MD Anderson leverages deep learning for radiation targeting tool
MD Anderson Cancer Center researchers have successfully used deep learning and a supercomputer at the Texas Advanced Computing Center to develop and test a treatment target identification tool for head and neck cancers.
Before performing radiation therapy, oncologists review gross tumor volume using medical imaging and then—in a process called contouring—establish the dose of radiation a patient will receive and the method of delivery. The problem is that there can be wide variability in how physicians contour the same patient’s computed tomography scan.
However, Carlos Cardenas, a graduate research assistant and PhD candidate, and a team of researchers at MD Anderson, developed a new method for automating the contouring of high-risk clinical target volumes using artificial intelligence and deep neural networks.
Specifically, the deep learning algorithm Cardenas developed uses auto-encoders—a form of neural networks that can learn how to represent datasets—to identify physician contouring patterns.
In a study, published recently in the International Journal of Radiation Oncology*Biology*Physics, he and the research team analyzed data from 52 oropharyngeal cancer patients who had been treated at MD Anderson and had their gross tumor volumes and clinical tumor volumes contoured for their radiation therapy treatment. What they found was that their results were comparable to the work of trained oncologists.
“Automating and standardizing the contouring of clinical target volumes (CTVs) can reduce interphysician variability, which is one of the largest sources of uncertainty in head and neck radiation therapy,” concludes the study’s authors. “These predicted high-risk CTVs provided close agreement to the ground-truth compared with current interobserver variability. The predicted contours could be implemented clinically, with only minor or no changes.”
In addition, by leveraging the Maverick supercomputer at the Texas Advanced Computing Center, researchers were able to produce CTVs in less than minute. By comparison, it takes a radiation oncologist two to four hours to complete the same task.
“I think it's going to change our field,” says Cardenas. “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.”