Eccentricity versus Form Factor in MeasureObjectSizeShape

I’m analyzing images of Drosophila cells and successfully gotten to the point where cp is identifying and measuring the nuclei. The measurements that I’m looking closely at (and will be doing statistics on later) are form factor (FF) and eccentricity (Ecc). The reason being that our lab is studying certain genes that cause deformaties of the nucleus shape. The question is: what is a better measurement of deformed shape, Ecc or FF?

I’m leaning toward eccentricity at this point because it “looks” to be more closely matching reality. To explain, I’ve attached two images that have measurements labeled on the nuclei (the file name denotes the measurement). For the one with Ecc labels, the leftmost nucleus has a low .22 value versus the nucleus to the right with a value of .79 (for reference, 0 would be a perfect circle and 1 is a straight line). For the one with the FF labels, the leftmost nucleus has a high .66 value while the nucleus immediately to the right of it has a value of .71 (for reference, 1 is a perfect circle and 0 is a squiggly mess). To me, it doesn’t make sense that leftmost nucleus can be close to a perfect circle on its eccentricity measurement but far away from a perfect circle with its form factor. Perhaps I’m not fully understanding what form factor actually means but I’d like to know if anyone else has done something similar to this? And did you end up using Ecc or FF as a measurement of deformed shape? Or some combination of the two?

Thanks.



Hi Scott,

A couple of thoughts:

  • Re: the leftmost nucleus - The fact that this apparently-round nucleus has an FF of 0.66 seems suspicious. Have you confirmed the segmentation of this object, to check that it is indeed round? Even though this same nucleus has a ECC of 0.22, the eccentricity is computed from a fitted ellipse, so it is conceivable that you could have a disparity between the ECC and FF nonetheless.
  • If you have negative controls (i.e., undeformed) cells and positive controls (i.e, deformed), you might want to consider performing a Z’ factor test
    on the FF and ECC measures. Whichever measure gives you the higher score is probably the one you should proceed with.

Regards,
-Mark

Thank you. I will try that.