Clinicians commonly call radiologists to ask about CT intravenous contrast assignment, according to presenter Mian Zhong from the University of California, San Francisco's Big Data in Radiology (BDRAD) research group.
"Even though it may be straightforward to radiologists, most referring clinicians are not necessarily trained on when to assign CT contrast or not," Zhong told AuntMinnie.com. "An automated program that can assign intravenous contrast can reduce these phone calls and also improve our CT protocoling efficiency. We now have big radiology report text data available and also have the natural language processing (NLP) technology necessary to analyze the free-text clinical indications."
The software could be deployed as a decision-support system for referring clinicians, yielding an immediate benefit of fewer unnecessary phone calls from clinicians asking whether CT contrast should be assigned, said senior author Dr. Jae Ho Sohn. By preassigning CT protocols, the software could also enhance workflow efficiency for radiologists.
"In conjunction with our already published automated MR contrast assignment algorithm, we are using these studies as the stepping stones for a more general intelligent CT and MR protocoling system that can take in the free-text clinical indication, requested imaging modality, and body part and then automatically preassign the appropriate protocol for final review by the radiologist," Sohn said. "So far, we have been impressed by our algorithm's speed and accuracy in clinical data, but, as with any machine-learning model, it needs external and prospective validation before clinical adaptation."
Stick around for this late Thursday morning talk to get all the details.
