An idealised model of one reentrant loop, with intra-atrial reentrant tachycardia after congenital heart surgery as the clinical setting in mind: how conduction velocity and refractoriness set the wavelength, and what is left over as the excitable gap.
The ring is a loop of excitable tissue of fixed path length with no branch points. One wave travels round it. Each cell obeys the two-current membrane model of Mitchell & Schaeffer 2003, and neighbouring cells are coupled by diffusion, so the wave propagates at a velocity the model works out for itself rather than one that is imposed.
The wave is drawn as one arc, with the arrowhead at its wavefront. Its action potential is the part that is depolarised. Behind that the refractory tail is tissue that has repolarised but whose gate has not yet passed the capture threshold for one stimulus strength. The rest, in teal, is the excitable gap. Wavelength is the action potential plus the tail, so wavelength plus gap is always the whole path length. A wavefront is a stronger stimulus than that threshold assumes, so it can keep going through tissue inside the tail. The cycle length starts to swing when the wavefront reaches tissue so recently recovered that the action potential’s length depends steeply on the interval before it.
Does the head hit the tail? In this model, usually not head-on when a circuit is shortened gradually. As the head nears the tail it enters tissue that has repolarised but not recovered, and slows, as Chen 2005 describe. The action potential it leaves there is short, which moves the tail away, so it can chase its tail for dozens of beats. When it does fail, it stops in that tissue with the gate ahead of it near 0.2, usually a few millimetres short of tissue still depolarised. A sudden or large rise in velocity or refractoriness is different: the depolarised arc can fill the ring, and the head then runs straight into it.
Every slider acts on the circuit while it runs. Changing the path length adds or removes tissue inside the excitable gap, so the wave itself is never cut. Vinet 1994 and Chen 2005 also reached their short rings by shortening a long one, though they removed depolarised tissue instead. A large change takes a few beats to complete, because only the gap can be cut and only the part of it away from the recording site. The slider stops at 7 cm; cutting pauses while the gap is closed.
Velocity. With refractoriness at 1.00, a running 12 cm circuit stays steady up to about 65 cm/s and a 20 cm one up to about 110 cm/s; beyond that it oscillates or stops, and a restart fails above about 85 cm/s. Shorten refractoriness to go faster.
Why the refractoriness slider has to move. At 1.00 the published membrane values give a refractory period of about 340 ms, far longer than the 114 ms Rensma 1988 measured in canine atrium. A conduction velocity in the physiological range therefore needs a shorter refractory period for the circuit to settle. Below about 0.4 the separation of time scales the model’s closed forms assume is getting weak, though the equations are still integrated exactly as written.
What the drug control is. A direct shift of conduction velocity and refractoriness, nothing more. At 100% it applies the changes Rensma 1988 measured at Bachmann’s bundle in conscious dogs after propafenone: conduction velocity ×0.78 and refractory period ×1.30. It is not sodium-channel kinetics and has no use dependence, so it cannot show why a class Ic effect grows with heart rate. Calling propafenone class Ic is standard usage, not a statement in the papers read for this page unverified.
The wavefront met tissue that had not recovered: reentry stopped. ✕ marks where. Change a slider, then restart.
measured from the running model
MEASURED FROM THE RUNNING MODEL
model a property of this model, not a measurement in any species.
—
Experimental teaching animation. A circuit, not a patient: no treatment, drug or dosing advice, and its numbers belong to the model. Model and sources.
Rensma 1988 measured refractory period and conduction velocity directly in chronically instrumented conscious dogs and multiplied them to get the wavelength of the atrial impulse. Wavelength classified the response to a test stimulus better than either refractory period or conduction velocity alone: 75% of 750 responses correct, against 48% for refractory period and 38% for conduction velocity (their Table 5).
| Condition | Refractory period ms | Conduction velocity cm/s | Wavelength cm |
|---|---|---|---|
| Control (n = 18) | 114 ± 7 | 127 ± 10 | 14.5 ± 1.7 |
| Propafenone (n = 5) | 148 ± 12 | 99 ± 8 | 14.6 ± 1.0 |
| Lidocaine (n = 11) | 147 ± 11 | 118 ± 9 | 17.2 ± 1.7 |
| Quinidine (n = 14) | 153 ± 9 | 115 ± 10 | 17.7 ± 1.6 |
| d-Sotalol (n = 4) | 161 ± 19 | 123 ± 4 | 19.9 ± 2.6 |
During normal rhythm the wavelength of the atrial impulse was 14 to 18 cm. Shortening it below the cut-offs the authors derived by discriminant analysis marked the onset of each arrhythmia: rapid repetitive responses below 12.3 cm, atrial flutter below 9.7 cm, and atrial fibrillation below 7.8 cm.
