Researchers from The University of Texas MD Anderson Cancer Institute in Houston, Texas developed a deep learning process that automatically identifies the precise size and location of tumors and determines how much radiation should be administered.
The process, known as contouring, is a time-intensive approach that varies from doctor to doctor. The problem, in the case of head and neck cancer, is that tumors sit in a very vulnerable region. If a doctor administers too much radiotherapy, some of the natural tissue may be compromised, too little, and cancer may continue to grow. To help prevent bias and minimize the inter-physician variability, the researchers developed an automated solution that leverages the power of AI.
“If we think about the problem in a smart way, we can replicate the patterns that our physicians are using to treat specific types of tumors,” said Carlos Cardenas, a graduate research assistant and Ph.D. candidate at The University of Texas MD Anderson Cancer Center.
Using NVIDIA Tesla GPUs on the Maverick supercomputer at the Texas Advanced Computing Center (TACC), and the cuDNN-accelerated TensorFlow deep learning framework, the team trained their system on images from 52 cancer patients.
The study focused on patients diagnosed with oropharyngeal cancer, cancer that forms in tissues of the throat.
“I think [deep learning] is 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 higher accuracy than some radiologists can,” Cardenas said. “ I hope that the clinical translation of these tools provides physicians with additional information that can lead to better patient treatments.”
The research was supported by grants from the National Cancer Institute and the National Institutes of Health.