Center tube-like samples during 2D classification?

2D Classification
1

Hello All,

I am trying to center tube-like samples during 2D classification. The issue I face constantly is that the classes slowly drift from decently centered helices to only converging on one “strand” of the filament (see image. The image is a mix of different 2D classifications, but the general trend is always the same).

image
image1226×266 41 KB

I wonder if there are any settings that I could try out. I have been trying to play around with:

  • Increasing the size of the particle extraction box size
  • Window inner & outer radius values
  • Enabling “Align Filament Classes Vertically”
  • Disabling “Use Circular mask on 2D classes”
  • Disabling “Re-Center 2D classes”
  • Enabling “Min over Scale after first iteration”

Do you have any other tips that could help centering these tubes? Going 3D is difficult without helical parameters, and I need good 2D classes to find twist/rise.

Thanks in advance!
Tom

2

Try switching on “Recenter mask binary”, then tweak the mask threshold based on the output masks (which will be displayed below the class averages during classification).

3

Dear Oli,
thanks for your quick response! I have played around a lot with binary mask thresholds and a value of 0.037 seems to get the best results for this sample.
Nevertheless the 2D classes still converge on one strand of the helix. Is this preventable at all? I am trying to re-center the tubes to be able to sort out different diameters via 2D classification.

image
image547×460 54.6 KB

Thanks!