Given fMRI's nature as a secondary measurement of brain activity, I'd have expected this kind of research to focus on EEG, which, as a direct measurement of brain electrical activity, should have much lower latency. I'd think that latency would be pretty crucial for neurofeedback.