That's what the treatment of chronic pain boils down to: you have to send feedback - be it visual, tactile, auditory, ... to the brain so it can update the VR model. Only if the model and the corresponding motor command generated by it and the information from the external sensors matches up will the pain stop - very rapidly as several studies have shown.
Now - the really great thing about any type of feedback therapy is that the corresponding map which was altered in a maladaptive way starts to get reorganized while doing the therapy. That's what the studies report: during the first few training sessions the pain diminishes very rapidly - but comes back after a short amount of time (15 minutes,...).
But after the 5th or 6th session most people are able to get more lasting pain relief - and after just 3 weeks of daily practice most people can stop doing mirror box therapy altogether because the pain doesn't return.
This is all due to the positive effects of brain plasticity: the brain is able to re-organize very very quickly - provided it has the right incentive to do so.
As yet there are no long terms studies out that show how people with phantom limb pain that were treated with mirror box therapy have done over the years - this method is simply to new. I expect the first big trials to show up in the next 5 years or so in the wake of the studies done now at Walter Reed Medical Hospital with Veterans of the Iraq War.
That should give us an even better picture about what causes the process of maladaption in the first place.
My best guess is that initial pain (a painful limb that hurts before amputation) is the single most important factor - the brain learns the pain.
That's why amputations done where the limb was anaesthesized first show much better outcomes than traumatic amputations.
If you look at the case of CRPS - and to some extent after a stroke - you see another interesting thing:
the immobilization phase you have to go through after breaking a bone or suffering from a stroke tells your brain that - despite it sending motor commands to your limb - there is no proprioceptive feedback from the sensors in that limb.
If you can't move your hand because it's in a cast it can't send enough feedback to the brain.
So gradually over the weeks the hand map is adjusted - resulting in a limb that actually can't be moved even after the cast is removed. The brain has un-learned that the hand can move a great deal. And since the VR simulation in our brains is the basis for our actual perceptions and thoughts and feelings and everything else a lot of people describe their afflicted hand as no longer belonging to them.
An even more extreme example with different causes is the so called Alien Hand syndrome in which people don't recognize the hand being theirs at all.
What happens with the hand during CRPS - the swelling, osteoporosis and all the other stuff is secondary to the changes in the brain!
By using a mirror to give the brain visual feedback of an intact, healthy and freely movable limb re-organizes the map and symptoms vanish very quickly.
There are some preliminary results that show that even in stroke - at least part of the paralysis that results is not "real" paralysis - but learned paralysis - the brain just thinks that one side can't move any longer.
And even if that only accounts for say 20% - every bit of function you are able to regain after a stroke is better than nothing.
What sets CRPS apart from other chronic pain syndromes is that in addition to the hand map becoming distorted is that people aren't able to distinguish between a picture of a left hand and a right hand. They loose the sense of laterality.
This could be due to the fact that the brain uses the representations of the hands in it's internal model to construct laterality in the first place.
This would amount to the loosing of one of the models of the physical world that was established during childhood. Imagine yourself loosing your sense of gravity - one could no longer throw things accurately because you wouldn't be able to plan a parabolic trajectory - you would think that things always just fly in straight lines.
The NOIgroup Institute has developed a set of cards and a computer program with which you can re-learn this basic first step called Recognize.
If CRPS is in full swing however and sort of movement might aggravate pain.
That's why in severe cases you have to wait a little before you can do visual feedback therapy (mirror box).
The treatment in these cases starts by doing imaginary movements.
Those activate the same brain areas as actual movements - after all the brain does the same while running the VR program - the only thing that's missing is the actual motor command which is actively inhibited.
That way you can train the neuronal/mental circuitry which later will be activated for real.
As for areas of the body that have no limb like features - like the back:
you could have the patient lie down and point a video camera at his back and show him the image on a monitor.
But - as a few clever scientists have found out long ago - visual and tactile feedback are treated as equal in the brain.
Here finally the hands of the therapist come into play. ;-)
By touching the skin over the area you want the patient to move you can facilitate the whole process.
Once he has established motor control over that area again he is instructed to activate the muscles in that area every day so that the brain receives proprioceptive information from that area on a constant basis.
I advise some variation of primate grooming for home use: back rubs, brushes, whatever is available.
Regardless of which type of condition you have or want to treat:
the basis is always to send feedback about a limb or another part of the body to the brain (the comparator).
Feedback in daily life is important: remember how hard you had to hit the keys on an old typewriter?
Today's keyboards - especially the new one's from Apple are awesome!
Touch screens however are tricky because they don't provide any feedback. If the manufacturers were clever they'd introduce a very slight vibration - that would do the trick.