Nothing involves any real risks to respondents
While potential risks count against approval, no risk doesn't imply no need for approval.
E.g., if I'd want to measure native blood cells (no treatment but immediately dilute with physiological NaCl, place under the microscope, measure, and destroy). As for the amount, a drop from the finger-tip is more than enough.
For this*, the safest possible sample is my own: I cannot infect myself due to the contact with my own blood. Nevertheless, ethics approval does not distinguish who the human in question is, and I'm treated the same as any one else: I need approval.
There are even ethical guidelines that just plain forbid this:
1.5 No one should work with his or her own blood.
Obviously, for all experiments on students & colleagues there is always the question how to ensure that there is no pressure for them to participate. As this pressure could come from superiors, it logically has to include oneself. IMHO this alone is sufficient to require ethical approval for such experiments.
* things would be completely different if the sample were cultured or transformations were involved, etc.: there risks would increase if working with own samples.