Dr. Alan Moody says X-rays, MRIs and CT-scans could provide a treasure trove of information if they were systematically studied.
But currently medical image are used simply to diagnose or rule out a problem for an individual, then gather dust in the patient's file.
Moody says if databases of images were amassed, they could be studied in the way researchers currently mine anonymized patient records for clues to drug efficacy or side-effects and other medical questions.
He says part of the beauty of such a system would be the images have already been paid for, so this would be low-cost research.
Moody, who is chairman of the department of medical imaging at the University of Toronto, is making the argument in a commentary published in the journal Nature.
Medicine has long stored preserved tissue samples taken from biopsies for study. Medical images could be used in the same manner, Moody said in an interview.
"Akin to pathology specimens which are sitting there, this is even richer data in a sense. We have whole body scans, we have brain scans, we have fundamental whole organ information," Moody says.
"The power of that comes when you put a lot of that data all in one space and then analyze that data, which is this 'Big data, big picture' idea."
One example of how images could be used in this way relates to the study of dementia. If researchers could look at brain scans of individuals who have experienced early symptoms of what might be Alzheimer's disease and are being followed, the scientists might be able to identify what an early Alzheimer's brain looks like.
"Often we see snapshots of individuals coming with a plethora of different signs which in that small snapshot cannot be put together to give you any larger information. But if you imagine instead of that one patient ... in Toronto alone we'd probably have a thousand of these patients a year," Moody says.
"You would then have this large data bank of brain scans that would then start giving you the bigger picture."
Moody says a system would need to be developed whereby images could be shared in a way that didn't jeopardize individual patient privacy. But these kinds of systems are already in place for the study of other types of data.