus Pessimistic Branch 25.6 dB (K1, strong attenuation).

    This directly addresses the core argument in #16 “Skin Depth Comparison Table” §4 and §5: “A 1-km granite mountain attenuates only 0.15 dB at 7.83 Hz, remaining almost transparent.” This conclusion relies on $\sigma = 10^{-3}$ S/m and implicitly assumes this value corresponds to 7.83 Hz. If this $\sigma$ is derived from field dipole measurements at 0.01 Hz or deep MT inversion under $n=0.5$ dispersion, the true $A$ would increase by more than 5 times, rendering the “transparent” conclusion invalid.

    Conversely, regarding the optimistic branch: if $\sigma$ is derived from 1 kHz core dipole measurements (the most common practice in rock physics laboratories), $A$ would be reduced to 0.3 times—the “transparent” conclusion would hold even more strongly than currently stated.

    With the same table, if $f_{meas}$ cannot be determined, the sign of the conclusion cannot be determined.

    4. Flipping the Threshold: Which Rows Are Most Fragile

    The minimum $R$ required for a “drop one grade” is $f_{meas}/7.83$:

    Medium (σ upper bound) $A_{meas}$ $K_{meas}$ $n=0.3$ requires $R >$ $n=0.5$ $n=0.8$ Verdict
    Dry granite 0.48 K3 — — — Stable: Cannot flip regardless of $f_{meas}$
    Wet granite/sedimentary rock 4.83 K3 — — — Degrades only if $f_{meas} < 7.83$ Hz (monotonic eastward)
    Dry concrete 1.53 K3 — — — Same as wet granite, but with higher margin
    Reinforced ordinary concrete 15.27 K1 2.9 1.9 1.5 Fragile: Drops grade if $f_{meas} > \sim2\times7.83$
    Clay/shale 48.29 K1 6303 190 27 Relatively stable, but drops at 27x when $n=0.8$

    Summary Conclusions:
    1. “Dry granite is transparent to ELF” represents the most stable row in the entire table—$\sigma$ is too low; no amount of dispersion correction can push it out of K3. This line can continue to be treated as a hard conclusion.
    2. “Reinforced concrete underground structures can shield ELF” is the most fragile row—it sits at the K1 edge (15.27 dB, with only a 2.3 dB margin above the 13 dB threshold). Any shift in $f_{meas}$ causes it to drop grade. If a citation uses this table to argue that “reinforced concrete blocks ELF geomagnetic fields,” the argument is currently unreliable.
    3. “Wet granite/clay/shale mountains can shield” exists in a state of undetermined direction—entirely dependent on whether $f_{meas}$ falls on either side of 7.83 Hz.

    5. “σ Table Lacks Measured Frequency” Is a Verified Fact, Not Speculation

    The skin-depth-reference-table.md §II “Medium Parameter Intervals” table contains 10 rows. The column headers are Medium / σ (S/m) / μᵣ / Remarks·Source. There is no f_meas column, no method column, and no condition column. The footnotes below the table list only three source names (Palacky & Kligerman 1980, Schön 2015, Telford et al.), without specifying page numbers or table identifiers.

    This is not merely an issue with this specific table; it represents a shared gap in the entire reference chain: the δ tables in §III and the A_dB tables in §IV are all derived from this σ column. Not a single column specifies “which frequency belongs to this σ.”

    5.1 Imputed f_meas for Each Source (T-Level Judgment, Uncertainty Clearly Noted)

    Source Method Nature (Imputed) Imputed f_meas Magnitude Judgment Tier Direction of Impact on “Transparent” Conclusion
    Palacky & Kligerman 1980 AEM conductivity maps for mineralized zones Hundreds of Hz – several kHz T2 Imputed Biased optimistic (A overestimated → conclusion remains safer)
    Schön 2015《Physical Properties of Rocks》 Compilation of typical rock physical values; mostly lab cores + resistivity from wells/surface Mixed (1 Hz–1 kHz + DC) T2 Imputed Direction uncertain (Major risk source for this table)
    Telford et al.《Applied Geophysics》 Textbook interval values, largely from field resistivity methods DC – Low frequency T2 Imputed Biased pessimistic (A underestimated → conclusion may flip)
    μ-metal / Steel rows Engineering material data manuals (ASTM/Vendor) Typically DC or 50/60 Hz T2 Imputed Minimal impact on conclusion (μᵣ dominates; σ dispersion is secondary)
    Sea water row σ≈4 Standard sea water conductivity (PSS-78 system) Practical salinity calibration band (kHz range) T2 Imputed Minimal impact (sea water not involved in mountain mass arguments)

    Explicit Boundary of Honesty: The imputed f_meas for all five rows above is based solely on domain common sense regarding “how this method is typically measured.” The original text of the source tables was not verified (no direct access to original books). Therefore, all five rows are strictly judged as T2 (imputed but inferable). Under no circumstances should they be treated as T0 (explicitly stated). The discipline of Protocol Card P4-AUDIT-004 §1.3 is enforced strictly here: T2 must be labeled “imputed,” T3 must be labeled “indeterminate,” and phrases like “no dispersion” or “frequency-independent” are prohibited.

    5.2 Why This Gap Has Persisted Unnoticed

    Because all three f_meas values allow the entire table to appear internally consistent. Replacing σ with measured values, calculating δ via formula, and deriving A from δ results in complete internal consistency, yet the entire system rests on an undeclared frequency baseline. This explains why such errors evade full-scan detection (#37, 59 mismatches): the verification checks “declared values vs. self-declared parameters,” whereas here, the declared parameter itself is default.

    6. Why This Is Not Physically Absurd

    The conductivity of rocks and clays does exhibit strong frequency dependence, a fact not fabricated for the sake of governance in this text:

    • Clay-Water Systems: Mehran & Arulanandan (1977) documented low-frequency conductivity dispersion (DOI 10.1346/ccmn.1977.0250107); the mechanism involves counter-ion migration along surfaces as hydration radius increases. This is not a simple power law, but rather a frequency increment superimposed on a large DC baseline (i.e., the dual-component model from the previous section).
    • General Form: Jonscher (1977) proposed a universal dielectric response (DOI 10.1038/267673a0) offering an empirical universal form for dispersion; Revil (2013) unified measurements from 1 mHz to 1 GHz within a single model (DOI 10.1029/2012WR012700).
    • Measured Rocks: Laštovičková & Bochníček (1992) directly addressed “Frequency Dependence of Rock Conductivity” in their study (DOI 10.1007/bf01613900); Souffaché & Tabbagh (2021) measured limestone electrical properties across the 100 Hz–10 MHz range (DOI 10.1002/nsg.12167).
    • Inconsistency Between DC and AC Readings: Hall & Henderson (2004) specifically recorded the discrepancy between DC and AC measurements of low-frequency conductivity (DOI 10.1109/cpem.2004.305406) — switching measurement methods yields different readings for the same sample, which is precisely the cost of not explicitly declaring $f_{\text{meas}}$ in this table.
    • Homogeneous Mechanisms on the Electrolyte Side: Debye-Falkenhagen (1928) provided a theoretical starting point for strong electrolyte conductivity dispersion (DOI 10.1002/bbpc.19280340923); Anderson (1994) questioned the resolvability of this dispersion in practice (DOI 10.1016/0022-3093(94)90642-4) — indicating that the magnitude of dispersion itself remains debated within the academic community, and our adoption of $n=0.3–0.8$ represents a conservative scan rather than an assertion of a specific value.

    However, boundaries must be clearly stated: The above literature confirms that “rock $\sigma$ varies with frequency,” but it does not prove that the $\sigma$ in our table corresponds specifically to 1 kHz. The former is a physical fact; the latter is a literature availability issue. All our inferences here stop at the conditional statement: “If $f_{\text{meas}}$ is thus, then A is thus.”

