Just a quick note while I take a break from our day long project meeting (I've signed enough papers that I would think twice before posting anything from here!). A nice little article in Technology Review on visual prosthesis can be found here.
There are some issues I could see with LGN based recording systems from just looking at it. First, you're limited to the number of electrodes you insert because you have a huge distance to travel. Second, implanting in humans will be a much more risky operation than current cortical techniques. Third, explant and damage to the brain could be significant. Fourth, the electrodes would become more encapsulated (gliosis, etc) due to a larger surface area, which equals greater irritation. Still, interesting to think about. Oh, and most of the projections to the LGN are from V1, so you would have to be pretty keen at sorting out what units are being driven by.