These are canine numbers, and they must not be read against the simulator. Every value in this section was measured in dogs. At its default settings the simulator’s conduction velocity is about 26 cm/s against the 127 ± 10 cm/s above, and its refractory period about 337 ms against 114 ± 7 ms: about a fifth of the canine velocity and about three times the canine refractory period. Its default wavelength of about 8.7 cm looks plausible only because those two errors oppose each other, and it also depends on where the model’s gate threshold is set. That it falls between the flutter and fibrillation cut-offs is an accident of those errors and says nothing about which arrhythmia this circuit resembles. No equivalent wavelength thresholds for children, or for a repaired congenital heart, are given here. Propafenone shortened nothing: it lengthened the refractory period and slowed conduction by almost the same proportion, so the wavelength did not move.
This is the question the animation is built around, and it is open. A sodium-channel blocker moves two things at once: it slows conduction, which lengthens the cycle length on a fixed circuit, and it lengthens refractoriness. Which moves more decides whether what is left over grows or shrinks, and the measured studies do not agree with each other.
Two different gaps. The total excitable gap runs from the shortest coupling interval that captures to the cycle length. Its fully excitable part is the flat portion of the reset response curve, where the return cycle does not prolong (Kus 1991, Derakhchan 1994). The simulator’s gap is the model’s analogue of the total gap. At the default settings and on all four scenarios this circuit has no fully excitable gap: its reset curve has no flat portion, because the gate never fully recovers (0.85 at the default when the next wave arrives).
Four studies in dogs measured the excitable gap: Kus 1991, Derakhchan 1994, Kawase 2003 and Niwano 1994. The table is under Model, equations and sources.
Unresolved. Kus and Kawase were read in full; Derakhchan and Niwano as abstracts, so the conditions behind their numbers are not restated here. The teaching diagram that shows a class Ic drug simply closing the excitable gap is not supported by the class Ic data here: propafenone unverified left the duration unchanged (Derakhchan), with the fully excitable part reduced or abolished in some cases, and pilsicainide widened the gap, in fibrillation (Kawase). Procainamide, class IA, also left it unchanged (Kus). The simulator does not settle it, and this page does not assert a direction.
What the model above does is worth watching for one specific reason: move the drug slider on a fixed path length and the gap measured in centimetres and the gap measured in milliseconds need not move together, because slowing conduction stretches the same arc of tissue over a longer time. A study that reports the gap in milliseconds and one that reports it as a fraction of the circuit can disagree while both are right about their own measurement.
What this model does, and why it is not evidence. Take the drug control from 0 to 100% on a 12 cm circuit. It multiplies conduction velocity by 0.78 and refractoriness by 1.30, so the wavelength barely moves, as Rensma 1988 measured for propafenone in one preparation; in Kawase 2003 it fell. With the wavelength pinned on a fixed circuit, the gap in centimetres is nearly pinned too (3.3 to 3.5 cm), and the gap in milliseconds, that arc divided by a velocity about a quarter lower, can only grow: 130 to 176 ms. Both are consequences of the two numbers put in, not findings. The direction agrees with Kawase 2003, a mother rotor in fibrillation, which this one-dimensional ring does not represent; Kus 1991 (procainamide) and Derakhchan 1994 found no change in gap duration.
This is a published membrane model run on a uniform one-dimensional ring, built to show how wavelength and the excitable gap behave. The numbers it prints are properties of the model, not measurements from any species. What it does and does not reproduce is set out below.
Mitchell & Schaeffer’s two-current model. The voltage v is dimensionless on 0 to 1; the authors note it can be scaled back with v̅ = vmin + v(vmax − vmin), “for example vmin = −70 mV and vmax = 30 mV”. A gate h opens in rest and closes above a changeover voltage vgate:
dv/dt = h·v2(1 − v)/τin − v/τout
dh/dt = (1 − h)/τopen if v < vgate, else −h/τclose
The parameter values are the only set the paper prints, from its Figure 1: τin 0.3 ms, τout 6 ms, τopen 120 ms, τclose 150 ms, with vgate = 0.13 “throughout this paper” and vstim = 0.1. The refractoriness slider multiplies τopen and τclose together, which leaves their ratio, and so the model’s dynamical regime, unchanged; at 1.00 the published values are in force. Action potential duration is the time v is above vgate, the paper’s own definition.
