Problem with stability of Courtemanche model

[robbedaelman-crypto]

I am experiencing problems with the Courtemanche model after recent updates, the last working version (at least for me) is de1f52e, dating from March 9th.

When previously running simulations with the Courtemanche model with dr = 0.3 and dt = 0.025, I did not experience any problems. However, after the update, the ionic kernel throws a divide by zero error. Reducing dt to 0.020 solves this, but nowthe recorded EGM signals have a completely wrong form and in videos of the simulation, no stimuli seems to be propagating. Reducing dt any further (down to 0.001) does not help to relieve the problem. I think this is a stability issue, but I'm not completely sure.

[TiNezlobinsky]

Hi Robbe! Thanks for the report.

The issue was caused by an incorrect diffusion coefficient in the updated Courtemanche setup: it was set to 1.0 instead of the intended 0.154. With dr = 0.3 and dt = 0.025, this pushes the explicit scheme into an unstable regime, which explains both the divide by zero error and the abnormal EGM behaviour.

The diffusion coefficient can be changed simply as model.D = value, but I've already corrected the values, so you can just update the package.

I'll leave this issue open, as it's a reminder that in addition to cell-level testing as we currently do, we also need tissue-level tests to catch this kind of problem earlier.