In a small study from Helen Mayberg and colleagues at Emory University, depressed participants were randomized to either a medication or CBT, and underwent PET scans before treatment.

Then the scans were analyzed to see if they were predictive of treatment response. The goal was to find a biomarker in order to determine in advance who will respond better to which type of treatment.

However, as The Neurocritic points out, often people respond to a combination of both psychotherapy and meds, and writes an excellent analysis of the study and where it needs more work. Look for lively debate from scientists in the comments.

Also check out a great write up from David Dobbs at Neuron Culture.