Until recently, radiologists would receive patient information directly from the referring physician during image review and consultation. Although clinical information can now be shared and exchanged easily and remotely across an enterprise, this has led to a siloed effect, requiring radiologists to collect pieces of a patient's history from various places such as the electronic medical record (EMR), said presenter Merlijn Sevenster, PhD, from Philips.
"This new process is time-consuming and can disrupt workflow," Sevenster told AuntMinnie.com.
To make it easier for radiologists to access patient information, Philips applied a natural language processing algorithm to extract key patient information from a set of radiology reports provided by the study's principal investigator, Dr. Paul Chang of the University of Chicago. Testing showed that the method was nearly 95% accurate for retrieving a pertinent condition included in a radiology report, Sevenster said.
"The study also identified disparities between the radiology report and EMR when it came to including information about a patient's condition (e.g., HIV), indicating that a radiologist may not have all the clinical details needed at the time of image review," Sevenster said. "Our findings suggest and reinforce the need for automated workflow solutions that aggregate and synthesize patient information in real-time from multiple data sources, including EMRs and radiology reports."
