It seems to me that retrospective analysis of data such as symptoms, diagnoses and images poses zero risk to subjects. Has anyone developed standard operating procedures to minimize informed consent requirements for such research? I would like to be able to pull images for two groups and compare them as a way to introduce students to the techniques of various neuroimaging research (e.g., VBM). Is there a way to avoid getting informed consent from all such subjects, since there is "minimal risk," or burden?

Hi Stephen
You should  contact the IRB at your institution for recommendations.