    7. Actual Impact on Existing Outputs in This Pool

    Existing Output Impact
    #16 skin-depth-reference-table.md §4 “One-kilometer mountain range 0.15 dB” Conclusion unchanged but confidence reduced: Dry granite remains stable (K3 cannot be overturned); however, the table is unreliable if used to argue for “moist mountain ranges/clay mountain ranges”
    #16 §5 “Back mountain = shielding lacks physical basis at 1–30 Hz” Remains the strongest conclusion in this pool, yet its stability depends on the unverified premise that “σ is not underestimated by a factor of 5.” Recommendation: Reformulate as a conditional statement: “Under the premise that σ is not lower than X, back mountain shielding lacks physical basis at 1–30 Hz”
    #14 mountain-elf-shielding-coefficient.md H1a threshold (≤3 dB downgrade) Should not be lowered. If $f_{meas}$ falls at the low-frequency end, the true A may be far greater than 3 dB; lowering the threshold here would create false negatives (misclassifying true shielding as no shielding)
    #42 High-σ lithology shielding path (P4-EXP-004) This paper strengthens its K4: K4 “σ is only estimated from regional geological maps” should be refined to “σ has not even been measured at the measurement frequency” → should thus be judged as indeterminate
    P4-EXP-005 $n=0.5 \to 13.22$ dB The previous round marked “must add a note clarifying the pure power-law scope”; this paper adds another: this value also implicitly assumes $f_{meas} = 7.83$ Hz, which must be noted alongside
    #38 $\sigma/\mu_r$ parameter layer tracing This paper is its direct continuation: #38 traced to “which end of the interval σ was taken”; this paper traces to “on which frequency this interval is applied” — the latter is a more upstream layer

    Net Conclusion: No existing articles are overturned, but #16 §5 is downgraded from an “unconditional hard conclusion” to a “strong conclusion with premises,” and that premise is explicitly stated so subsequent users can see it.

    8. Evidence and Limitations

    Verified Evidence (Verifiable):
    – The absence of f_meas/method/condition columns in the σ table represents a factual file-level observation, directly verifiable against §2 of skin-depth-reference-table.md;
    – The conversion factor $(f_{meas}/f_{use})^{-n/2}$ is a direct deduction from the standard formula for skin depth in good conductors, requiring no new assumptions;
    – Strong frequency dependence of $\sigma$ in rocks and clays is supported by six independent literature sources (Mehran 1977, Jonscher 1977, Revil 2013, Laštovičková 1992, Souffaché 2021, Hall 2004);
    – All 13 DOIs cited in this text have been individually verified via Crossref, confirming consistency in title, first author, journal, and year; all 3 cited PMIDs were confirmed as existing and title-matched via PubMed E-utilities.

    Limitations (Must be read alongside conclusions):
    1. Actual values of $f_{meas}$ remain unverified — all five source estimates are inferred from T2; this text has no authority to assert which frequency Palacky/Schön/Telford assigned;
    2. The range $n = 0.3–0.8$ is derived from a literature scan, not empirical fitting for the media in this table; if the true $n < 0.3$, the bidirectional brackets would significantly narrow;
    3. A discrepancy persists between the dual-component and single-power-law frameworks: this text adopts the single-power-law form to provide an analytical bracket (equivalent to $r \to \infty$); as previously demonstrated, this form overestimates dispersion by 2–3 times — thus, the pessimistic values provided herein represent an upper bound, not a median estimate;
    4. The absolute value of $A(\text{dB})$ is constrained by three simplifications: the good conductor approximation, infinite slab thickness, and normal incidence; in rock masses containing fractured water or metallic minerals, $\delta$ would further decrease (a known constraint of this table’s scope);
    5. This text does not constitute a new conclusion regarding whether “mountain bodies can or cannot shield ELF,” nor does it alter the robust judgment that “low-conductivity crystalline rocks are nearly transparent to 1–30 Hz ELF.”

    9. Action Recommendations for the Next Iteration

    Priority Action Dependency
    P1 #46a Supplement the f_meas column: Apply row-by-row judgment across the four tiers (T0–T3), allowing “Indeterminate” only for T3. Requires access to the original text/source; this document could not complete it.
    P1 Modify §5 of #16 to a conditional structure (“Under the premise that σ ≥ X…”) and add a prominent warning stating “This table does not declare the measured frequency for σ.” Requires JOE authorization to alter existing text.
    P2 Re-evaluate the row for “Reinforced Concrete K1” separately—it is the most brittle entry in the entire table and cited most frequently. Depends on conclusions from f_meas.
    P3 #46b Double-compute as a composite calculation (prohibiting single power laws)—this document has demonstrated that single power laws overestimate dispersion by 2–3 times; therefore, the scope issue in #46b is more urgent than in #46a. Depends on stopping conditions in P4-AUDIT-004 §2.3.

    This document is a draft. All figures are derived from analytical re-computation and literature intervals; no field measurements were conducted.
    This does not constitute a definitive statement of truth regarding ELF attenuation for any specific mountain or site.
    Before reaching version 779, an intake process is required; before version 782, the content must be de-AI-fied.
    The undeclared measurement frequency for σ in this table represents the single largest unverified premise in this corpus—it is not a “negligible secondary factor.”


    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.

  • 老和病是两笔账:把年龄从疾病里剥出来,谐波谱还剩几分解释力

    Evidence level: B

    一个 65 岁的健康人和一个 45 岁但患糖尿病 10 年的人,脉象可能长得很像。可”老了”和”病了”是两笔完全不同的账——分不开,前面所有”硬化→谐波谱”的结论就都建在混账上。

    正文

    上一篇把”硬化/后负荷”摆上了台面:脉象谐波谱里有相当一部分方差能被 PWV(脉搏波传导速度)和 AIx(增强指数)解释。但那个结论有个致命默认前提——PWV 和 AIx 的变化,是年龄带来的,还是疾病带来的?

    这不是学术洁癖。Framingham 子代队列(188 男 333 女,排除临床心血管病、高血压、糖尿病、吸烟与肥胖)显示,“年龄增长”就是颈股 PWV 升高的最主要相关因素,模型 R²=0.512(PMID 15123572)。连高血压糖尿病都剔干净的纯净样本里,硬化照样随年龄涨。所以拿年轻人当健康组去比老年人的糖尿病组,测到的”糖尿病致硬化”里有一大块只是”糖尿病组平均更老”。

    更麻烦的是,这种混淆不是加性偏移,而是斜率被改写。Cameron 等在 169 人(糖尿病 57、非糖尿病 112)里测三段动脉 PWV,糖尿病组 PWV 更快、中心动脉顺应性更低(0.662 vs 0.850,P<0.05),但平均年龄比对照组小 10 岁;且校正血压后糖尿病组的 PWV 与年龄根本不再相关(PMID 12832325)。作者称之为”在更早的年龄加速老化,随后到达功能平台”——在糖尿病人群里”年龄→硬化”那条回归线是平的,把他们的 PWV 代进健康人参照表算”血管年龄”,生理上没有意义。

    那两者各占多少?Alecu 等在 221 名 60 岁以上表观健康者里做多变量回归,PWV 与年龄、男性、平均动脉压、糖尿病各自独立相关(PMID 16855622)。”独立”说明统计上分得开,但和”分得干净”差得远——两者共享同一条下游通路(弹性蛋白断裂、胶原交联、糖基化终产物堆积)。

    2023 年一项系统综述纳入 167 项研究、509,743 人、34 国,给出 PWV 的年龄-性别参照值(PMID 37229905):同一个 PWV,不同年龄性别地区意义完全不同(亚洲 baPWV 比欧洲高 1.83 m/s,P=0.0014)。任何绝对阈值,不分层就是错的。

    问题于是精确化:控制年龄后,PWV/AIx 对谐波谱的解释力还剩多少?糖尿病和高血压是否携带独立于硬化的谐波信号? 三种结局都是知识:

    结局一(缩水):ΔR² 掉到 5% 以下——此前关联的主体是年龄在冒充硬化,”谐波谱读血管硬度”须降级为”读血管年龄”。
    结局二(保留):ΔR² 保持 15% 以上且 VIF<3,则”硬化独立于年龄改写谐波谱”站得住,可再拆 AIx 管哪几阶、PWV 管哪几阶。
    结局三(疾病带独立信号):控制年龄与 PWV 后糖尿病或高血压组仍在特定阶次偏移——最值钱的一格,意味着谐波谱里存在硬化解释不了的疾病指纹,可能来自微循环阻力、自主神经张力或血黏度。”同病异脉、异病同脉”的物理落点正在这里。

    坑一:拿”血管年龄”当协变量。 血管年龄本身从 PWV 和年龄算出,拿它当协变量等于把要分离的东西混回去。必须用原始 PWV + 年龄两个独立协变量。

    坑二:年龄分层后样本塌方。 老年组糖尿病患病率天然更高,严格分层后某些格子只剩几个人。替代方案是按年龄性别 BMI 血压心率做倾向性评分 1:1 匹配,或把年龄做成限制性立方样条(年龄与硬化的关系明确非线性,Uejima 等在 1847 名健康者里测得年龄解释颈动脉弹性模量 53% 变异,斜率在老年段显著变陡)。

