5D Library › 自主神经交感迷走平衡、六经辨证状态空间与多模态标志物反演 · English

第二章 交感与迷走的双室模型与五种状态

一个可以计算的双室模型

把交感活动 SS 与迷走活动 VV 写成两个相互作用的「室」:各自向中枢驱动回落,彼此抑制(系数 cc),压力反射再把交感引起的血压上升转为迷走制动(增益 gg):

dSdt=uS−S−cV2,dVdt=uV+(g−c)S−V.\frac{dS}{dt} = \frac{u_S – S – cV}{2},\qquad \frac{dV}{dt} = u_V + (g – c)S – V.

这是一个示意模型,不是生理定律;它的用处在于把「交感过载」「迷走撤退」「压力反射坍缩」「双侧耗竭」这些说法写成可以比较的参数设定。心率与 RMSSD 由 SS、VV 线性读出。模型运行到静息后施加一次应激脉冲,记录交感活动回到静息值一成以内所需的时间。

推导

# Chapter 2. p03: A two-compartment model of autonomic tone. S is
# sympathetic and V vagal activity (arbitrary units, >= 0). Each relaxes
# to its central drive, they inhibit each other (c = 0.3), and the
# baroreflex turns sympathetic pressure into vagal braking with gain g:
#   dS/dt = (uS - S - c V) / 2      dV/dt = (uV + g S - c S - V) / 1
# Heart rate and RMSSD are read off as HR = 70 + 40 S - 25 V and
# RMSSD = 12 + 40 V - 6 S (illustrative). Each regime is run to rest,
# then given a stress pulse (uS + 1 for 5 time units); the recovery time
# is when S returns within 10% of its resting value. Euler step 0.01.
REGIMES = (("balanced", 0.5, 0.8, 0.6),
           ("sympathetic overload", 1.4, 0.8, 0.6),
           ("vagal withdrawal", 0.5, 0.25, 0.6),
           ("baroreflex collapse", 0.5, 0.8, 0.0),
           ("dual depletion", 0.15, 0.15, 0.6))
C = 0.3

def step(S, V, uS, uV, g, dt=0.01):
    dS = (uS - S - C * V) / 2.0
    dV = (uV + g * S - C * S - V) / 1.0
    return max(0.0, S + dt * dS), max(0.0, V + dt * dV)

def run(uS, uV, g):
    S = V = 0.5
    for _ in range(5000):
        S, V = step(S, V, uS, uV, g)
    rest = (S, V)
    t, back = 0.0, None
    for i in range(6000):
        pulse = 1.0 if i < 500 else 0.0
        S, V = step(S, V, uS + pulse, uV, g)
        t = (i + 1) * 0.01
        if i >= 500 and back is None and abs(S - rest[0]) <= 0.1 * max(
                rest[0], 0.05):
            back = t - 5.0
    return rest, back

if __name__ == "__main__":
    print("   regime                  S     V     HR  RMSSD  recovery")
    for name, uS, uV, g in REGIMES:
        (S, V), back = run(uS, uV, g)
        hr, rm = 70 + 40 * S - 25 * V, 12 + 40 * V - 6 * S
        print("   {:<21} {:5.2f} {:5.2f}  {:4.0f}  {:5.1f}  {:6.2f}".format(
            name, S, V, hr, rm, back))
#    regime                  S     V     HR  RMSSD  recovery
#    balanced               0.24  0.87    58   45.4    6.14
#    sympathetic overload   1.06  1.12    85   50.4    3.66
#    vagal withdrawal       0.39  0.37    76   24.3    5.33
#    baroreflex collapse    0.29  0.71    64   38.9    8.21
#    dual depletion         0.10  0.18    69   18.6    7.63

均衡状态下,心率 58,RMSSD 45.4,恢复时间 6.14。交感过载时心率升到 85,但 RMSSD 反而是 50.4:压力反射把升高的交感转成了更强的迷走制动,两者同时升高。迷走撤退时 RMSSD 降到 24.3;压力反射坍缩时静息读数与均衡相近(心率 64,RMSSD 38.9),差别在于恢复变慢(8.21);双侧耗竭时 RMSSD 只有 18.6。