Mitchell & Schaeffer is a membrane patch. The authors describe “a uniformly polarized membrane patch… a spatially clamped ventricular myocyte”: their paper contains no diffusion term, no space and no conduction velocity. The cable term D·∂2v/∂x2 and the diffusion coefficient are therefore not theirs. D = 1.12 cm2/s is the value Chen 2005 used for a one-dimensional ring of cardiac tissue, read in the corrected and republished version of record and confirmed identical to the original. Pairing that coefficient with these membrane constants is still this page’s own choice and appears in no paper read for this page unverified, so everything the pairing produces is measured here rather than quoted: the conduction-velocity slider sets the coupling strength needed to deliver the velocity you ask for, and the velocity in the results panel is then measured from the running wave. The animation uses 0.04 cm cells, which is coarser than anything Chen tested: they report that halving their cell size from 0.02 to 0.01 cm moved conduction velocity by under 5% and action potential duration by under 0.1%. On our grid the measured velocity runs about 2 to 3% below a fine grid above 25 cm/s, and about 7% below it at 15 cm/s.
Integration is explicit, with the step size held below both the diffusion limit and a fraction of τin. This page starts a wave by seeding a long loop and then shortens it to the path length you choose, because its seeding method needs recovered tissue ahead of the wave. Vinet 1994 and Chen 2005 also shortened long rings, but cut depolarised tissue: Vinet removed “a ring section corresponding to the action potential plateau” and Chen eliminated “a 1-cm portion of the cable 0.2 cm behind the wavefront”. The slider cuts only the excitable gap, so moving it never deletes part of the wave; the cost is that it pauses when that gap closes. Either way a cut disturbs the beat it lands in, so that beat is discarded rather than reported; Chen likewise excluded the first beat after each cut from their figures “because of the cut-induced disturbance”.
The refractory period printed in the results is the model’s own: the time from activation until the gate h recovers past the threshold value hthr of the paper’s equation (18). It is tagged model because it is not a clinically measured effective refractory period, and how the two correspond is an open question rather than something this page can state. Wavelength is that refractory period times the measured conduction velocity, after Rensma 1988, and the excitable gap is the path length minus it, so the two always add to the path length. When that product cannot be formed, because a path-length change has just discarded the beat, because the gate at the recording site had not recovered, or because that beat’s local velocity times its refractory period comes to more than the whole circuit, the panel reads both lengths from the drawn ring instead.
The excitable gap shown is the cycle length minus that refractory period. That is how Kus 1991 and Kawase 2003 construct their excitable gap, the total in Niwano’s terms, with a capture threshold at its edge, so it is the model’s analogue of the total gap and its size depends on the stimulus strength assumed model. At the gap’s edge tissue conducts at about 60% of the rested velocity. A wavefront is a stronger stimulus, so a beat can show no gap at the recording site while reentry continues: in the oscillating range described below about half the beats are like that. A flat portion of the reset curve appears when the gate has time to recover before the next wave, for example at 20 cm with refractoriness 0.40, or at 20 cm and 20 cm/s with refractoriness 1.00. Recovery is exponential, so it comes on gradually. Kus 1991 note that earlier animal models had a gap of only partially refractory tissue and purely increasing reset curves; the default here behaves like those.
What it reproduces. As a ring is shortened, this model goes from steady reentry to sustained cycle-length oscillation to block: the sequence Vinet 1994 reports and two of Chen 2005’s three parameter sets show. At D = 1.12 cm2/s (about 36 cm/s requested) oscillation begins smoothly at about 6.1 to 6.2 cm depending on the grid, its swing grows to about 100 ms, and reentry fails at about 4.7 cm. At the page’s default coupling (30 cm/s requested, D = 0.78 cm2/s) the same sequence sits lower, with oscillation from about 5.1 cm and block near 4 cm: lengths scale with √D. Below onset a pattern of long and short intervals repeats every 1.5 to 1.6 turns, against Vinet’s 1.73 to 1.79 and Chen’s 1.44 and 1.90 (10.1 and 13.3 cm on a 7 cm ring). At one site that reads as long-intermediate-short rather than strict alternation, which Vinet and Frame & Simson 1988 also describe. After a premature stimulus on a longer ring, action potential duration alternates long-short for 4 to 5 beats, in the order Quan 1990 report at 6 cm, though theirs lasted about 30 beats, and the cycle length settles more slowly the closer the ring is to onset.
What it does not reproduce. Vinet’s narrow aperiodic zone; Quan’s finding of no alternans in conduction velocity (Vinet and Chen do see it vary; here it moves with action potential duration after a premature stimulus and swings about three-fold in the settled oscillation); diastolic intervals and action potential durations swinging much more than cycle length, as Frame & Simson report (about 1.2 times here); Chen’s branching of restitution. Not looked for: Vinet’s third factor, triggered waves of secondary repolarization.
