四个嫌疑人共用一张脸:EIT 的「低阻带」到底该归谁

Written by

in

Language

Evidence level: B

想用 EIT 分出”这条低阻带是经络、不是血管不是神经不是筋膜”,难的地方不在分辨率——四个人里有两个在电导率上本来就分不开,唯一能分开神经的那个信号,恰好不在环形阵列的电流方向上。

正文

“低阻”这两个字,先就把神经排除了

源设计列了四个候选结构:血管(高电导)、神经(活动依赖)、筋膜(低电导)、脂肪(低电导)。这套分类在纸面上很清楚,一落到数字上就出问题。

按组织介电参数汇编(Gabriel 参数模型,PMID 8938025、8938026),50 Hz 到 500 kHz 区间内,血液稳定在 0.7 S/m 上下,骨骼肌 0.23 至 0.45 S/m,脂肪 0.02 至 0.05 S/m,神经约 0.03 至 0.11 S/m。

也就是说:神经静止时的电导率跟脂肪基本重叠,只有肌肉的三分之一到十分之一。它根本不”低阻”,它是”低”——低到跟绝缘层差不多。

要找一条”低阻带”,最可能的两个主人是血管(0.7,最高)和肌肉(0.35,且带各向异性)。神经和脂肪在这个量级上是同一团。这条推论有点不舒服:源设计把神经排在候选第二位,但从电导率量级看,它是最不可能被”低阻”这两个字指认的那个。

更麻烦的是脂肪。10 kHz 下脂肪电导率被报到 0.02(Gabriel 1996)、0.08(Kangasmaa 2022)、0.09(Gabriel 2009)、0.15(Wake 2016)、0.41(Stoneman 2007)S/m——同一个组织,文献值跨了 20 倍(PMID 39825696 综述了这批数据,指出差异可能来自含水量)。一个连自己都定不下来的数,当不了参照物。

各向异性:最像经络的那个”假朋友”

肌肉不是一块均匀导体。低频下横向与纵向电导率之比约 1:5;10 kHz 到 1 MHz 之间,这个比值从 37% 升到 64%(PMID 36228622,10 名受试者、MRI 建模、有界 EIT 反演)。

这句话的意思是:沿肌纤维方向电流好走,横着走难走五倍。

四肢的肌纤维大体沿肢体长轴排列。经络也沿肢体长轴排列。

于是在一个肢体横断面的重建图上,”沿长轴的一条高电导带”和”沿长轴的一条经络”会落在同一个方位、有相近的宽度。DT-MREIT 的对照说明这不是小事:同一个肌肉样本,各向异性重建出的电场图覆盖范围比各向同性方法宽两倍以上(PMID 28446015)。把肌肉当各向同性处理,误差不是 5%,是”整张图都歪了”。

结论很难听:“低阻带”最可能的解释不是新增了一个结构,而是肌肉的方向性被投影到了截面上。

方向悖论:环形阵列的电流,不走神经信号最强的那个方向

现在处理源设计里最有希望的一条区分路径——神经兴奋时阻抗会变,拍两张做差。

这条路在实验室里成立,在四肢环形阵列上不成立。

Hope 等人的神经袖套模型给了两个数(PMID 29547133):1 kHz、电极间距 10 mm 时,神经束内组织活动时的阻抗变化比例,横向 −8.8×10⁻⁴,纵向 −0.30。

差 340 倍。

而环形电极阵列的电流,从体表一对电极注入、沿截面绕过去,主要分量垂直于肢体长轴——也就是横向。这个几何恰好落在最差的那个方向上。

所以”神经(活动依赖)”这条路径,前提是要有纵向电流几何(沿神经走行注入、纵向差分采集)。那就不是四肢环形 EIT,是神经袖套 EIT。两个装置不能借用彼此的结论。

补一句量级:整根神经的实测阻抗变化约 1%,单根纤维的膜阻抗变化可到 10%,但后者要在一根纤维上测,不是在一条腿上测。

出路在时间轴,不在空间轴

四个人里两个在电导率上分不开,一个方向不对,那还剩什么?