五种设定的静息点(圆点)与应激脉冲后的轨迹(p03)。

短时记录分得开吗

推导

# Chapter 2. p04: Can a short recording tell the five regimes of p03
# apart? Each simulated person sits at a regime's resting point, with
# person-to-person spread and 3-minute measurement noise: HR sd 5 bpm,
# RMSSD sd 7 ms, recovery time sd 1.2. A nearest-centre rule (scaled by
# those sds) assigns the regime. Compared: resting HR and RMSSD only,
# and the same plus the recovery time after a standard stressor
# (for example a timed breathing or stand-up test). 2,000 people per
# regime, seed 86.
import importlib.util
import random

def load(name):
    spec = importlib.util.spec_from_file_location(name, name + ".py")
    m = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(m)
    return m

p03 = load("p03_two_compartment")
SD = (5.0, 7.0, 1.2)

def centres():
    out = []
    for name, uS, uV, g in p03.REGIMES:
        (S, V), back = p03.run(uS, uV, g)
        out.append((name, (70 + 40 * S - 25 * V, 12 + 40 * V - 6 * S,
                           back)))
    return out

def classify(x, cs, k):
    return min(range(len(cs)), key=lambda j: sum(
        ((x[i] - cs[j][1][i]) / SD[i]) ** 2 for i in range(k)))

if __name__ == "__main__":
    rng = random.Random(86)
    cs = centres()
    for k, label in ((2, "HR + RMSSD"), (3, "HR + RMSSD + recovery")):
        hits, conf = 0, {}
        for j, (name, c) in enumerate(cs):
            for _ in range(2000):
                x = [c[i] + rng.gauss(0, SD[i]) for i in range(3)]
                g = classify(x, cs, k)
                hits += g == j
                conf[(j, g)] = conf.get((j, g), 0) + 1
        print("   {:<22} correct {:.3f}".format(label, hits / 10000))
        worst = max(((a, b) for (a, b) in conf if a != b), key=conf.get)
        print("     most common mix-up: {} read as {} ({:.3f})".format(
            cs[worst[0]][0], cs[worst[1]][0], conf[worst] / 2000))
#    HR + RMSSD             correct 0.783
#      most common mix-up: baroreflex collapse read as balanced (0.238)
#    HR + RMSSD + recovery  correct 0.876
#      most common mix-up: baroreflex collapse read as balanced (0.130)

只用三分钟的心率与 RMSSD,五种状态的正确分类率是 0.783;最常见的错误是把压力反射坍缩读成均衡(0.238)。加上一项标准应激(例如定时呼吸或站立试验)后的恢复时间,正确率升到 0.876,同一种错误降到 0.130。这是在模型完全正确、只有测量噪声的情况下的上限。

假说 · 尚未验证

作者的诠释(尚未验证): 医家所说的「阳亢」「阴虚」「气陷」「亡阳」,可能分别对应交感过载、迷走撤退、压力反射增益坍缩与双侧耗竭四种自主神经状态。什么会证明它错: 在由中医师独立辨证、再测量静息心率、RMSSD 与应激恢复时间的研究中,四种证候在这三项指标上的分布没有可重复的差别。迷走神经背核与疑核的分工(多迷走神经理论)本身在学界也仍有争议 (Porges 2011; Grossman 2023)。

小结

在双室模型中,「交感过载」并不一定降低 RMSSD,因为压力反射会同时提高迷走制动;压力反射坍缩在静息读数上几乎看不出来,要靠应激后的恢复时间才能分辨。把中医证候映射到这些状态,是一项需要检验的假说。

参考文献

Grossman, Paul. 2023. “Fundamental Challenges and Likely Refutations of the Five Basic Premises of the Polyvagal Theory.” Biological Psychology 180: 108589.
Porges, Stephen W. 2011. The Polyvagal Theory: Neurophysiological Foundations of Emotions, Attachment, Communication, and Self-Regulation. W. W. Norton.

Fan, Y. (2026). Autonomic Sympathovagal Balance, Six-Meridian State-Space Models and Multimodal Biomarker Inversion: Heart-Rate Variability Spectral Decomposition, Thermoregulatory Phase Drift and Bayesian Syndrome Classification [自主神经交感迷走平衡、六经辨证状态空间与多模态标志物反演] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090227 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).