Categorical channel in image segmentation


I have an existing Unet network which is trained to find “interesting stuff” in CT brain scans. The brain scans I have are decomposed from a single input channel to 5 input channels. Now I want to add an extra channel containing a brain atlas (outlining the anatomy of the brain). This channel will contain categories, say, 0 to 10, but the fact that 0 < 5 has no meaning.
What is the preferred way to add such a channel? I guess I want to add some kind of embedding of the categorical channel. Does anyone have an example of this?

Kind regards,