Q&A: Some cilia have no cilium length value. Why?
Question: "For some cilia, I obtain no cilium length value, with the # of found skeletons value equal to 0. Why? And how to fix it? Should I include those NA values as 0 length?"
Answer provided on 11th of January 2023 based on a Question in the CiliaQ Discussion forum.
The problem with these cilia is that they are spherical. CiliaQ uses skeletonization of the cilium object to determine the length and because skeletonizing a sphere yields a point, the results has no length value.
Let's first look at why CiliaQ uses skeletonization:
- In CiliaQ, the length of the cilium is determined based on the skeleton of the ciliary "mask". This is valid since cilia have a width that lies at the diffraction limit - thus, there appearance of a cilium in light-based microscopy images appears "blurred", even at a very good "scanning" rate / with a very good camera that allows you to reach pixel widths below the wavelength of light (e.g., 0.05 µm). Consequentially, the actual length of the cilium corresponds only to the central line of a ciliary "appearance"/ segmentation mask:
- Deriving the skeleton and using that as the length measurement has also a second advantage. It makes the analysis less sensitive to the segmentation method selected by a user. For different segmentation methods you may detect wider or narrower ciliary masks (e.g. wide masks with Canny3D since it includes the edges of the appearing cilium, while narrower masks with a harsh threshold algorithm like "RenyiEntropy") - however, the skeleton is the almost the same even when applying it to images generated with different segmentation methods.
- If one would go with the length measurements from begin of a mask to end of a mask the measurement would be very sensitive to different segmentation methods and easily biased by the segmentation method.
- Now if we see a "mask" for a cilium that is like a sphere the skeleton will be a point and thus has no length:
Of course in reality the "spherical" cilium might have a length. Such "spherical" cilia structures in the image may relate to:
- In the case of analyzing a 2D image in CiliaQ (no different z planes, e.g., acquiring an epifluorescent image or only one confocal plane, or analyzing a maximum intensity projection of a stack): the sphere cilium may be only a cross-section through the actual cilium or it may relate (in non-confocal microscopy) to a "standing" cilium (a cilium aligning with the imaging axis = the z axis = the objective).
- In case you analyze a 3D image: It could be that it relates to a cilium at the edge of the stack and thus you see only a fraction of it - or it could be an extremely short cilium (few nm or hundreds of nm long), such as a ciliary "stub" (e.g. emerging from a degrading, shedding of, or forming cilium, which your marker of choice labels only as a "point" that appears as a spherical mask in CiliaQ after segmentation).
In any case, it makes sense to include these cilia into a quantification. To do so, one may either include them with a length of zero or one may just add to all cilia in the experiment (those that give length values and those that give zero length values) a fixed value corresponding to one pixel length or less (e.g., 0.1 µm). It is justifiable since the "spherical" zero-length may actually refer to a ciliary length below the diffraction limit. This addition of a fixed value to all cilia will not bias the data but instead improve the possibility to compare different groups.