The potential for big data in radiology is limitless—but the industry has a good amount of work ahead if it’s going to capitalize on potential, according to a webinar hosted Jan. 29 by the Society for Imaging Informatics in Medicine.
Led by Eliot Siegel, MD, with the University of Maryland School of Medicine, the event focused on defining what exactly big data is and the role it can play in radiology today.
“There are lots of meanings out there,” Siegel said. “But to the National Institute of Standards and Technology, big data is that which exceeds the capacity or capability of our current or conventional methods and systems.”
As Siegel explained it, big data could potentially play an important role in defining the way radiologists use clinical decision support systems to assist them in reading images.
Citing a recent survey circulated, nearly 89 percent of radiologists said they always use the clinical decision support software computer-aided diagnosis (CAD).
“What was eye-opening to us, however, was that when we asked how often CAD influenced a radiologists decisions, only 2 percent said they often change their interpretation based on CAD,” Siegel said. “Almost 62 percent said they rarely or never change their interpretations based on CAD.”
Siegel said, in light of the numbers presented, there’s an extraordinary need better utilize the computer-radiologist relationship—and he believes big data can play a valuable role.
To illustrate his point, he dubbed everything in a radiology image that does not go into a specific report “dark matter.”
“There is so much in our images that we just aren’t aware of because the images are untagged and not mineable. All of that information we’re not using is like universal dark matter—it’s vast,” he said. “This must change if medical imaging is going to play a substantial role in this era of big data, medical guidelines, decision support and personalized medicine.”
Siegel said the future of radiology lies in making radiological data machine accessible and intelligible and he used oncology as an example.
“Our goal would be to identify molecular pathways for a cancer rather than simply its diagnosis or appearance in one pathology or histology. Rather than relying on small studies, each clinical course and data would be saved and made available for decision support rather than just 2 or 3 percent of oncology patients that are currently enrolled in clinical trials.”
For Siegel, the ideal would be to capture the data of 100 percent of patients in electronic medical records and to capture their radiology data, as well. These large data sets could be used in the future in clinical decision support systems like CAD to study patients with similar characteristics.
“Bottom line, computers are much better at calculating likelihoods of things like malignancy and can use huge datasets to reach their conclusions,” he said.
Siegel offered organizations simple steps to follow going forward in an effort to get ready for the new wave of big data and how it relates to radiology.
“The first thing you want to do is to look at the ways you capture your own data. How do you mine your own radiology reports and how can you make it easier to put it in a format that you can mine them in the future?” he asked. “Radiologists are now being asked to do a little extra coding in our department and as they do that, cases are easier to organize and search through. It starts with individual organizations tagging their own studies and making them available.”