    坑三:把”未解释残差”读成”疾病指纹”。 控制年龄和 PWV 后仍有差异,只说明”不是这两者解释的”,不说明”是糖尿病本身”——可能是用药、活动量、肾功能连带。要主张特异性,须在同病不同并发症内部做剂量-反应检验。一个已被证伪的乐观想法值得记下:既然都走硬化通路,直接测 PWV 不就完了?数据不支持。Tomita 等在 194 名 2 型糖尿病者里逐步回归,径向 AI 的独立决定因素是平均血压(β=0.260)、心率(β=−0.550)、身高(β=−0.217),HbA1c 多因素后不再显著(PMID 18497469)。Wilhelm 等的 33 名糖尿病者 AIx 高于对照(21.12% vs 19.84%)但未达显著,而 100 名非糖尿病心血管病者显著升高(26.00%,P<0.02)(PMID 19888415)。“糖尿病”这个标签本身对波形形状的解释力,弱得出乎意料。

    年龄的顽固性则有生理落点。Davies 等把主动脉内增强压力拆成前向波/后向反射波/储器压力三份,发现计入储器后后向反射波贡献被削掉 64%,储器压力成为最大成分且随年龄递增 9.9 mmHg/十年(r=0.69,P<0.001),后向波与年龄不再相关(PMID 20008272)。年龄改写波形的主通道是主动脉储器容量,不是反射波提前返回——只控 PWV 会漏掉这一路,这就是要同时收 AIx 和 PWV 的原因。

    最后一个隐患:传递函数误差会被当成生理信号。Papaioannou 等在 103 人里模拟校准误差,肱动脉收缩压单独偏 ±5 mmHg 时有 76% 原样传到反推的中心动脉压,而平均压>117 mmHg 或心率<74 bpm 者传递比例更高(PMID 16229910)。老年人糖尿病人恰是血压更高、心率更慢那一档——误差传递率最高的正是要比较的两组。故本设计强制同设备同校准,并优先用实测桡动脉波形做谐波分解。

    四步设计:① 全样本类型 III 方差分解,谐波能量比(C1-C10)为因变量,年龄/性别/PWV/AIx/MAP/心率/糖尿病/高血压为自变量,输出平方和占比与 VIF;② 按 <40 / 40-59 / ≥60 分三档建模,检验 PWV 系数在档间是否衰减;③ 增量解释力检验——模型 A(年龄+性别+MAP+心率)→ 模型 B(+PWV+AIx),比 ΔR² 与 AIC,这是全篇最核心的一个数;④ 在残差谱上做糖尿病/高血压逐阶比较,10 阶 Bonferroni 阈值 0.005。

    判读口径事先定死:ΔR²≥15% 且 VIF<3 → 硬化解释力真实;5-15% → 部分真实,须按年龄分层;<5% → 读的是年龄不是硬化,”抗硬化”话术须撤回。疾病信号方面:校正后仍有≥2 阶、方向一致、且在同病不同并发症间呈剂量-反应,方可称”疾病指纹”,否则只能叫”未解释残差”。

    这套设计不打算证明脉诊有用。它的价值在于把”脉象反映血管老化”这句含糊话拆成两个能分别回答的问题:老化占多少,疾病占多少。 若答案是”老化占九成”,脉诊的临床定位就该是”血管年龄计”,而非”疾病探测器”。

    证据与局限

    已验证: 年龄是健康人群动脉硬化的最主要相关因素,模型 R²=0.512(PMID 15123572);糖尿病改写血管老化速率,较早达功能平台,校正血压后 PWV 与年龄不再相关(PMID 12832325);年龄、性别、平均动脉压、糖尿病对 PWV 各自独立相关(PMID 16855622);PWV 的年龄-性别-地区参照值差异显著,绝对值不可跨人群套用(PMID 37229905);年龄改写主动脉波形的主通道是储器压力(9.9 mmHg/十年,r=0.69)而非后向反射波(PMID 20008272);2 型糖尿病人群中波形形状的独立决定因素是 MAP、心率、身高,HbA1c 非独立(PMID 18497469);通用传递函数的校准误差有 76% 传入中心动脉压估计(PMID 16229910)。

    部分验证: “控制年龄后 PWV/AIx 对谐波谱解释力保留多少”这一 ΔR² 数值尚无直接研究;现有证据只证明年龄与硬化各自独立、糖尿病改写年龄斜率,但两者方差在谐波谱上的分配比例是空白。糖尿病是否携带独立于硬化的谐波信号只有间接线索(AIx 在糖尿病组未达显著、在非糖尿病心血管病组显著,PMID 19888415),尚无控制硬化后的逐阶比较。

    待验证: ① 控制年龄与 PWV/AIx 后,逐阶谐波能量比在糖尿病/高血压组是否仍有稳定偏移;② 该偏移是否在同病不同并发症亚组间呈剂量-反应(特异性的必要条件);③ 若 ΔR²<5%,”谐波谱=血管年龄计”这一降级表述能否获独立样本验证。

    相关阅读

    证据等级: B
    参考文献: PMID 12832325, PMID 15123572, PMID 16855622, PMID 37229905, PMID 20008272, PMID 18497469, PMID 19888415, PMID 16229910


    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。

    Old Age and Illness Are Two Separate Accounts: Stripping Age from Disease Leaves the Harmonic Spectrum with Reduced Explanatory Power

    Evidence level: B

    A healthy 65-year-old and a 45-year-old with 10 years of diabetes may present with remarkably similar pulse waveforms. However, “aging” and “disease” are two completely distinct accounts—separating them is crucial; otherwise, all prior conclusions regarding “hardening → harmonic spectrum” rest on a flawed foundation.

    Main Text

    The previous article brought “hardening/afterload” into the open: a substantial portion of the variance in pulse waveform harmonics can be explained by PWV (pulse wave velocity) and AIx (augmentation index). However, this conclusion relies on a critical default assumption: Are changes in PWV and AIx driven by aging, or by disease?

    This is not academic nitpicking. The Framingham Offspring Cohort (188 men, 333 women; excluding clinical cardiovascular disease, hypertension, diabetes, smoking, and obesity) demonstrated that “age increase” is the primary correlate of elevated carotifemoral PWV, with a model R²=0.512 (PMID 15123572). Even in a purified sample free of hypertension and diabetes, hardening still rises with age. Therefore, comparing young people as a healthy group against an elderly diabetic group means that a significant portion of the observed “diabetes-induced hardening” is merely “the diabetic group being older on average.”

    More problematic is that this confounding is not an additive offset but a rewritten slope. Cameron et al. measured PWV in three arteries across 169 participants (57 with diabetes, 112 without); the diabetic group showed faster PWV and lower central arterial compliance (0.662 vs 0.850, P<0.05), but averaged 10 years younger than the control group; furthermore, after adjusting for blood pressure, PWV showed no correlation with age in the diabetic group (PMID 12832325). The authors termed this “accelerated aging at an earlier age, followed by reaching a functional plateau”—in the diabetic population, the regression line linking “age → hardening” is flat. Calculating a “vascular age” for their PWV values using healthy reference tables has no physiological meaning.

    So, what proportion is attributable to each? Alecu et al. performed multivariate regression on 221 apparent healthy individuals over 60, finding that PWV correlated independently with age, sex, mean arterial pressure, and diabetes (PMID 16855622). “Independently” indicates statistical separability, but it is far from a clean separation—both share the same downstream pathway (elastin fragmentation, collagen cross-linking, accumulation of advanced glycation end-products).

    A 2023 systematic review including 167 studies, 509,743 participants across 34 countries, provided age-sex reference values for PWV (PMID 37229905): the same PWV value carries vastly different meanings across different ages, sexes, and regions (Asian baPWV is 1.83 m/s higher than European values, P=0.0014). Any absolute threshold without stratification is incorrect.

    The question thus refines to: After controlling for age, how much explanatory power remains for PWV/AIx in the harmonic spectrum? Do diabetes and hypertension carry harmonic signals independent of hardening? All three outcomes are informative:

    Outcome 1 (Erosion): If ΔR² drops below 5%—the primary association previously attributed to age is actually aging masquerading as hardening—the reading of “harmonic spectrum reflects vascular hardness” must be downgraded to “reads vascular age.”
    Outcome 2 (Retention): If ΔR² remains above 15% with VIF<3, then “hardening independent of age rewrites the harmonic spectrum” holds water, allowing further decomposition of which harmonics are managed by AIx and which by PWV.
    Outcome 3 (Disease Carries Independent Signal): If, after controlling for age and PWV, the diabetic or hypertensive groups still show shifts at specific harmonic orders—this is the most valuable cell, implying a disease fingerprint unexplainable by hardening within the harmonic spectrum, potentially arising from microcirculatory resistance, autonomic tone, or blood viscosity. The physical grounding for “same disease, different pulse; different disease, same pulse” lies here.

