MRI shows anatomical changes and thus until there are gross physical changes in the brain you won't be able to detect such disease processes. MRI is wonderful for brain imaging, as it can differentiate between gray and white matter better than any other modalities but given the cost per scan, time required per scan and long queue of higher priority patients (stroke, head trauma, etc) it isn't effective given its low sensitivity.
While I am biased towards Molecular Imaging a lot of focus has been on Pittsburgh compound B for imaging amyloid plaques. This type of imaging has the advantage of being extremely sensitive and specific however the cost and availability are even greater than that of MRI. Positron Emission Tomography (PET) scanners aren't widely available and where they are it is mainly reserved for oncology.
Many theories for Alzheimer's disease exist but given the evidence linked to beta-amyloid mechanisms I don't think the missing link is some unknown mechanism but rather no cost effective way of dealing with it considering preventative treatments exist but are limited and no intervention exists to reverse the effects. That and there are a number of non-Alzheimer's dementias that people are less aware of.