Countering is an instrumental process for radiation oncologists and their patients—but the method is highly subjective. Researchers at MD Anderson Cancer Center in Texas created a deep neural network (DNN)-based method to automate the contouring of high-risk clinical target volumes in CT scans that proved as accurate as physicians, resulted in less subjective practices and time savings.
Results were published in the June 2018 issue of the International Journal of Radiation Oncology Biology Physics.
“By implementing a Dice similarity coefficient (DSC)-based threshold selection function, our DNN auto-delineation algorithm accurately identified physician patterns to predict clinically acceptable high-risk clinical target volumes (CTV) contours,” wrote first author Carlos E. Cardenas, with the Department of Radiation Physics at The University of Texas MD Anderson Cancer Center and colleagues. “Our models allowed for the prediction of new volumes within a few minutes and have the potential to greatly reduce physician contouring time.”
Researchers analyzed 52 oropharyngeal cancer patients who had been treated at MD Anderson between January 2006 and August 2010. Each had previously had their gross tumor volumes and clinical tumor volumes contoured for their radiation therapy treatment.
The deep learning model used the gross tumor volume and distance map information from surrounding anatomic structures as inputs. A subset of data was withheld from the training data to test the algorithm.
Results showed their algorithm was comparable to the work of trained oncologists. The predicted contours also closely resembled with the ground truth, according to authors.
Additionally, the DNN method provided clinical target volumes much quicker than if a radiation oncologist were to manually determine them. Traditionally the process could take anywhere from two to four hours, whereas the algorithm, with the help of the Maverick supercomputer at the Texas Advanced Computing Center (TACC), produced CTVs in under a minute.
"If we were to do it on our local GPU [graphics processing unit], it would have taken two months," Cardenas, a graduate research assistant at MD Anderson Cancer Center in Texas said in a May 9 TACC story.
"I think it's going to change our field. Some of these recommender systems are getting to be very good and we're starting to see systems that can make predictions with a higher accuracy than some radiologists can. I hope that the clinical translation of these tools provides physicians with additional information that can lead to better patient treatments," Cardenas added.
Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.