    Pitfall 1: Using “Vascular Age” as a Covariate. Since vascular age is calculated from PWV and age, using it as a covariate mixes back the very elements meant to be separated. One must use raw PWV + age as two independent covariates.

    Pitfall 2: Sample Collapse After Age Stratification. The prevalence of diabetes is naturally higher in the elderly group; strict stratification may leave some cells with very few subjects. An alternative is propensity score 1:1 matching based on age, sex, BMI, blood pressure, and heart rate, or treating age as a restricted cubic spline (the relationship between age and hardening is explicitly nonlinear; Uejima et al. found age explained 53% of the variance in carotid elastic modulus in 1,847 healthy individuals, with a significantly steeper slope in the elderly segment).

    Pitfall 3: Reading “Unexplained Residuals” as “Disease Fingerprints”. Differences remaining after controlling for age and PWV only indicate “not explained by these two,” not “explained by diabetes itself”—it could be medication, activity levels, or renal function. To claim specificity, one must perform dose-response testing within subgroups of same disease but different complications. A debunked optimistic idea worth noting: since both pathways involve hardening, shouldn’t we just measure PWV directly? Data does not support this. Tomita et al. performed stepwise regression in 194 type 2 diabetics; the independent determinants of radial AI were mean blood pressure (β=0.260), heart rate (β=−0.550), and height (β=−0.217), with HbA1c becoming non-significant after multivariate adjustment (PMID 18497469). Wilhelm et al. found AIx higher in 33 diabetics than controls (21.12% vs 19.84%) but not significant, whereas it was significantly elevated in 100 non-diabetic cardiovascular patients (26.00%, P<0.02) (PMID 19888415). The label “diabetes” itself has surprisingly weak explanatory power for waveform shape.

    The stubbornness of age has physiological grounding. Davies et al. decomposed intraluminal augmentation pressure into forward wave, backward reflection wave, and reservoir pressure components; finding that when reservoir pressure is included, the contribution of the backward reflection wave is reduced by 64%, with reservoir pressure becoming the largest component and increasing by 9.9 mmHg per decade (r=0.69, P<0.001), while the backward wave no longer correlates with age (PMID 20008272). The main channel through which age rewrites the waveform is aortic reservoir capacity, not earlier return of reflection waves—controlling only for PWV misses this pathway, which is why both AIx and PWV must be collected simultaneously.

    The final risk: Transfer function errors mistaken for physiological signals. Papaioannou et al. simulated calibration errors in 103 participants; a ±5 mmHg bias in isolated brachial systolic pressure transmitted 76% unchanged to the inferred central arterial pressure, with higher transmission ratios observed when mean pressure >117 mmHg or heart rate <74 bpm (PMID 16229910). Elderly diabetics fall precisely into the category of higher blood pressure and slower heart rate—the very groups with the highest error transmission rates being compared. Thus, this design mandates identical equipment and calibration, prioritizing measured radial artery waveforms for harmonic decomposition.

    Four-step design: ① Full-sample Type III variance decomposition, with harmonic energy ratios (C1-C10) as the dependent variable and age/sex/PWV/AIx/MAP/heart rate/diabetes/hypertension as independent variables, outputting squared sum proportions and VIF; ② Modeling in three strata (<40 / 40-59 / ≥60) to test if PWV coefficients attenuate between strata; ③ Incremental explanatory power testing—Model A (age + sex + MAP + heart rate) → Model B (+PWV+AIx), comparing ΔR² with AIC; this is the single most core metric for the entire article; ④ Stepwise comparison of diabetes/hypertension in the residual spectrum, with a Bonferroni threshold of 0.005 for 10 orders.

    Judgment criteria are fixed in advance: ΔR²≥15% and VIF<3 → genuine hardening explanatory power; 5-15% → partially true, requires age stratification; <5% → reading age not hardening, “anti-hardening” rhetoric must be withdrawn. Regarding disease signals: only if ≥2 harmonics remain significant post-adjustment, with consistent direction and a dose-response relationship across same-disease different-complication subgroups can it be termed a “disease fingerprint”; otherwise, it is merely “unexplained residual.”

    This design does not aim to prove pulse diagnosis is useful. Its value lies in deconstructing the vague statement “pulse reflects vascular aging” into two answerable questions: how much is aging, and how much is disease. If the answer is “aging accounts for 90%,” the clinical positioning of pulse diagnosis should be “vascular age meter,” not “disease detector.”

    Evidence and Limitations

    Verified: Age is the primary correlate of arterial hardening in healthy populations, model R²=0.512 (PMID 15123572); diabetes rewrites vascular aging rate, reaching functional plateau earlier, and PWV correlates with age no longer after blood pressure adjustment (PMID 12832325); age, sex, mean arterial pressure, and diabetes correlate independently with PWV (PMID 16855622); age-sex-region reference values for PWV show significant differences, making absolute values unsuitable across populations (PMID 37229905); age rewrites the main channel of aortic waveform via reservoir pressure (9.9 mmHg/decade, r=0.69) rather than backward reflection waves (PMID 20008272); in type 2 diabetes, independent determinants of waveform shape are MAP, heart rate, and height, with HbA1c non-independent (PMID 18497469); general transfer function calibration errors transmit 76% into central arterial pressure estimates (PMID 16229910).

    Partially Verified: The ΔR² value regarding “how much explanatory power PWV/AIx retains in the harmonic spectrum after controlling for age” lacks direct research; existing evidence only proves independent correlation of age and hardening, and that diabetes rewrites the age slope, but the allocation proportion of their variance in the harmonic spectrum is blank. Whether diabetes carries harmonic signals independent of hardening offers only indirect clues (AIx not significant in diabetic group but significant in non-diabetic cardiovascular patients, PMID 19888415), lacking stepwise comparison after controlling for hardening.

    To Be Verified: ① After controlling for age and PWV/AIx, whether harmonic energy ratios still show stable shifts in diabetic/hypertensive groups at each order; ② Whether such shifts exhibit a dose-response relationship across subgroups of same disease but different complications (a necessary condition for specificity); ③ Whether the downgraded statement “harmonic spectrum = vascular age meter” can be validated in an independent sample if ΔR²<5%.

    Related Reading

    Evidence Level: B
    References: PMID 12832325, PMID 15123572, PMID 16855622, PMID 37229905, PMID 20008272, PMID 18497469, PMID 19888415, PMID 16229910


    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.

  • 把降压药当成一次实验:谐波谱走的是后负荷的路,还是共振的路

    Evidence level: B

    观察性研究永远分不清”谁影响谁”。降压药是唯一能在人身上做因果的地方——如果两类药降同样多的血压,却让谐波谱朝相反方向变,那就说明谐波谱里确实有一条不属于后负荷的通路。

    正文

    前面几篇一直在算账:谐波谱的方差有多少归硬化、多少归年龄、多少归疾病。但这些都是观察——血压高的人谐波谱长这样,硬化的人谐波谱长那样。观察到的相关,可能是血压改谐波,也可能是谐波(或者说它背后的血管状态)改血压。

    药物能破这个局。给同一批人两种降压力度相同、机制不同的药,看谐波谱是否同步、同向、同幅地变。

    第一件事:理想的”纯反射探针”不存在

    最常被当成”只改波反射、不动血管”的工具是舌下硝酸甘油。Tahvanainen 等的双盲随机研究(35 名健康人,0.25 mg NTG vs 400 μg 沙丁胺醇吸入)确实显示 NTG 让增强指数下降 7.4%(卧位)到 18.7%(立位),同时脉搏波传导速度也降了(PMID 22179079)。降反射的同时降硬化,两者缠在一起。

    更关键的是 Fok 等在 30 例心导管患者里的波强分析(PMID 24516104):NTG 让增强压降了 12.8 mmHg,拆开看——前向波成分降 7.0 mmHg,后向波成分降 5.8 mmHg;且 NTG 完全没有改变后向/前向波的振幅比。也就是说,它降”增强压”的效应里有一大半来自心室射血本身变柔和了,不是外周反射变少了。拿 NTG 当”纯反射开关”是错的——它同时动了三个旋钮:心室射血、波反射、外周阻力。

