CellParser.determine_nseg¶

single_cell_parser.cell_parser.CellParser.determine_nseg(f=100.0, full=False, max_seg_length=None)¶

Determine the number of segments for each section according to the d-lambda rule.

Parameters:¶

f : float, optional¶: frequency used for determining discretization. Default is \(100.0 Hz\).
full : bool, optional¶: Whether or not to use full spatial discretization (one segment per morphology point). Default is False.
max_seg_length : float, optional¶: Maximum segment length. Default is None.

Note

The d-lambda rule predicates the spatial grid on the AC length constant \(\lambda_f\) computed at a frequency \(f\) that is high enough for transmembrane current to be primarily capacitive, yet still within the range of frequencies relevant to neuronal function. Hines and Carnevale [7] suggested that the distance between adjacent nodes should be no larger than a user-specified fraction (“d-lambda”) of \(\lambda_{100}\), the length constant at 100 Hz. This frequency is high enough for signal propagation to be insensitive to shunting by ionic conductances, but it is not unreasonably high because the rise time τr of fast EPSPs and spikes is ~ 1 ms, which corresponds to a bandpass of \(1/\tau \, 2 \, \pi \, r \approx 400 Hz\). At frequencies where \(R_m\) can be ignored, the attenuation of signal amplitude is described by

\[\log \left| \frac{V_0}{V_x} \right| \approx 2 x \sqrt{\frac{\pi f R_a C_m}{d}}\]

So the distance over which an e-fold attenuation occurs is

\[\lambda_f \approx \frac{1}{2} \sqrt{\frac{d}{\pi f R_a C_m}}\]