Compute the Cure: How GPU-Driven Cancer Therapies Overtook One Man’s Astronaut Dreams

by Liz Austin

John Neylon thought he’d be an astronaut.

But instead of probing space, he now explores new frontiers in cancer treatments, using adaptive radiation therapy. His work aims to improve patient outcome and quality of life, and he’s doing it using NVIDIA GPUs.

To help him pursue his studies at UCLA’s Physics and Biology in Medicine graduate program, Neylon recently received a Fellowship Award from the NVIDIA Foundation. The award forms part of the Foundation’s Compute the Cure program, an initiative to advance the fight against cancer.

Head and neck model volumetric rendering with critical structures, by UCLA's John Neylon.
UCLA’s John Neylon’s head and neck model volumetric rendering with critical structures.

Neylon is using the money to help fund the development of a framework that uses image registration and predictive biomechanical models.

Accelerating these tasks with GPUs will allow a seamless shift into existing clinical workflows and provide physicians with comprehensive data for optimizing treatments to the patient’s daily anatomy.

Patient Imaging

Treatment plans detail the required radiation dose, such as an external beam with photons, as well as the angle, energy needed and rate of delivery.

While the fundamental physics of this complex treatment are well established, variables such as weight loss or a patient moving on the treatment table can lead to less effective therapy.

“Technologies are evolving to offer more precise treatment that improves the curative rates and decreases the side effects,” Neylon said.

Optimizing treatment plans by using adaptive radiation therapy to address changes can reduce risks. Biomechanically modeling a patient’s anatomy can reveal variances such as posture changes and tumor regression.

The drawback: overlong computation time on a CPU, leaving little flexibility or speed in offering a new treatment plan.

“We had to get our calculation times down to seconds and minutes, from minutes and hours,” said Neylon, who taught himself programming languages and CUDA.

Developing a GPU-based model allowed for more complex algorithms and sophisticated treatment simulations at much faster computational speeds. Also, coding to a cloud-based GPU server would mean faster processing, less cost and ready access to all the calculation tools needed.

GPU-based high-resolution multi-level biomedical skeletal anatomy with critical structures undergoing deformation due to head rotations by UCLA's John Neylon.
GPU-powereed anatomical models undergoing deformation due to head movements, by UCLA’s John Neylon.

Earthbound Ties

Neylon’s journey to medical physics, after dreams of space, was formed due to the most earthbound of ties — family.

He did his undergrad work at Purdue University, drawn by its competitive science programs. And the knowledge that at least 23 of its graduates became astronauts, including Neil Armstrong, the first person to walk on the moon, and Eugene Cernan, the last.

But one summer spent calculating the probabilities of how exchange particles would spontaneously split and recombine drove Neylon to focus “on something more connected to humanity.”

He got his wish, but not in the way he expected.

Soon after, Neylon’s mother was diagnosed with cancer. While helping her navigate surgery and chemotherapy and their devastating side effects, he learned about medical physics and realized his training had prepared him perfectly to enter that field.

Neylon’s mother recovered, and now her son is working on ways to help others do the same.