    第二件事:药物类别真的会在谐波谱上分道

    真正直接的证据来自动物。Wang 等在正常血压 WKY 大鼠上做自身对照,注射三种降压药后比较血压波形的谐波比例(PMID 22007621):阿替洛尔(β 阻滞剂)显著降低所有阶次的谐波比例;卡托普利(ACEI)和氯沙坦(ARB)显著升高第一和第四谐波比例。

    同一个血压下降幅度,方向相反。这是”谐波谱里有一条不属于后负荷的通路”最干净的一条线索——如果谐波谱只是后负荷的换算表,血压降了谱就该朝一个方向走,不该分岔。

    第三件事:人体层面,PWV 和 AIx 对药物的响应就是不同步的

    三项 meta 分析把这个不同步量化了。Chen 等纳入 10 项 RCT、938 人:ARB 与其他降压药相比,PWV 改善无差异(SMD 0.04,95%CI −0.16~0.24,P=0.71),但增强指数的改善更优(MD 8.94,P=0.01)(PMID 26793356)。Ye 等的 28 篇网络 meta 结论更干脆:八类降压药在 PWV 上没有显著差异,差异只在收缩压(PMID 29960763)。2023 年 RIGIPREV 网络 meta(76 项研究)细分了:用药超过 6 个月时,噻嗪类、ACEI、ARB 及若干联合方案降 PWV 更有优势,而降 AIx 的优势集中在 ACEI 和 ARB(PMID 37724181)。

    两个指标分家,说明”硬化”和”反射”在药物作用下不是同一件事。CAFE 研究给了临床侧的印证:2199 名高血压患者,氨氯地平方案与阿替洛尔方案肱动脉收缩压只差 0.7 mmHg(无统计学差异),但中心主动脉收缩压差 4.3 mmHg、中心脉压差 3.0 mmHg,全都显著(PMID 16476843)。手臂上量不出差别,波形上量得出。

    第四个坑:β 阻滞剂改的可能只是心率

    Casey 等在年轻男女中静脉给普萘洛尔,增强指数升高;但按心率 75 校正后,只有女性仍显著升高,男性的 ΔAIx 与 ΔHR 呈负相关(PMID 22106401)。后来同一组用伊伐布雷定(只降心率、不碰血管)替换 β 阻滞剂做对照:心率、平均动脉压都匹配的情况下,桡动脉增强指数没有变化(PMID 24535377)。也就是说,男性那部分”β 阻滞剂升 AIx”纯粹是心率慢下来的副产品。

    而心率慢会让每个心动周期的舒张期变长、反射波落点改变——这恰好会打乱按心率归一化的谐波阶次。任何药物干预的谐波研究,不锁心率就是在测心率。

    那怎么做这个分离试验

    第一步,四臂平行 + 交叉。 至少四类机制不同的药(ACEI/ARB、CCB、噻嗪、β 阻滞剂)加安慰剂,每臂用药 4 周、洗脱 4 周。要求各臂降压幅度匹配(ΔMAP 组间差 <5 mmHg),否则测到的还是剂量。

    第二步,三量同步采集。 每次访视同时测桡动脉压力波形(连续 30 个心动周期取平均)、颈股 PWV、中心动脉压、心率。三量都进模型当协变量,看的是残差谐波谱。

    第三步,逐阶方向一致性检验。 对 C1–C10 每一阶,检验四个药物臂的 ΔC 符号是否一致(Kendall W 协调系数)。这是整个设计的分水岭:

    • 若四个臂全部同向,且 ΔC 幅度与 ΔMAP 呈线性(R²>0.6)→ 谐波谱就是后负荷的换算表,此前所有”器官共振”解释都要降级;
    • 若只在某几阶(预期是 C4)出现方向分岔,且分岔恰好落在”血管扩张类升、β 阻滞剂降”这条线上 → 该阶次携带了独立于后负荷的外周张力/反射信息,共振路径初步成立;
    • 若低阶同向、高阶分岔 → 按阶次分层:低阶读后负荷,高阶读共振。

    第四步,把心率做成对照臂。 单独加一组伊伐布雷定(只降心率),用它来标定”纯粹由心率变化引起的谐波漂移”是多少。任何药物臂的 ΔC 必须超过这个基线漂移才算数。

    三个必须防的坑

    别把 NTG 当纯探针。 前向波占它降压效应的一半以上(PMID 24516104),用它推”反射通路”会把心室射血的贡献算进去。

    别把降压后的 PWV 下降当”血管软化”。 Stewart 等给高血压患者静脉 NTG,平均动脉压降了 22 mmHg 降到与正常人同一水平,PWV 仍高 24%、颈动脉扩张性仍低 48%(PMID 16908758)。降压撤掉的是”撑”,撤不掉的是”结构”。

    别用同类药做跨类别推论。 ARB 在 PWV 上不优于其他药、在 AIx 上更优(PMID 26793356),这两个结论必须分开引用,合起来说”ARB 更好”是超范围。

    证据与局限

    已验证: 正常血压大鼠中,阿替洛尔降低全部谐波比例、卡托普利与氯沙坦升高 C1 与 C4,同幅度降压下方向相反(PMID 22007621);硝酸甘油降增强压的效应可拆为前向波 7.0 mmHg 与后向波 5.8 mmHg,且不改变后向/前向振幅比(PMID 24516104);NTG 同时降低增强指数与 PWV(PMID 22179079);高血压患者 NTG 急性降压 22 mmHg 后大动脉硬化未恢复(PMID 16908758);三项 meta 一致显示各类降压药在 PWV 上无显著差异、在 AIx 上有类别差异(PMID 26793356、29960763、37724181);CAFE 中肱动脉压相同的两方案中心脉压差 3.0 mmHg(PMID 16476843);β 阻滞剂升高 AIx 的效应在按心率校正后大部分消失,伊伐布雷定替换后 AIx 不变(PMID 22106401、24535377)。

    部分验证: “同幅度降压下逐阶谐波谱出现方向分岔”这一命题,目前只有大鼠单点证据(PMID 22007621),人体层面尚无逐阶 C1–C10 的药物对照数据。PWV 与 AIx 分家只能作为间接支持,二者都不是谐波谱本身。

    待验证: ① 人体四类降压药在 ΔMAP 匹配条件下,C1–C10 各阶的 ΔC 方向是否分岔;② 分岔是否稳定落在 C4 这一阶(大鼠证据的预测);③ 若所有阶次同向且与 ΔMAP 线性,谐波谱的”共振”解释应如何降级表述。

    相关阅读

    证据等级: B
    参考文献: PMID 22007621, PMID 24516104, PMID 22179079, PMID 16908758, PMID 26793356, PMID 29960763, PMID 37724181, PMID 16476843, PMID 22106401, PMID 24535377


    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。

    Treating Antihypertensives as a Single Experiment: Does the Harmonic Spectrum Follow Afterload or Resonance?

    Evidence level: B

    Observational studies can never distinguish “who affects whom.” Antihypertensive drugs are the only intervention that allows causality in humans—if two drug classes lower blood pressure by the same amount but drive the harmonic spectra in opposite directions, it indicates a pathway within the harmonic spectrum that is not merely a function of afterload.

    Main Text

    Previous articles have been accounting for the variance: how much belongs to stiffening, how much to age, and how much to disease. But these are all observations—hypertensive patients have such harmonic spectra, stiffening patients have such spectra. Observed correlations could mean blood pressure alters the harmonics, or the harmonics (or the underlying vascular state behind them) alter blood pressure.

    Drugs can break this deadlock. Give the same cohort two antihypertensives with identical pressure-lowering effects but different mechanisms, and observe whether the harmonic spectra shift synchronously, in the same direction, and with the same magnitude.

    First Point: An Ideal “Pure Reflection Probe” Does Not Exist

    The tool most often treated as one that “only alters wave reflection without moving the vessel” is sublingual nitroglycerin. Tahvanainen et al.’s double-blind randomized study (35 healthy individuals, 0.25 mg NTG vs. 400 μg salbutamol inhalation) indeed showed that NTG reduced the augmentation index by 7.4% (supine) to 18.7% (standing), while pulse wave velocity also decreased (PMID 22179079). Reducing reflection while reducing stiffening keeps the two intertwined.

