loading...
Restitution curve APD(DI) with the working point and cobweb of the last beats (grey line: slope = 1)
Experiment: bifurcation diagram, APD against pacing cycle length (the branch splits at CL*)
running...
The science
This is an educational model with synthetic parameters. It is not a diagnostic tool and it does not predict the behaviour of any patient or any drug.
Action potential duration and the diastolic interval. A heart cell fires an action potential (AP) when stimulated, stays depolarised for the plateau, then repolarises. The action potential duration (APD) is not constant: it depends on how much rest the cell had since the previous beat, the diastolic interval (DI = CL − APDprevious, CL is the pacing cycle length). This dependence, APD(DI), is the restitution curve. At a short DI the gates that carry the plateau have not recovered, so the next APD is short.
The model. The Mitchell–Schaeffer (2003) two-variable cell has a voltage-like variable v (0 to 1) and a gate h (0 to 1):
dv/dt = h v2(1 − v)/τin − v/τout + Jstim + D ∂2v/∂x2
dh/dt = (1 − h)/τopen if v < vgate, −h/τclose otherwise
While v is above vgate the gate h decays as e−t/τclose. The plateau ends when h falls to hmin = 4 τin/τout (the fold of the v-nullcline), which gives a closed form for the restitution curve:
APD(DI) = τclose ln[ (1 − (1 − hend) e−DI/τopen) / hmin ] + δ
Here hend is the gate value at the end of the previous AP and δ is a small constant delay caused by the slow passage of v through the fold. The page measures δ once from a resting beat; the dependence on DI is entirely the analytic curve. The slope of the curve is f′(DI) = (τclose/τopen) u/(1 − u) with u = (1 − hend) e−DI/τopen.
Why alternans. If every beat is paced at the same CL, the APD obeys the map APDn+1 = f(CL − APDn). The steady state is the fixed point APD*. A small deviation ε from it changes the next DI by −ε, and the next APD by −f′ε: a deviation is multiplied by −f′(DI*) each beat. If |f′| < 1 deviations die out (stable), if |f′| > 1 they grow and the cell settles into a long–short–long–short pattern: alternans (Nolasco and Dahlen, 1968; Guevara et al., 1984). The onset CL* solves f′(CL* − APD*) = 1. On the ECG this shows as T-wave alternans, a clinical marker of arrhythmia risk. Below CL* the short beats become too short to follow, and the tissue shows 2:1 block.
The cable. The 120 cells are coupled by diffusion (the term D ∂2v/∂x2), stimulated at the apex end, so a wave travels to the base with conduction velocity roughly proportional to √D. Because the wave reaches the far cells later, their diastolic interval differs from the near cells; with a steep restitution curve this can make the ends alternate out of phase (spatially discordant alternans), a setting in which conduction block and re-entry become likely. The pseudo-ECG is the line-source approximation −∑ (∂v/∂x)(x − xe)/R3 for an electrode above the fibre.
Numerical checks run in the page. (1) The simulated time for h to fall from 1 to hmin equals τclose ln(1/hmin) to 1 %. (2) The resting beat APD equals that value plus the measured delay. (3–5) Premature stimuli (S1–S2 protocol, five beats at 700 ms then one early stimulus) at three coupling intervals: the simulated APD vs the closed-form restitution curve, typically within 0.1 %. (6) The steady APD at CL 500 ms equals the fixed point of the map. (7–8) At CL 320 and 360 ms the alternans is damped; the simulated decay factor per beat of the APD difference equals the map slope |f′|. (9) The simulated alternans onset (found by bisection on the full ODE) agrees with CL* from the closed form. (10) Cable theory: quadrupling D doubles the conduction velocity. (11) The state stays in its physical range (0 ≤ v, h ≤ 1).
Honest simplifications. The Mitchell–Schaeffer cell has two variables and no calcium cycling; real alternans also has a calcium-driven (intracellular) mechanism that this model cannot show. The restitution curve is that of the model, not a measurement. The fibre is one-dimensional with uniform properties; real hearts have fibre orientation, heterogeneity and a three-dimensional geometry. The pseudo-ECG is a qualitative line-source estimate, not a body-surface ECG. Parameter ranges were chosen so the effects are visible, not fitted to a patient. The 3D heart in the companion page only paints this same one-dimensional fibre onto the surface from apex to base.