剩时间。血管有脉搏(约 1 Hz),神经有活动(毫秒级),筋膜什么都没有。Halter 等人用脉搏血氧门控的前臂 EIT,把搏动性电导变化和血氧波形对上了——频谱峰值分别在 1.19 Hz 和 1.21 Hz(PMID 19162879),并报告体模上小于 1 mm 的电导变化可被这套动态系统检出。

空间上分不开的东西,可以在时间上分开。这是本条路线真正该走的方向,而源设计把它放在了第四位。

最后一个坑:别用答案去改题干

还有一个陷阱比前三个隐蔽。

如果反演时用 MRI 分割图做硬先验或软先验约束,重建图会显著变好看——结构性先验能明显提升重建图与解剖图的结构相似度。但那时你面对的已经不是”EIT 看见了什么”,而是”EIT 被允许看见什么”。

所以区分度检验必须是盲的:反演阶段不给解剖信息,反演完成后再与 MRI/超声比对;预注册比对指标(Dice、中心距、半高宽、连续性断点数),并同时报告”先验关闭”与”先验开启”两组结果的差值。这个差值本身,就是这篇设计最该产出的一个数。

古人说的是”路上跑什么”,不是”墙是什么砖”

《灵枢·本脏》给经脉下的定义是:”经脉者,所以行血气而营阴阳,濡筋骨,利关节者也。”

“行血气”三个字里,血是血管里的东西,气是功能性的东西。古人把它们放在同一条通道上讲,从没说过这条通道本身是血管、是神经、还是筋膜。

《素问·调经论》说得更直白:”五脏之道,皆出于经隧,以行血气。”——隧。隧道的本质不是墙壁的材质,是”能过东西”。

所以现代研究问”经络是四种结构里的哪一种”,这个问法古人压根没提过。古人提的是”一条路上跑着三种东西”。

这也给出了本条路线最值钱的一个转向:如果 EIT 测出的低阻带,能被”肌肉各向异性 + 血管 + 筋膜 + 神经”的联合模型完整解释掉,那不等于”经络不存在”——那等于”经络就是那个联合”。把四个嫌疑人拼起来,可能才是那张脸。

证据与局限

已验证: 各组织电导率谱的量级与频率依赖已有系统测量(PMID 8938025、8938026);在体人体肌肉与脂肪的各向异性电导率已被有界 EIT 结合 MRI 建模定量估计,横向/纵向比随频率从 37% 升到 64%(PMID 36228622);忽略各向异性会导致重建电场图显著失真,覆盖范围差两倍以上(PMID 28446015);神经活动的阻抗变化在纵向几何下可达 −0.30,而横向只有 −8.8×10⁻⁴(PMID 29547133,有限元神经袖套模型);前臂 EIT 可捕获与心搏同步的搏动性电导变化,频谱与脉搏血氧波形吻合(1.19 Hz vs 1.21 Hz,PMID 19162879)。

部分验证: EIT 的对比度检出能力高度依赖电极协议与噪声水平(PMID 18215899 建立检出概率框架;PMID 21646709 证明传统相邻注入/测量模式的表现比优化模式差 6.9 倍);高对比小目标(1 cm 量级)在超声耦合阵列下可辨(PMID 20647618);EIT 与 CT 融合时对低对比介质(蒸馏水/生理盐水)的对比判别可达 47%,但结构相似性明显低于 CT(PMID 24877814)——即 EIT 的强项是功能对比度,弱项恰是结构保真度。

待验证: ① 四肢环形阵列下,重建”低阻带”中可归因于肌肉各向异性的比例有多大——目前无任何研究做过这项归因分解;② 提高阵列密度后,横向神经活动信号能否被推到可测范围(模型预测为 0.09%,实测噪声底待标定);③ 若把”肌肉各向异性 + 血管 + 筋膜 + 神经”作为联合正演模型,其对实测边界电压的解释力是否已接近饱和——若是,则”新增一个结构”在统计上不再是必要假设。第三项是本路线最关键的判决性实验,尚无直接研究。

相关阅读

证据等级: B(各组织电导率、各向异性、神经阻抗变化幅度、搏动性电导变化均有 PubMed 直接支撑;但”四肢环形 EIT 能否把经络低阻带与血管/神经/筋膜分离”这一具体问题尚无直接研究,属方法学设计与归因分析提案)
参考文献: PMID 8938025, PMID 8938026, PMID 39825696, PMID 36228622, PMID 28446015, PMID 29547133, PMID 19162879, PMID 18215899, PMID 21646709, PMID 20647618, PMID 24877814


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。

Four Suspects Share One Face: Who Belongs to the EIT \”Low-Resistivity Band\

Evidence level: B

The goal of using Electrical Impedance Tomography (EIT) to distinguish whether “this low-resistance band belongs to meridians, not vessels, nerves, or fascia” is not hindered by resolution. The difficulty lies in the fact that two of the four candidates are inherently indistinguishable by conductivity alone; the only signal capable of distinguishing nerves happens to lie outside the current direction of the ring array.

