Studies are already being done to see if an MRI can detect lies any better than the traditional polygraph (which doesn't do a very good job of it). These studies are all testing the reactions to spoken questions though. An easier way, based on the latest findings, would be to ask nothing and just watch their reactions to pictures. The logic behind this approach is based on a recent study where people without any background in physics were shown a picture of two balls falling to the ground, a very large one and a small one. When the big ball appeared to be falling faster than the smaller one, the error-correcting portion of the brain didn't light up (contrary to reality, this was what the test subjects expected to happen). When a picture was shown with the balls falling at the same rate, the error-correction neurons fired up because the subjects thought something was amiss. People familiar with physics didn't react the same way.
So here's how it would work using crime scenes. Suspects would be shown pictures of the crime scene but a single detail in each of them would be radically altered (the blond turned into a redhead for example). Only that single detail would be different in the pictures. The error-correcting neurons wouldn't fire up in the brains of the innocent because they wouldn't have any idea of what the crime scene looked like to be begin with. The criminal on the other hand would realize something was wrong and his error-correcting neurons would fire up to resolve the discrepancy. It would be easy to find the top suspect this way without asking a single question.
The following is an excerpt from an article in a magazine called New Scientist that gave rise to this idea.
Why is physics so difficult?
Kevin Dunbar, a cognitive scientist at Dartmouth University in Hanover, New Hampshire, and his colleagues scanned the brains of students while they watched a video demonstrating either classical Newtonian physics, in which a large and a small ball fall to the ground at the same speed, or the naive scenario, in which the larger ball drops faster.
Those who had never studied physics showed activity in a part of the brain associated with error processing when they watched the Newtonian model, implying they thought there was something wrong with what they were watching. But the naive model sparked activity in the medial prefrontal cortex, normally active when someone thinks about a theory accepted as correct. Students of physics showed the opposite patterns, though even they had some prefrontal activity when watching the naive model, indicating they were still attached to this false but intuitive notion.
Substitute the words 'guilty' and the 'innocent' in place of non-physics and physics students and you've got this idea.