espenhgn / icsd Goto Github PK

In the current implementation inverse methods will fail if contacts are not equally spaced. Furthermore, constant radii (and thickness) of sources are implicitly assumed

Need unittest for spline iCSD method, and tests for failiures e.g., if input is given in different units than m, V, S/m etc.

The link to the original matlab version appears to be broken - do you know whether it is still hosted anywhere? Thanks -

Hi, I'm Marta from Daniel's lab at Nencki. We're trying to compare iCSD methods to kCSD. I found a discrepancy in your Spline iCSD implementation. Could you please verify?

In file icsd.py at line 311, 426 and 474:

z_js = pl.zeros(el_len+2)
z_js[1:-1] = self.coord_electrode
z_js[-1] = z_js[-2] + pl.diff(self.coord_electrode).mean()

Here, I suspect that there must be the following line:
z_js[0] = z_js[1] - pl.diff(self.coord_electrode).mean()

Similarly in, line 386:
coord_ext[0] = 0
should be

coord_ext[0] =  self.coord_electrode[0] - \
            pl.diff(self.coord_electrode).mean()

Thanks!

It seems that all CSD estimation methods assume equal spacing of recording depths and hence what they actually need for calculation is the interval between the depths (i.e. local variable h_val).
Then, isn't it simplar to give as an argument only the interval, rather than the coordinates of all the recording depths?
I also spotted a bug in SplineiCSD() related to this point, i.e., if you give coord_electrode so that the first element of this array is not just below zero (e.g, coord_electrode = (1000E-6, 1100E-6, ...)), the class returns a wrong result (and I cannot manage to fix this bug).