IBM analytics solving cancer
The new technique uses a high-energy particle accelerator — a room-sized version of the gigantic accelerators used to unravel physics — that directs a proton beam to precisely kill the cancer cells inside tumors, leaving adjacent tissue untouched. IBM Research (Austin, Texas) is aiming to reduce the cost of this promising new therapy, using software analytics running on a Power7 cluster supercomputer.
The stakes are huge. Last year there were over 12 million cancer patients worldwide receiving various therapies, and that number is predicted to increase to 21 million by 2030. Unfortunately, today there are only 10 centers offering proton therapy, however 17 new ones are currently under construction at a cost of over $200 million each. Over $3.4 billion is being invested in current-generation proton accelerators, and smaller, less expensive accelerators are also being designed to make the technique more affordable.
The big bottleneck, however, is the computational workload required to utilize proton therapy.
Today proton therapy requires a long involved preparation process, starting with a magnetic resonance imaging (MRI) or computational tomography (CT) scan to identify the tumor’s location, after which a bevy of doctors and technicians spend over a week mapping out exactly how to use a proton beam to destroy it. Unlike traditional radiation therapy, proton beams do not affect human tissue as they pass through it, only releasing their energy at the very end of the path, which must be carefully plotted to end precisely within the tumor.
Today it takes a team of doctors and technicians a week to map out the path for a proton beam to kill tumors, but in the meantime it grew, making success less likely. IBM’s analytics running on a Power 730 cluster computer maps out the same proton beam path in 15 minutes. SOURCE: IBM
"The protons are accelerated to half the speed of light, but do not loose that energy until they hit a threshold, after which they release burst of kinetic energy," said IBM Research scientist, Sani Nassif. "Therefore, they can go deep into the body — almost six inches deep — where they create a very dynamic hot spot while not touching anything between the skin and that point."
Like traditional radiation therapy, the release of energy disrupts the internal chemistry of the affected cells, damaging their DNA and thus preventing them from reproducing or even performing their normal metabolic functions. The cancer cells thus destroyed are then subject the body’s own cleansing mechanisms which flush the dead tissue out of the body.
The details of mapping the beam’s path are daunting, since a nozzle must be used to simultaneously deliver multiple proton beams from different angles each with different energies. The problem — besides the expense of a week’s worth of high-priced labor — is that while the doctors and technicians are plotting out the paths for the beam to follow, the tumor continues to grow, thus making success less likely.
IBM’s solution is to use a supercomputer to quickly plot out the necessary path for the proton beam to follow, presenting numerous alternative therapy plans to the attending physician in just 15 minutes (instead of a week). As a result, by the time that patients can be shuttled from the MRI or CT scanner to the proton-beam therapy room, they can be treated immediately, ensuring that the tumor has not had time to grow further, thus greatly enhancing its chance of success.
"Applying design automation techniques to proton cancer treatment, we also improve the model that predicts what happens when beam hits tumor, compared to manual methods, said Nassif. "It produces thousands of different way to perform the treatment, just like it produces thousands of alternative ways to optimize chip fabrication."
Working with its partners at the University of Texas’s Anderson Medical Cancer Center (Houston) IBM is hoping to reduce the cost of proton therapy by as much as 60 percent, as well as speed-up the treatment planning time using Power7 cluster supercomputers, which perform the computational tasks up to 1000-times faster than the manual methods used by doctors today.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
