Disclaimer: I'm a graduate student in medical imaging -- but not on the hardware side of things, so my remarks in these areas are likely to be incorrect.

I'm sure some are reading this hoping for applications to uploading (as the comparison to neurons is compelling), but IMHO (emphasis on humble) I don't expect MRI to be the path to nondestructive human brain imaging.

Problems:

1. Increasing the strength of the magnetic field is an easy way to increase resolution, but there are hard limits as to how much magnetic field a human brain can take. Already at 2 Tesla an MRI can heat and sometimes burn patients; the risk at 11.75 Tesla (as in the article) is proportionally greater.

2. Not all brain-related information that may be necessary for reconstruction is visible to an MRI, e.g., some kinds of neurotransmitters.

3. MRI resolution can be high, but contrast can be low -- particularly for those applications to sodium and potassium nuclear resonance the article briefly mentions. You can improve contrast by injecting contrast agents, but this also comes with safety risks, and the blood-brain barrier makes it difficult to chose good contrast agents.

4. Increased resolution implies increased acquisition time, and increases in acquisition time imply increased artifacts from motion blurring. In the early days of MRI they used to sedate or briefly kill patients to obtain decent imagery, so this problem is not insurmountable.

For a while, at least, destructive imaging is going to be my best bet for uploading.

EDIT: Removed some confused remarks in 4.