Right, it'll most likely be low frequency, near field inductive. The body is plenty conductive enough to short out electric fields (hence your finger looks very much like a lossy capacitor say at 100MHz), but is still pretty transparent to magnetic fields. Typical skin depth at 100MHz is 6cm -- even in the very center of the body, you can probably expect a modest say 20dB loss, and obviously, less at lower frequencies.
High frequencies aren't even that bad, as the high dielectric constant makes antennas a lot more feasible! The waves don't radiate far, but if your link gain works out alright, eh, there you go.
Indeed, because of the large κ, internal fields don't radiate much at all into free space -- total internal reflection hits hard, and most radiation is quickly absorbed. The preferred approach is patch antennas stuck to the body, also scaled appropriately for the medium.
I wonder if they use antenna diversity to improve reception in all orientations; if the camera is a simple dipole (solenoid), it has a null plane, which will be crossed sooner or later as it waggles its way through the intestine.
Tim