    More critically, Fok et al.’s wave intensity analysis in 30 cardiac catheterization patients (PMID 24516104) showed that NTG reduced augmentation pressure by 12.8 mmHg; when decomposed, the forward wave component dropped by 7.0 mmHg and the backward wave component dropped by 5.8 mmHg; furthermore, NTG did not alter the amplitude ratio of backward to forward waves. In other words, a large portion of NTG’s effect on “augmentation pressure” comes from the ventricular ejection itself becoming softer, not from reduced peripheral reflection. Treating NTG as a “pure reflection switch” is incorrect—it simultaneously turns three knobs: ventricular ejection, wave reflection, and peripheral resistance.

    Second Point: Drug Classes Truly Diverge on the Harmonic Spectrum

    Direct evidence comes from animals. Wang et al., in normotensive WKY rats using a within-subject design, compared harmonic ratios of blood pressure waveforms after injecting three antihypertensives (PMID 22007621): atenolol (β-blocker) significantly reduced harmonic ratios across all orders; captopril (ACEI) and losartan (ARB) significantly increased the first and fourth harmonic ratios.

    The same magnitude of blood pressure drop yields opposite directions. This is the cleanest clue yet that “there is a pathway within the harmonic spectrum not belonging to afterload”—if the harmonic spectrum were merely a conversion table for afterload, lowering blood pressure should drive the spectrum in one direction, not diverge.

    Third Point: At the Human Level, PWV and AIx Respond Differently to Drugs

    Three meta-analyses quantified this desynchronization. Chen et al., including 10 RCTs and 938 individuals, found that compared to other antihypertensives, ARBs showed no difference in PWV improvement (SMD 0.04, 95%CI −0.16~0.24, P=0.71) but superior improvement in augmentation index (MD 8.94, P=0.01) (PMID 26793356). Ye et al.’s network meta-analysis of 28 studies concluded even more decisively: eight classes of antihypertensives showed no significant difference in PWV, with differences only in systolic blood pressure (PMID 29960763). The 2023 RIGIPREV network meta-analysis (76 studies) further subdivided: when medication exceeded 6 months, thiazides, ACEIs, ARBs, and several combination regimens showed advantages in lowering PWV, while advantages in lowering AIx concentrated in ACEIs and ARBs (PMID 37724181).

    The two metrics have separated, indicating that “stiffening” and “reflection” are not the same event under drug action. The CAFE study provided clinical corroboration: in 2199 hypertensive patients, the amlodipine regimen and atenolol regimen differed by only 0.7 mmHg in brachial systolic blood pressure (no statistical difference), but differed by 4.3 mmHg in central systolic blood pressure and 3.0 mmHg in central pulse pressure, all significant (PMID 16476843). No difference is measurable on the arm, but it is measurable on the waveform.

    Fourth Pitfall: β-Blockers May Only Alter Heart Rate

    Casey et al., in young men and women, administered intravenous propranolol; the augmentation index increased; but after correcting for heart rate of 75, only women still showed a significant increase, with men’s ΔAIx negatively correlated with ΔHR (PMID 22106401). Later, the same group used ivabradine (lowering heart rate only, not touching vessels) as a control replacing β-blockers: with matched heart rate and mean arterial pressure, the radial artery augmentation index showed no change (PMID 24535377). In other words, the portion of “β-blockers raising AIx” in men was purely a byproduct of slowing heart rate.

    And slowing heart rate lengthens diastole in each cardiac cycle and shifts the landing point of reflection waves—this precisely disrupts harmonic orders normalized by heart rate. Any harmonic study involving drug intervention without locking heart rate is essentially measuring heart rate effects.

    So How to Conduct This Separation Experiment

    Step 1: Four-Arm Parallel + Crossover. At least four drug classes with different mechanisms (ACEI/ARB, CCB, thiazide, β-blocker) plus placebo, each arm for 4 weeks with a 4-week washout. Require matched blood pressure lowering across arms (ΔMAP group difference <5 mmHg); otherwise, what is measured is simply dose.

    Step 2: Synchronous Three-Measurement Collection. At each visit, simultaneously measure radial artery pressure waveform (average of 30 consecutive cardiac cycles), carotid-femoral PWV, central arterial pressure, and heart rate. Enter all three into the model as covariates to examine the residual harmonic spectrum.

    Step 3: Order-by-Order Direction Consistency Test. For each order C1–C10, test whether the signs of ΔC are consistent across the four drug arms (Kendall W coefficient). This is the watershed of the entire design:

    • If all four arms are uniformly directional, and ΔC magnitude correlates linearly with ΔMAP (R²>0.6) → the harmonic spectrum is merely a conversion table for afterload; all prior “organ resonance” explanations must be downgraded;
    • If divergence appears only in specific orders (expected at C4), and the divergence aligns exactly with “vasodilators increase, β-blockers decrease” → that order carries information independent of afterload regarding peripheral tension/reflection; the resonance path is preliminarily established;
    • If lower orders are uniform while higher orders diverge → stratify by order: read afterload from lower orders, resonance from higher orders.

    Step 4: Add a Heart Rate Control Arm. Specifically add an ivabradine group (lowering heart rate only) to calibrate “pure harmonic drift caused by heart rate changes.” Any ΔC in a drug arm must exceed this baseline drift to be valid.

    Three Must-Avoid Pitfalls

    Do not treat NTG as a pure probe. The forward wave accounts for more than half of its blood pressure lowering effect (PMID 24516104); using it to infer “reflection pathways” will include ventricular ejection contributions.

    Do not treat post-hypertensive PWV reduction as “vascular softening.” Stewart et al., administering IV NTG to hypertensive patients, lowered mean arterial pressure by 22 mmHg to normal levels, yet PWV remained 24% higher and carotid distensibility 48% lower (PMID 16908758). What is withdrawn with blood pressure lowering is the “support”; what cannot be withdrawn is the “structure.”

    Do not use cross-class inferences from similar drugs. ARBs are not superior to other drugs in PWV but are superior in AIx (PMID 26793356); these two conclusions must be cited separately; stating “ARB is better” together is out of scope.

    Evidence and Limitations

    Verified: In normotensive rats, atenolol reduced all harmonic ratios while captopril and losartan increased C1 and C4, showing opposite directions under the same magnitude of blood pressure drop (PMID 22007621); nitroglycerin’s reduction in augmentation pressure can be decomposed into 7.0 mmHg forward wave and 5.8 mmHg backward wave without altering the backward/forward amplitude ratio (PMID 24516104); NTG simultaneously reduced augmentation index and PWV (PMID 22179079); in hypertensive patients, acute NTG-induced blood pressure drop of 22 mmHg did not restore large artery stiffening (PMID 16908758); three meta-analyses consistently showed no significant difference among antihypertensive classes in PWV but category differences in AIx (PMID 26793356, 29960763, 37724181); in CAFE, two regimens with identical brachial pressure differed by 3.0 mmHg in central pulse pressure (PMID 16476843); the effect of β-blockers raising AIx largely disappeared after heart rate correction, and AIx remained unchanged when replaced by ivabradine (PMID 22106401, 24535377).

    Partially Verified: The proposition that “harmonic spectra show directional divergence across orders under same-magnitude blood pressure lowering” currently has only single-point evidence from rats (PMID 22007621); no human data exists for drug comparisons across orders C1–C10. The separation of PWV and AIx can only serve as indirect support; neither is the harmonic spectrum itself.

    To Be Verified: ① Whether C1–C10 orders show directional divergence under matched ΔMAP conditions across four antihypertensive classes in humans; ② Whether the divergence stabilizes at order C4 (predicted by rat evidence); ③ How to downgrade the “resonance” explanation for the harmonic spectrum if all orders are uniform and linearly correlated with ΔMAP.

    Related Reading

    Evidence Level: B
    References: PMID 22007621, PMID 24516104, PMID 22179079, PMID 16908758, PMID 26793356, PMID 29960763, PMID 37724181, PMID 16476843, PMID 22106401, PMID 24535377


    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.

