Spine fractures are one of the most frequently encountered emergent findings in the emergency room (ER). While CT spine imaging may be used in the ER to detect a number of disease processes, acute fracture is by far the most important and common finding, according to senior study author Dr. Peter Chang of the University of California, Irvine (UCI).
At UCI -- a level I trauma center located at the intersection of three major freeways -- dozens of patients come to the ER every day for CT spine imaging to rule out an acute fracture, he said.
"The ability of an AI algorithm to quickly evaluate a CT scan for the presence of fracture significantly improves the day-to-day radiologist workflow and, importantly, expedites patient care," Chang said.
The deep-learning approach, based on two residual convolutional neural networks, yielded near human accuracy in detecting spine fractures on CT, according to the researchers.
"While it will still be many years before a computer can independently interpret diagnostic images, there are a handful of key, focused tools like this that will tremendously benefit the day-to-day practice of radiology in the near future," Chang told AuntMinnie.com.
The deep-learning technique is also flexible, he noted. Its ability to identify and separate out each individual vertebral body can be broadly applied to a number of related spine applications, such as assessing spine alignment or detecting spine metastasis. Both applications are currently being developed at UCI's Center for Artificial Intelligence in Diagnostic Medicine.
Want to learn more? Sit in on this Tuesday morning talk to hear all about their AI initiative.
