Change in dendritic diameter tracing across acquisition periods in Cell Types reconstructions?

Hello,

While fitting passive cable models to Cell Types Database cells (human and mouse, ~580 reconstructed cells with Short Square responses), I noticed what looks like a systematic change over time in how dendritic diameters were traced. I’d be grateful if someone from the morphology team could confirm whether this reflects a known change in the reconstruction pipeline.

What I see in the SWC files

  • Two diameter conventions coexist. In some reconstructions no dendritic node has a diameter below 0.454 µm , with a large fraction of dendritic length sitting exactly at that value. In others, dendritic diameters go down to ~0.23 µm.
  • The convention tracks specimen ID, i.e. acquisition order. In mouse cells the floored convention accounts for 0% of reconstructions with specimen IDs below 5.0 × 10⁸, ~8% between 5.0 and 5.5 × 10⁸, and ~81% above.
  • Dendrites are traced thicker overall in later cells, not only at the floor. The median length-weighted mean dendritic diameter is ~0.49 vs 0.58 µm in mouse and ~0.59 vs 0.68 µm in human (unfloored vs floored).
  • Within a cell type, the electrophysiology doesn’t change. For example, Pvalb interneurons reconstructed early (n = 37) vs late (n = 17) have the same τ/Rin capacitance and the same traced dendritic length, but traced diameters differ by a factor of ~1.5, and so does reconstructed membrane area.
  • Human cells were almost all reconstructed in the later period, so the change is partly confounded with species.

This matters for anything normalized by membrane area: fitted specific capacitance and axial resistivity, conductance densities, surface-area features.

Questions

  1. Was there a change in the tracing or radius-estimation procedure (tool version, automatic radius estimation, default minimum radius, imaging or processing) around these specimen IDs?
  2. Is the procedure used for each reconstruction recorded anywhere (metadata, SWC header, API field) so users can group cells by protocol?
  3. Is there a recommended way to harmonize diameters across periods, or a view on which convention is closer to the true dendritic diameter?

I’m happy to share the analysis code and the per-cell list of conv

entions and specimen IDs if that would help.

Thank you,
Tanguy Damart