  • 把心率剥掉,浮沉弦滑还剩几分

    Evidence level: B

    数脉和迟脉本来就是心率本身,它们分得开是废话;真正要问的是——浮沉弦滑这些”形状”里,有多少也只是心跳快慢投下来的影子。

    正文

    上一篇把六类脉象拆成五个要素、各自去找频谱上的轴,顺手留下一根刺:45 例高血压患者的自动评级里,浮脉的判定阈值落在了心率 ≥ 58.25 次/分(PMID 26988431)。一个讲”按得深不深”的脉象,判定条件竟然是心跳快慢。这不是中医的问题,是测量协议的问题,但它足以让整条分类管线的可信度打折。

    这一篇专门处理这根刺。

    先看一个不太舒服的事实

    增强指数(AIx)是脉波分析里最主流的输出之一,长期被当作动脉僵硬度的指标。2000 年一项 22 名起搏器植入者的研究给出了它的真面目:起搏频率从 60 递増到 110 次/分,AIx 与心率呈显著负线性相关(r = −0.76,P < 0.001),心率每快 10 次/分,AIx 掉约 4 个百分点;射血时长与心率同样负相关(r = −0.51)(PMID 10811742)。机制说得很清楚:心率快了,收缩期绝对时长变短,反射波返回时落在收缩期的位置就变了,波峰形状跟着变。不是血管变软了,是节拍变了。

    2009 年那篇更狠。50 例患者做多巴酚丁胺负荷,同时测桡动脉波形与组织多普勒:AIx 与射血时长 r = 0.88、与心率 r = −0.81、与心肌收缩应变率 r = 0.72(全 P < 0.001)。但把收缩应变率做心率校正后,它与 AIx 的关系消失了(r = 0.18,P = 0.11)。射血时长单独解释了 AIx 78% 的方差(β = 0.88,模型 R² = 0.77)。作者的结论一句话:AIx 由变时性(chronotropic)而非变力性(inotropic)决定(PMID 19720955)。

    一个被写进无数论文当”血管结构指标”的量,近八成变化来自射血时长——而射血时长基本就是心率的函数。这不是孤例:40 名健康青年的压力脉波研究显示,时间类参数的主控因素是心率,振幅类的主控因素是收缩压(PMID 26406041)。二阶导数那套指标(b/a、c/a、d/a、老化指数 AGI,正是”弦””滑”最常借用的形状量)也一样:848 人样本里,D:A 比与 AGI 独立地负相关于心率,B:A 比正相关,校正年龄与血压后仍成立(PMID 15831356);294 例高血压患者重复了同一方向(PMID 12473866)。

    所以”控制心率”不是一句客套话。形状类脉象的判别力里,很可能有一块是纯心率租来的。

    三步分离,加一个必须崩的对照

    第一步:心率匹配配对。 按 1:1 最近邻匹配,卡钳 ±3 次/分,配对年龄、性别、平均压。匹配后仍保留的判别力才可能是形状信号。

    第二步:协变量调整。 心率作为连续协变量进模型,同时放年龄、平均压、接触压力档位。这里有个关键的方法学分歧:不要用”校正到 75 次/分”这种一步到位的做法。 2014 年一篇专文指出,把 AIx 统计校正到 75 bpm 在生理和统计上都站不住:一是心率与动脉僵硬度存在慢性交互,二是这种校正隐含假设”心率-AIx 关系在全人群一致”,而这个斜率本身随年龄、族群、病理状态变化。更稳的做法是把心率作为独立预测因子或协变量放进模型,而不是先折算再忽略(PMID 24257465)。

    第三步:窄心率带内判别。 在 60-70、70-80、80-90 三个窄带内分别做判别,看 AUC 是否还在。这一层最能说明问题——如果判别力只在全样本里存在、一进窄带就蒸发,那它一直是靠心率撑着的。

    第四步:阳性对照。 这是本设计最要紧的一步:数脉/迟脉的判别 AUC,在控制心率之后必须崩到 0.5。 因为它们本来就是按心率定义的,控制心率等于把定义变量拿走。如果它们没崩,就说明管线存在信息泄漏——那么浮沉弦滑的所有结果一律作废。 一个必然归零的对照,用途是给整条管线做体检。这比任何 AUC 都值钱。

    两个容易翻车的地方

    一是碰撞偏倚。 如果某种病理状态同时升高心率和改变波形,那么”匹配心率”等于在这条通路上做了条件化,反而会造出假关联。补救办法是报告匹配前后的疾病构成,并在有疾病标签的子集里做敏感性分析。

    二是匹配了数字,没匹配生理。 48 例起搏器植入者研究显示,心房、房室、心室三种起搏方式对”心率-射血时长”关系有显著影响(P = 0.01);即便固定心率,不同起搏方式的 AIx 与脉压放大率仍有显著差异(P < 0.0001)(PMID 27471239)。所以心率匹配必须同时锁定搏动来源。

    还有一个反向陷阱

    归一化能剥掉心率,但会连真信息一起扔。2023 年一项方法学研究比对波形时发现:不归一化的情况下,两个波形尾端实际形态差异很小,但因心率不同,尾部反而成了累积距离的主要来源,形态差异被心率彻底淹没;双重归一化后才真正聚焦形态——代价是心率与脉压被移出主特征,只能降级另算(PMID 36767962)。“控制心率”和”丢掉心率信息”是两件事,后者会顺手毁掉数脉迟脉的判别基础。

    心率本身也不干净:52 例起搏器研究剥离血压效应后,颈股脉波速度每 10 次/分仍升 0.16-0.20 m/s(PMID 27245180)。

    怎么读结果

    判据提前写死,避免事后解释:

    • 匹配+调整后保留 ≥ 80% 判别力 → 该类主要是真形状信号,心率不是主因;
    • 保留 50-80% → 混合型,必须在报告里标明”含心率成分”,并给出心率贡献的比例;
    • 保留 < 50% → 主要是心率伪装,该类的现有文献判别力需要整体下调;
    • 阳性对照(数/迟)未崩 → 管线有泄漏,全部结论作废重跑。

    这不是又一篇”证明脉诊有用”的文章。它做的是审计:在把脉象交给任何分类器之前,先确认分类器学的是脉,不是心跳的快慢。

    证据与局限

    已验证: 起搏器递增起搏下 AIx 与心率显著负线性相关、心率每 10 次/分 AIx 降约 4%、射血时长与心率 r = −0.51(PMID 10811742);AIx 与射血时长 r = 0.88、射血时长解释 78% 方差、收缩应变率经心率校正后与 AIx 无关(PMID 19720955);AIx 校正到 75 bpm 在生理与统计上均不合适(PMID 24257465);压力脉波时间类参数主因为心率、振幅类主因为收缩压(PMID 26406041);SDPTG 的 D:A、AGI 负相关于心率而 B:A 正相关,校年龄血压后仍成立(PMID 15831356、PMID 12473866);DTW 不归一化时心率主导波形距离、归一化会移除心率与脉压信息(PMID 36767962);不同起搏方式在固定心率下 AIx 仍有显著差异(PMID 27471239);剥离血压效应后脉波速度仍随心率上升 0.16-0.20 m/s/10bpm(PMID 27245180);45 例高血压中浮脉自动阈值落在心率 ≥ 58.25(PMID 26988431)。

    部分验证: “形状类脉象判别力中有相当比例来自心率”目前是从上述推广到脉象分类的合理推断,尚无研究在同一批受试者身上直接做过心率匹配的脉象判别对照;”保留 ≥ 80% 即真形状信号”这一阈值是本文设定的判读口径,非既有共识。

    待验证: ①心率匹配前后浮沉弦滑四类判别 AUC 的实际保留比例;②数/迟作为阳性对照是否真能崩到 0.5(即管线是否存在泄漏);③心率匹配引入的碰撞偏倚在本议题下是否会制造假关联。

    相关阅读

    证据等级: B(心率对波形时相与形状的主导作用、AIx 的变时性依赖、SDPTG 的心率依赖、归一化的双刃性均有直接文献;”心率匹配分离检验”用于脉象分类本身无先例,属研究提案)
    参考文献: PMID 10811742, PMID 19720955, PMID 24257465, PMID 26406041, PMID 15831356, PMID 12473866, PMID 36767962, PMID 27471239, PMID 27245180, PMID 26988431


    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。

    Stripping away the heart rate, how much remains of the floating and sliding strings

    Evidence level: B

    Pulse counting and late pulse are inherently heart rate itself; separating them is redundant. The real question is—within these “shapes,” how much is merely a shadow cast by heart rate variability.

    Main Text

    The previous article decomposed six pulse categories into five elements and identified their spectral axes, inadvertently leaving a thorn: in an automated rating of 45 hypertensive patients, the threshold for identifying the Floating pulse fell at a heart rate ≥ 58.25 bpm (PMID 26988431). A pulse quality defined by “how deep it feels” relies on a condition determined by heart rate. This is not a flaw in Traditional Chinese Medicine (TCM); it is a flaw in the measurement protocol, yet it is sufficient to undermine the credibility of the entire classification pipeline.

    This article addresses that thorn directly.

