There is evidence that MRI can be used to detect biological differences and that there is some genetic component as well. The two major challenges though are that these diseases are not "binary" but multivariate and that it's very difficult to collect enough data to explore that space. So it's certainly hard, but not impossible!

http://www.sciencedirect.com/science/article/pii/S0920996401...

http://www.nature.com/ng/journal/v43/n10/full/ng.940.html