Hi Barz,
We've never tried this, but I would think the easiest thing to do would be to use an 'effective density' and 'effective sound speed' in the regions where the US contrast agent is. In other words define a region where the density and sound speed are both lower than in tissue. With the current release of k-Wave the tissue properties are not time-dependent, so it would not be possible to simulate the flow-through of a contrast agent, but that might in future be possible within this sort of pseudospectral framework.
Modelling the bubbles or gas-filled microspheres directly, and especially modelling the dynamics of a cloud of them, is quite a complex task in itself (with a large literature) and that is not currently incorporated into k-Wave.
Hope that helps,
Ben