For your situation, I might consider not using annotations and using 3D objects. I would create 3D disks or circular planes with textures for numbering. I would then constrain the disks to always face the camera.
Then you can use the onclick even of the discs to do whatever you need.
Alternatively, you could create HTML elements and bind them to your empties….Then you could send an HTML onclick event in the parent window to the V3D scene to take some action.
I am sure there are some other ways to do this but these are just two simple methods.