Ilastik : Object Classification Table

Hi everyone,

I’m sorry in advance if this kind of questions have already been answered.

I’m a PhD student working on particles analysis in 3D, I use the FIB-SEM 3D technique to analyse nanocomposites. In this sample, I have one type of spherical particles of different size, and I need to analyse size distribution, coordination etc…

I’m working with Ilastik to segment my particles with Pixel Classification then I do Object Classification to do some measurements: Size in pixels, Radii of the Object, Object Centre. I need some precisions on the measurement made.

Is the size in pixels the measurement of the projected diameter of the object?
Is the “Radii of the object_0”, “Radii of the object_1”, “Radii of the object_2” correspond to the X, Y, Z directions of the radii for the ellipse?

Best regards,
Manon

Hi @MgTomo3D,

welcome to the image.sc community!

  • Size in pixels: Total size of the object in pixels. No correction for anisotropic resolution or anything else.
  • Center of the object: Average of the coordinates of this object’s pixels.
  • Radii of the object: Eigenvalues of the PCA on the coordinates of the object’s pixels. Very roughly, this corresponds to the radii of an ellipse fit to the object. The radii are ordered, with the largest value as first. So please first bear in mind that this is not an ellipse fit, but a principle component analysis of the object voxel/pixel coordinates. The radii are then to be interpreted in the direction of the principle components, which are 3 vectors with 3 values each, ordered descending from the one that explains the most variance of the data.

@Sith pointed out that they use the result of object classification in Fiji to compute ellipse fits using MorphoLibJ: Object size underestimated by the Radii - #3 by Sith

Cheers!

Sorry for reopening this thread, I am doing something very similar. May I clarify regarding this:

  • Size in pixels: Total size of the object in pixels. No correction for anisotropic resolution or anything else.

For 3D pixel classification and tracking, does this pixel size refer to the number of voxels?

Hi @zhoutianxun,

welcome to the image.sc community! And great that you found this thread :slight_smile:

yes!

Thank you so much!

May I clarify one more thing if you don’t mind :sweat_smile:

For the principle components that I get when I export tracks to csv, I am correct to assume that PC0, PC1 & PC2 make up the first principle component. PC3, 4, 5 the second and PC6, 7, 8 the third? And are they representing x, y, z in that order?

Principal_components_of_the_object_0 Principal_components_of_the_object_1 Principal_components_of_the_object_2 Principal_components_of_the_object_3 Principal_components_of_the_object_4 Principal_components_of_the_object_5 Principal_components_of_the_object_6 Principal_components_of_the_object_7 Principal_components_of_the_object_8
-0.976181 -0.003298 0.216934 -0.189131 0.502859 -0.843423 0.106306 0.864362 0.491505

yes exactly - we should really document this properly somewhere…