An algorithmic approach developed by the researchers to mechanically combine diagnoses of medical images from a group of 12 radiologists outperformed 83% of radiologists alone when distinguishing any kind of abnormality from a normal case, said co-author Dan Palmer, PhD, a computer science professor at John Carroll University.
Palmer spent the spring of 2013 on sabbatical at the Cleveland Clinic working on the research with Dr. David Piraino, a radiologist and chief informatics officer for the Cleveland Clinic's Imaging Institute.
Palmer's area of interest is in "swarms," biologically based strategies of creatures that act in large numbers to solve computational problems. Piraino works with Palmer's students as a software engineering client, and Palmer's students developed a project for him that supports collaborative diagnosis of medical images using a social media paradigm.
The exhibit will also show how for a radiological consult, radiologists should present the image(s)/case without any interpretation and get an uninfluenced opinion, rather than presenting a colleague with a diagnosis and asking for confirmation, Palmer said.
"This will provide a wider range of potential diagnoses, increasing the likelihood of finding the correct one," Palmer told AuntMinnie.com.
