How can we improve the relationship between computers and radiologists? How can we make use of dark matter (all the information currently not being mined from our images)? How can we combine data sets to calculate critical conditions like malignancy?
The answer to all of these questions is big data. And if it sounds a bit futuristic, that’s because it is.
Dr. Eliot Siegel of the University of Maryland School of Medicine led a webinar at the end of January, hosted by the Society for Imaging Informatics in Medicine. According to a recent HealthImaging article, Siegel both defined big data and discussed the role it could play in radiology. He explained that the National Institute of Standards and Technology defines big data as data that exceeds the capacity or capability of our current or conventional methods and systems.
Big data’s role in radiology begins with saving every clinical course and making its data available for decision support, rather than only using data from specific clinical trials. The idea is to encourage radiologists to rely more frequently on computer-aided diagnosis (CAD). Siegel noted that although a recent survey found that 89 percent of radiologists say they use CAD, only 2 percent say they often change their original interpretation based on its findings.
Second, 100 percent of radiology data would be saved in EMRs. Siegel explained that untagged, unmined images are the equivalent of dark matter — vast amounts of unusable information. He stressed that this must change if enterprise medical imaging is going to play a substantial role in the era of personalized medicine and computer-based decision support.
Third, the resulting large data sets would be used in CAD systems to study patients with similar characteristics and calculate the likelihood of things like malignancy. “Our goal would be to identify molecular pathways for a cancer rather than simply its diagnosis or appearance in one pathology or histology,” he said.
Finally, Siegel offered the following suggestions for radiologists who want to see this future become reality:
- Look at how you capture your data.
- Consider using formats that are easier for your reports to be mined.
- Consider tagging your studies and making them widely available.