Main Text

The Two Characters “Low Resistance” Already Exclude Nerves

The source design lists four candidate structures: vessels (high conductivity), nerves (activity-dependent), fascia (low conductivity), and fat (low conductivity). This classification is clear on paper but breaks down when applied to numerical data.

According to tissue dielectric parameter compendia (Gabriel Parameter Model, PMID 8938025, PMID 8938026), within the 50 Hz to 500 kHz range, blood stabilizes around 0.7 S/m, skeletal muscle ranges from 0.23 to 0.45 S/m, fat ranges from 0.02 to 0.05 S/m, and nerves range from 0.03 to 0.11 S/m.

In other words: The conductivity of nerves at rest overlaps significantly with fat, representing only one-third to one-tenth of muscle. It is not “low resistance”; it is simply “low”—low enough to be comparable to an insulator.

To identify a “low-resistance band,” the two most likely candidates are vessels (0.7, highest) and muscle (0.35, exhibiting anisotropy). Nerves and fat fall into the same magnitude range in this context. This conclusion is somewhat uncomfortable: while the source design ranks nerves as the second candidate, the magnitude of conductivity suggests it is the least likely to be identified by the term “low resistance.”

Complicating matters further is fat. At 10 kHz, fat conductivity is reported as 0.02 S/m (Gabriel 1996), 0.08 S/m (Kangasmaa 2022), 0.09 S/m (Gabriel 2009), 0.15 S/m (Wake 2016), and 0.41 S/m (Stoneman 2007)—a 20-fold variation across literature values for the same tissue (PMID 39825696 reviews this data, suggesting differences may stem from water content). A value that cannot even be defined consistently for itself cannot serve as a reference.

Anisotropy: The “False Friend” Most Like Meridians

Muscle is not a uniform conductor. At low frequencies, the ratio of transverse to longitudinal conductivity is approximately 1:5; between 10 kHz and 1 MHz, this ratio rises from 37% to 64% (PMID 36228622, based on 10 subjects, MRI modeling, and bounded EIT inversion).

This means: Current flows easily along the muscle fiber direction but faces five times more resistance flowing perpendicularly.

Muscle fibers in limbs are generally arranged along the longitudinal axis of the limb. Meridians also run along the longitudinal axis.

Consequently, in a reconstructed cross-section of a limb, “a high-conductivity band running along the longitudinal axis” and “a meridian running along the longitudinal axis” would occupy the same orientation with similar widths. DT-MREIT comparisons indicate this is not trivial: for the same muscle sample, anisotropic reconstruction yields electric field maps with a coverage area more than twice as wide as isotropic methods (PMID 28446015). Treating muscle as isotropic introduces errors not of 5%, but a “complete distortion of the entire map.”

The conclusion is difficult to hear: The most likely explanation for the “low-resistance band” is not the addition of a new structure, but rather the projection of muscle anisotropy onto the cross-section.

The Directional Paradox: Ring Array Currents Do Not Follow the Nerve Signal’s Strongest Direction

Now we address the most promising differentiation path in the source design—varying impedance during nerve excitation and comparing two acquisitions.

This approach holds in the laboratory but fails in four-limb ring arrays.

Hope et al.’s nerve sleeve model provides two values (PMID 29547133): at 1 kHz with an electrode spacing of 10 mm, the impedance change ratio within a nerve bundle during activity is -8.8×10⁻⁴ transversely versus -0.30 longitudinally.

A difference of 340 times.

However, the current in a ring electrode array is injected between a pair of surface electrodes and circulates around the cross-section, with the primary component perpendicular to the limb’s longitudinal axis—i.e., transverse. This geometry coincides exactly with the worst-case direction.

Therefore, the path relying on “nerves (activity-dependent)” requires a longitudinal current geometry (injection along the nerve course and longitudinal differential acquisition). That is not a four-limb ring EIT; it is a nerve sleeve EIT. The two devices cannot borrow conclusions from one another.

A note on magnitude: the measured impedance change for an entire nerve is approximately 1%, while membrane impedance change for a single fiber can reach 10%, but the latter must be measured on a single fiber, not on an entire leg.

The Solution Lies in the Time Axis, Not the Spatial Axis

Two of the four candidates are indistinguishable by conductivity, and one direction is incorrect. What remains?

Time. Vessels exhibit pulses (~1 Hz), nerves exhibit activity (millisecond scale), and fascia exhibits nothing. Halter et al. used pulse-gated EIT on the forearm to align pulsatile conductivity changes with blood oxygen waveforms—spectral peaks occurred at 1.19 Hz and 1.21 Hz (PMID 19162879)—and reported that conductivity changes smaller than 1 mm can be detected by this dynamic system on a phantom.

