New Technology for Brachytherapy to Test at U of Rochester
December 22, 1999. Scientists at the University of Rochester Medical Center are exploring ways to improve brachytherapy as a treatment for prostate cancer by using strings of binary digits in algorithms.
Using strands of 0s and 1s to stock a digital community, complete with mating, offspring, and the occasional mutation, is nothing new to scientists who create genetic algorithms. These mathematical formulas are governed by Darwin's "survival of the fittest" mantra and rely on the principles of evolution to create solutions in a variety of applications, including financial analyses, computer-assisted scheduling and target detection for the military.
Now scientists are using them to improve how radiation is delivered. The formulas have been developed by medical physicist Yan Yu, Ph.D., an expert in radiation treatment planning who heads a task force on the subject for the American Association of Physicists in Medicine. His algorithms are being used in brachytherapy, where tiny radioactive seeds are implanted into the prostate and destroy cancer cells in the organ over several weeks.
Medical physicists play an often unseen role in cancer treatment, ultimately deciding how to deliver radiation to best eradicate cancerous tissue or organs without hurting healthy tissue. In cases of prostate cancer, the stakes are very high: the bladder, rectum, urethra, and nerves that control sexual function are all packed together near the organ, making it an especially challenging disease to treat.
Choosing the right pattern for the seeds, which are radioactive particles about the size of a grain of rice, is daunting. Physicians commonly turn to commercial programs to help them decide how to place the particles. The recommended treatment plan is put in the hands of a surgeon, who actually inserts several dozen seeds into the prostate during a one- or two-hour surgical procedure.
Yu's work, dubbed PIPER for Prostate Implant Planning Engine for Radiotherapy, uses artificial-intelligence technology to recommend a radiation treatment plan. Based on an ultrasound scan of a patient's prostate and other pelvic organs, PIPER sets up a competition between the possible configurations. Yu and colleagues create a digital community with 64 "members" whose binary codes each represent a different radiation pattern. Community members compete to pass on their "genetic material" -- bits of binary code -- to the next generation. Each pattern's viability is determined by mathematical criteria which favor radiation plans that irradiate the prostate efficiently and knock out cancerous cells while sparing vital organs. In a game of virtual natural selection, binary code that embodies these qualities survives and multiplies, while poor code perishes.
Over the course of 200 generations the community evolves, with codes coming together randomly, combining their genetic material, and even mutating occasionally. In this way, the genetic algorithm creates a huge range of potential solutions that it constantly sifts to find the best candidate for a treatment plan. "A genetic algorithm can look at a much greater range of options than we otherwise could. There might be certain combinations that would never occur to a physicist to try," says Yu.
In just two minutes, PIPER presents to physicists and physicians the "winner" - the plan that the program decides is most likely to work best. The speed of PIPER, whose underlying technology has been patented by the University, may make it possible to do the radiation planning right in the operating room immediately before surgery, instead of several weeks beforehand as is now standard. That's a tremendous advantage, says surgeon Edward Messing, M.D., chair of the Department of Urology, who has performed scores of brachytherapy procedures at the University's Strong Memorial Hospital.
"The prostate you see in the operating room is never the same one you saw three weeks previously in your office," says Messing. "Hormone therapy before surgery can shrink the prostate, for instance, and even anesthesia can change the positioning of the pelvis the day of surgery. This oftentimes makes deviations from the now-dated plan necessary during surgery." Yu's goal is to make the process more precise. With a radiation treatment plan compiled just minutes before the operation, the plan is more likely to match what physicians actually confront in the operating room.
The University of Rochester has commercialized this new technology. They have joined with Real Time Enterprises, a Rochester software engineering firm, to create a new company, RTek Medical Systems LLC. RTek will focus on the development of new radiation treatment planning systems based on Yu's algorithms. The system must now undergo tests in order to apply for approval from the U.S. Food and Drug Administration, the FDA.
The FDA has approved a clinical trial of the current system as an investigational device on about 30 patients. Messing, Yu and their colleagues will study the radiation treatment plans recommended by the system, along with the effects on tumor control, quality of life, and complications in patients who get the PIPER treatment compared to patients who receive a commercially available treatment. The clinical study will build on more than five years of basic research that has been funded by a variety of sources, including the National Cancer Institute and the Whitaker Foundation.
University of Rochester