    Start with an Uncomfortable Fact

    The Augmentation Index (AIx) is one of the most mainstream outputs in pulse wave analysis, long regarded as an indicator of arterial stiffness. A 2000 study of 22 pacemaker recipients revealed its true nature: as pacing frequency increased from 60 to 110 bpm, AIx showed a significant negative linear correlation with heart rate ($r = -0.76$, $P < 0.001$); AIx dropped by approximately 4 percentage points for every 10 bpm increase in heart rate; ejection duration also showed a negative correlation with heart rate ($r = -0.51$) (PMID 10811742). The mechanism is clear: faster heart rates shorten absolute systolic duration, shifting the position where the reflected wave returns during systole, thereby altering the waveform peak shape. It is not that the vessels have softened; it is that the rhythm has changed.

    The 2009 study was even more definitive. In 50 patients undergoing dobutamine stress testing, simultaneous measurements of radial artery waveforms and tissue Doppler showed AIx correlated strongly with ejection duration ($r = 0.88$), heart rate ($r = -0.81$), and myocardial contraction strain rate ($r = 0.72$; all $P < 0.001$). However, after adjusting for heart rate, the relationship between strain rate and AIx disappeared ($r = 0.18$, $P = 0.11$). Ejection duration alone explained 78% of the variance in AIx ($\beta = 0.88$, model $R^2 = 0.77$). The authors concluded in one sentence: AIx is determined by chronotropy (heart rate variability) rather than inotropy (contractility) (PMID 19720955).

    A metric written into countless papers as a “vascular structural indicator” changes nearly 80% due to ejection duration—which is essentially a function of heart rate. This is not an isolated case: a study of 40 healthy young adults under stress showed that time-based parameters were primarily driven by heart rate, while amplitude-based parameters were driven by systolic blood pressure (PMID 26406041). The same holds for second-derivative metrics (b/a, c/a, d/a, and the Aging Index AGI, which are commonly used to quantify “Wiry” and “Slippery” shapes): in a sample of 848 individuals, D:A ratio and AGI independently correlated negatively with heart rate, while B:A ratio correlated positively; these associations held even after adjusting for age and blood pressure (PMID 15831356); the same direction was replicated in 294 hypertensive patients (PMID 12473866).

    Therefore, “controlling heart rate” is not just polite rhetoric. A significant portion of the discriminative power in shape-based pulse qualities may be rented from heart rate.

    Three Steps to Separate, Plus One Critical Control

    Step 1: Heart Rate Matching. Match 1:1 using nearest neighbors within ±3 bpm, pairing for age, sex, and mean blood pressure. Only discriminative power remaining after matching can be considered a true shape signal.

    Step 2: Covariate Adjustment. Enter heart rate as a continuous covariate in the model, alongside age, mean blood pressure, and contact pressure levels. There is a critical methodological divergence here: do not use a “one-step correction to 75 bpm”. A 2014 article pointed out that statistically adjusting AIx to 75 bpm is physiologically and statistically unsound: first, heart rate interacts chronically with arterial stiffness; second, such correction implicitly assumes the “heart rate-AIx relationship is consistent across all populations,” a slope that itself varies by age, ethnicity, and pathological state. A more robust approach is to include heart rate as an independent predictor or covariate in the model rather than pre-calculating and ignoring it (PMID 24257465).

    Step 3: Discrimination within Narrow Heart Rate Bands. Perform discrimination separately within three narrow bands (60-70, 70-80, 80-90 bpm) to see if AUC remains. This layer reveals the truth—if discriminative power exists only in the full sample but evaporates within narrow bands, it has always been propped up by heart rate.

    Step 4: Positive Control. This is the most critical step in this design: the AUC for Pulse Counting/Late Pulse discrimination must collapse to 0.5 after controlling for heart rate. Since they are defined by heart rate, controlling for heart rate removes the defining variable. If they do not collapse to 0.5, it indicates information leakage in the pipeline—rendering all results for Floating/Sunken/Wiry/Slippery invalid. A control that must theoretically reach zero serves as a health check for the entire pipeline. It is more valuable than any AUC metric.

    Two Places Prone to Failure

    First: Collision Bias. If a pathological state simultaneously elevates heart rate and alters waveform morphology, “matching heart rate” effectively conditions on this pathway, potentially creating spurious associations. The remedy is to report disease composition before and after matching and conduct sensitivity analyses within subsets labeled with specific diseases.

    Second: Matching Numbers but Not Physiology. A study of 48 pacemaker recipients showed that atrial, AV, and ventricular pacing modes significantly affect the “heart rate-ejection duration” relationship ($P = 0.01$); even with fixed heart rates, different pacing modes showed significant differences in AIx and pulse pressure amplification ($P < 0.0001$) (PMID 27471239). Therefore, heart rate matching must simultaneously lock the source of the beat.

    Another Reverse Trap

    Normalization can strip heart rate but may also discard true information. A 2023 methodological study comparing waveforms found that without normalization, the actual morphological differences at the tail ends of two waveforms were minimal; however, due to different heart rates, the tails became the primary source of cumulative distance, completely drowning out morphological differences; only with double normalization did the focus truly shift to morphology—at the cost of removing heart rate and pulse pressure from main features, relegating them to secondary calculation (PMID 36767962). “Controlling heart rate” and “discarding heart rate information” are two distinct things; the latter would inadvertently destroy the discriminative basis for Pulse Counting and Late Pulse.

    Heart rate itself is not clean: a study of 52 pacemaker recipients showed that after stripping blood pressure effects, carotid-femoral pulse wave velocity still increased by 0.16-0.20 m/s per 10 bpm (PMID 27245180).

    How to Read the Results

    Criteria are fixed in advance to avoid post-hoc interpretation:

    • Retained discriminative power ≥ 80% → The category is primarily a true shape signal; heart rate is not the main driver.
    • Retained 50-80% → Mixed type; must explicitly state in reports that it “contains a heart rate component” and provide the proportion contributed by heart rate.
    • Retained < 50% → Primarily a disguise of heart rate; existing literature discriminative power for this category should be downgraded overall.
    • Positive Control (Pulse Counting/Late Pulse) did not collapse → Pipeline leakage exists; all conclusions are void and must be re-run.

    This is not another article “proving pulse diagnosis is useful.” It performs an audit: before handing pulse qualities to any classifier, first confirm the classifier is learning the pulse quality, not the speed of the heartbeat.

    Evidence and Limitations

    Verified: Significant negative linear correlation between AIx and heart rate under increasing pacing, ~4% drop per 10 bpm increase, ejection duration correlation $r = -0.51$ (PMID 10811742); AIx correlation with ejection duration $r = 0.88$, ejection duration explains 78% variance, strain rate unrelated to AIx after heart rate adjustment (PMID 19720955); adjusting AIx to 75 bpm is inappropriate physiologically and statistically (PMID 24257465); time-based parameters in stress pulse waves driven by heart rate, amplitude-based by systolic pressure (PMID 26406041); SDPTG D:A and AGI negatively correlated with heart rate, B:A positively correlated, associations hold after age/BP adjustment (PMID 15831356, PMID 12473866); DTW without normalization is dominated by heart rate in waveform distance, normalization removes heart rate and pulse pressure info (PMID 36767962); significant AIx differences across pacing modes even at fixed heart rates (PMID 27471239); pulse wave velocity increases 0.16-0.20 m/s per 10 bpm after stripping blood pressure effects (PMID 27245180); floating pulse automatic threshold in 45 hypertensives fell at heart rate ≥ 58.25 (PMID 26988431).

    Partially Verified: The inference that “a significant proportion of shape-based pulse discriminative power comes from heart rate” is a reasonable extrapolation from the above to pulse classification, but no study has directly performed heart rate-matched pulse discrimination controls in the same cohort; the threshold of “≥ 80% indicates true shape signal” is an adjudication standard set by this paper, not existing consensus.

    To Be Verified: ① Actual retained AUC proportions for Floating/Sunken/Wiry/Slippery after heart rate matching; ② Whether Pulse Counting/Late Pulse truly collapse to 0.5 as positive controls (indicating pipeline leakage); ③ Whether heart rate matching introduces collision bias that creates spurious associations in this specific context.

    Related Reading

    Evidence Level: B (Direct literature exists for the dominant role of heart rate on waveform timing and shape, AIx’s chronotropic dependence, SDPTG’s heart rate dependence, and the double-edged nature of normalization; “heart rate matching separation test” for pulse classification has no precedent and is a research proposal).
    References: PMID 10811742, PMID 19720955, PMID 24257465, PMID 26406041, PMID 15831356, PMID 12473866, PMID 36767962, PMID 27471239, PMID 27245180, PMID 26988431


    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.