Things indistinguishable spatially can be separated temporally. This is the direction this line of inquiry should truly follow, whereas the source design placed it in fourth position.

The Final Pitfall: Do Not Modify the Question with the Answer

There is another trap more subtle than the previous three.

If inversion uses MRI segmentation maps as hard or soft prior constraints, the reconstructed image becomes significantly more visually pleasing—structural priors can markedly improve structural similarity between the reconstruction and anatomical maps. However, at that point, one is no longer asking “what did EIT see,” but rather “what was EIT allowed to see.”

Therefore, discrimination testing must be blind: do not provide anatomical information during the inversion stage; compare with MRI/ultrasound only after inversion is complete; pre-register comparison metrics (Dice coefficient, centroid distance, full width at half maximum, number of continuity breaks), and simultaneously report the difference between “prior-off” and “prior-on” groups. This difference itself is the number this design should most critically produce.

The Ancients Spoke of “What Runs on the Road,” Not “What Brick the Wall Is Made Of”

The Ling Shu · Ben Zang defines meridians as: “Meridians are those that circulate blood and qi, nourish bones and tendons, and benefit joints.”

Within the phrase “circulate blood and qi,” blood refers to substances within vessels, while qi refers to functional aspects. The ancients placed these on the same channel but never stated that the channel itself is a vessel, a nerve, or fascia.

The Suan Jing · Tiao Jing Lun states even more directly: “The pathways of the five organs all emerge from the jing-tui to circulate blood and qi.”——Tui (tunnel). The essence of a tunnel is not the material of its walls, but “the ability to pass things through.”

Thus, when modern research asks “which of the four structures are meridians,” this question was never posed by the ancients. The ancients asked “what three things run on one road.”

This also provides the most valuable shift for this line of inquiry: if the low-resistance band detected by EIT can be fully explained by a joint model of “muscle anisotropy + vessels + fascia + nerves,” that does not equal “meridians do not exist”—that equals “meridians are the joint combination.” Putting all four suspects together might indeed be the single face.

Evidence and Limitations

Verified: Magnitudes and frequency dependencies of conductivity spectra for various tissues have been systematically measured (PMID 8938025, PMID 8938026); anisotropic conductivity of in-vivo human muscle and fat has been quantitatively estimated using bounded EIT combined with MRI modeling, with the transverse/longitudinal ratio rising from 37% to 64% with frequency (PMID 36228622); ignoring anisotropy causes significant distortion in reconstructed electric field maps with coverage differences exceeding twice the original area (PMID 28446015); nerve activity-induced impedance changes reach -0.30 under longitudinal geometry, whereas transverse changes are only -8.8×10⁻⁴ (PMID 29547133, finite element nerve sleeve model); forearm EIT can capture pulsatile conductivity changes synchronized with cardiac beats, matching spectral peaks with pulse oximetry waveforms (1.19 Hz vs 1.21 Hz, PMID 19162879).

Partially Verified: EIT contrast detection capability is highly dependent on electrode protocols and noise levels (PMID 18215899 establishes a detection probability framework; PMID 21646709 proves traditional adjacent injection/measurement modes perform 6.9 times worse than optimized modes); high-contrast small targets (~1 cm scale) are discernible under ultrasound-coupled arrays (PMID 20647618); EIT-CT fusion achieves contrast discrimination of 47% for low-contrast media (distilled water/physiological saline), but structural similarity is markedly lower than CT (PMID 24877814)—indicating EIT’s strength lies in functional contrast while its weakness is structural fidelity.

To Be Verified: ① To what extent the “low-resistance band” can be attributed to muscle anisotropy in four-limb ring arrays—currently no study has performed this attribution decomposition; ② Whether increasing array density can push transverse nerve activity signals into a measurable range (model prediction 0.09%, actual noise floor pending calibration); ③ Whether a joint forward model of “muscle anisotropy + vessels + fascia + nerves” already saturates in explaining measured boundary voltages—if so, “adding a new structure” is no longer a statistically necessary assumption. The third item represents the most critical decisive experiment for this line of inquiry, with no direct studies yet.

Related Reading

Evidence Level: B (Direct PubMed support exists for tissue conductivity, anisotropy, nerve impedance change magnitudes, and pulsatile conductivity changes; however, the specific question of whether “four-limb ring EIT can separate meridian low-resistance bands from vessels/nerves/fascia” lacks direct research, representing a methodological design and attribution analysis proposal)
References: PMID 8938025, PMID 8938026, PMID 39825696, PMID 36228622, PMID 28446015, PMID 29547133, PMID 19162879, PMID 18215899, PMID 21646709, PMID 20647618, PMID 24877814


Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *