active contours 3D measurements

[Jingjing]

Dear Alexandre and Stephane, 

It might be a stupid question because computer never lies, but I'm quite confused about some results generated by active contours and ROI measurements so decided to double check with you. 

I have been using active contours to segment cells in an epithelial tissue with cell boundary staining and the segmentation itself works quite nicely. However when extracting the 1st axis of the fitted ellipsoid I find that although (by eye) cells stand quite perpendicular to the xy plane (small vx, vy and big vz for the axis vector), very often in the ROI statistics they have big y (x is mostly small, though). Also some apparently elongated objects (by 1st and 2nd diameters and by eye) show a sphericity of 100 meaning they are quite round. I checked this result by reslicing the z stack orthogonally and again more object vectors seem to have small xs and bigger ys and zs (which shouldn't happen?). 

As an example I uploaded the images and segmented rois (active contours seeds are also included) to the Dropbox (they are over 20MB) with the link https://www.dropbox.com/sh/132w2uhjbsqsinm/AACL-JZU3iXcds31lZ33SeEea?dl=0

I know it's such a specific question but could you kindly take a look at the data and maybe spot anything I did incorrectly? 

Thanks very much!

Jingjing