Why none of this validates the model. The membranes differ, so only the sequence and directions are shared. Several of these behaviours also appear with other time constants in this model, including ones with no oscillating range at all, and agreement in pattern does not mean agreement in mechanism. Lengths belong to their membrane and coupling: Vinet’s stable reentry above 19.6 cm and block below 12.8 cm to a modified Beeler-Reuter membrane, and Chen’s stability ending at 10 and 8 cm with failure at 6 cm, and failure at 3 cm in their third set, to a Fenton-Karma one.
Where the swing starts. In this model it starts where the action potential’s length begins to depend steeply on the interval before it: restitution slope 1, a factor Vinet names. From Mitchell & Schaeffer’s equation (17) that falls at gate value 1/(1+r), with r = τopen/τclose, which is 0.56 with the published constants (this page’s own derivation). Those constants also lie on the edge of the paper’s equation (45), so here the swing begins as the printed gap closes; that is a coincidence of the parameter set. Changing the time constants separates them: at τopen 30 ms oscillation starts near 9.4 cm with the gap still open, and at 300 ms near 2.6 cm with it closed, both at D = 1.12 cm2/s. The refractoriness slider moves both constants together, so it cannot show this.
Reaching that range here. The figures above come from the published shortening method, which removes tissue just behind the wavefront, here in 0.2 cm steps (Chen cut 1 cm at a time). The path-length slider stops at 7 cm. At the default velocity that is above onset, so length alone will not reach it. With refractoriness 1.00 it does from about 42 cm/s: dragged to 7 cm the circuit swings by about 60 ms at 45 cm/s and about 90 ms at 50 cm/s, and at 55 cm/s or more it stops. Cutting pauses when the printed gap closes and resumes on beats where the oscillation reopens it. At the default playback speed this can take a minute or two. Raising refractoriness or velocity on a running circuit also reaches it: try 12 cm, 60 cm/s and refractoriness 1.40. What a circuit does can depend on how it got there: at 20 cm, 120 cm/s and refractoriness 1.00 a circuit dragged there keeps going, while a restart at the same settings fails.
| Study | Preparation | Reported effect on the excitable gap |
|---|---|---|
| Kus 1991 | Canine atrial flutter around the tricuspid valve, 7 open-chest dogs | Procainamide, class IA, lengthened cycle length 144 ± 10 to 160 ± 13 ms, slowed conduction 73 ± 6 to 66 ± 6 cm/s and lengthened the shortest coupling interval achievable 101 ± 13 to 116 ± 17 ms, with no change in the excitable gap: 44 ± 17 to 39 ± 16 ms in the Results (the abstract prints 38 ± 9 and 40 ± 18). Its flat, fully excitable part, present in six of seven experiments, was 9 ± 4 and 8 ± 3 ms |
| Derakhchan 1994 abstract | Canine atrial flutter around the tricuspid valve | Procainamide and propafenone lengthened refractoriness and cycle length without changing the duration of the gap; on propafenone, in some cases, the fully excitable part of it was reduced markedly or even eliminated |
| Kawase 2003 | Canine atrial fibrillation, isolated perfused atria, a mother rotor, 10 experiments | Pilsicainide, class IC, a pure sodium-channel blocker, widened the excitable gap 17 ± 3 to 42 ± 6 ms (+147%). Fibrillation cycle length 81 ± 12 to 132 ± 15 ms, refractory period 64 ± 7 to 90 ± 8 ms, conduction velocity 75 ± 5 to 45 ± 5 cm/s, core perimeter 10 ± 4 to 20 ± 6 mm (+100%), and wavelength fell, 4.8 ± 0.4 to 4.1 ± 0.4 cm |
| Niwano 1994 abstract | Canine sterile pericarditis flutter, a functionally determined circuit | A fully excitable gap is present in at least part of the circuit, and the gap is not uniform round it: the total (fully plus partially) excitable gap was 12 ± 4 ms in areas of slow conduction against 22 ± 6 ms elsewhere, while the fully excitable gap was 4 ± 1 ms |
abstract marks a source read as its published abstract rather than its full text. model marks a quantity that is a property of the model and not a clinical measurement. unverified marks something this page relies on that has not been checked against a primary source. Every citation above was checked field by field against its PubMed record.
Not verified. This is an experimental prototype, and no official calculator or reference implementation of this model exists to check it against. It has only the comparison with published descriptions above, which is weaker. Its numbers belong to the model. No number on this page should be used for a clinical decision.
The model runs in this browser; nothing you set on this page is stored or sent.
Educational use only. Experimental teaching animation. A circuit, not a patient: no treatment, drug or dosing advice, and its numbers belong to the model.