Comment ERP (Score 3, Interesting) 552
First, recognize the need for empirical information on the state of your loved-on. It is of very little use to make subjective observations, since humans are incredibly good at finding patterns where none exist.
Second, recognize the difficulty of what you're undertaking. Humans are at the very beginning of understanding how our bodies work, and we have essentially no model to predict when patients will, or never will, recover from injury like this. What makes it hard is that this ignorance means that you will be trying to make decisions under extreme uncertainty - but that doesn't mean you shouldn't do so. For instance, there should probably be a time past which you withdraw life support when there are no signs of recovery. No one knows how long that should be, but the key thing is whether there are signs of hope.
What would be such signs? You've already read something about the locked-in phenomenon. First, CT cannot possibly provide any information about function: it measures x-ray density, and provides only structural information. At best, it might show which tissue has died - but unfortunately, we have very primitive knowledge of how that relates to function (or recovery). ERP (scalp electrodes) are MUCH more relevant: there is a huge literature describing the sorts of obligate responses made by sensory portions of the brain (our understanding of less sensory processes is rather spotty). PET can map metabolic activity, but that has a much less obvious relation to organized, functional brain activity. I think ERP monitoring should be your primary path forward. There is lots of research on this topic, and pretty much any university psychology/neuroscience/psychiatry department would have well-informed people you could talk to, often ones able to perform ERP tests for brain function. (The technology of ERP is very not hard, and designing effective tests is somewhat subtle. But if a test is supposed to guide a decision like continuation of life-support, it's not a casual trip-to-radioshack kind of project.)
In short, find a non-self-deluding way to gather empirical signs of functioning personhood; in the absence of such signs, figure out how long to wait.