Everyone has fMRI research going on. You don't need top of the line equipment that costs big bucks and you don't need a good understanding of MR physics to do it. So it's in vogue right now. Neurologists and Psychiatrists who have few other ways to study the brain can create a paradigm, press a button on the MR scanner, and input the results into SPM for VOILA! Results. Send to Human Brain Mapping (IF 7), rinse, repeat. Reproducibility may be just about zero, but eh, nobody can reproduce your paradigm *exactly* can they?
I did my PhD in Biophysics and I'm a more nuts and bolts MR engineering type. So the above paragraph just reflects the bias you'll hear from us. But that's ok. I do think if you have experience in this field as an undergrad you can easily make arguments for a neuroimaging PhD at many programs. My advice is to look at it more from an engineering perspective so you really understand the fundamentals of MR if you're going to do BOLD/ASL/VASO/etc... You can still get your PhD in Neuroscience or Biophysics (and I recommend this if you weren't an engineering undergrad), but make sure to take the hardcore MR courses. Baylor isn't the first place that comes to my mind as being strong in these things. It's more like Minnesota (BOLD fMRI birthplace), MGH, Penn (ASL birthplace, though the real brains behind it are at MGH), and I can give more recommendations if needed...