Diagnostics & Imaging Week

Radiation therapy has come a long way since the days of treating tumors by directing a radiation beam in the general vicinity of a patient's tumor and hoping for the best.

With intensely modulated radiation therapy (IMRT) —the newer approach for targeting radiation — each beam is composed of thousands of tiny "beamlets" that can be individually modulated to deliver a tighter distribution of the radiation dose right around the tumor.

IMRT is an "exciting technique," Richard Radke, assistant professor of electrical, computer, and systems engineering at Rensselaer Polytechnic Institute (Troy, New York), told Diagnostics & Imaging Week. But it "can require hours of manual tuning to determine an effective radiation treatment for a given patient."

Hoping to speed up the process, Radke is leading a team of engineers and medical physicists to develop a "machine learning" algorithm that could save radiation physicists hours of manual adjustment associated with the popular cancer treatment.

In a paper published in the Feb. 7 issue of Physics in Medicine and Biology, the Rensselaer researchers describe their computer-based approach that could automatically determine acceptable radiation plans in a matter of minutes, without compromising quality of treatment.

A subfield of artificial intelligence, machine learning is based on the development of algorithms that allow computers to learn relationships in large datasets from examples. Radke and his team have tested their algorithm on 10 prostate cancer patients. They found that for 70% of the cases, the algorithm automatically determined an appropriate radiation therapy plan in about 10 minutes.

"The main goal of radiation therapy is to irradiate a tumor with a very high dose, while avoiding all of the healthy organs," Radke said.

IMRT adds nuance and flexibility to radiation therapy, increasing the likelihood of treating a tumor without endangering surrounding healthy tissue. But the semi-automatic process of developing a treatment plan can be extremely time-consuming — up to about four hours for prostate cancer and up to an entire day for more complicated cancers in the head and neck, Radke said.

A radiation planner must perform a CT scan, analyze the image to determine the exact locations of the tumor and healthy tissues, and define the radiation levels that each area should receive. Then the planner must give weight to various constraints set by a doctor, such as allowing no more than a certain level of radiation to hit a nearby organ, while assuring that the tumor receives enough to kill the cancerous cells.

Radke said this is currently achieved by manually determining the settings of up to 20 different parameters, or "knobs," deriving the corresponding radiation plan, and then repeating the process if the plan does not meet the clinical constraints.

"So one of our goals was to see if we could, No. one, identify which of these knobs was really critical, and, second, to automate this knob-turning process," Radke said, which saves the planner time by removing decisions that don't require their expert intuition.

The researchers first performed a sensitivity analysis, which showed that many of the parameters could be eliminated completely because they had little effect on the outcome of the treatment. They then showed that an automatic search over the smaller set of sensitive parameters could theoretically lead to clinically acceptable plans.

The procedure was put to the test by developing radiation plans for 10 patients with prostate cancer. In all, 10 cases for this process took between five and 10 minutes, Radke said. Four cases would have been immediately acceptable in the clinic; three needed only minor "tweaking" by an expert to achieve an acceptable radiation plan; and three would have demanded more attention from a radiation planner.

Radke said his team also have applied their approach to breast tumors, and they were able to reduce radiation planning time from about a half-hour to mere seconds. "So for the breast we moved from minutes to seconds and for the prostate we moved from hours to minutes," Radke said.

The challenging part of the research, at first, Radke said, was "for us as electrical engineers to really immerse ourselves in the terminology of the clinic … learning the language of the radiation planners."

To overcome that challenge, Radke said a student on his team spent time at the hospital, working side-by-side with the radiation planners to make sure what the researchers were doing in the lab would work in a clinical setting.

"It's hard to duplicate what goes on in the clinic," Radke said. "It was surprising that there was a sizeable gap between what we could simulate here and what they actually did in the clinic."

But again, he said having a student spend time at the hospital helped the team more closely mimic what was happening in the clinic.

Radke and his team plan to extend their work to head and neck tumors. He said they also want to make the algorithms more general so that the approach can be applied to other types of tumors as well. "We're trying to get the idea into the clinics," he said.

His team, he added, hopes to work with some software manufacturers to integrate their tools with the clinical software programs.

In a related project, Radke is collaborating with colleagues at Boston's Massachusetts General Hospital (Boston) to create computer vision algorithms that offer accurate estimates of the locations of tumors. This automatic modeling and segmentation process could help radiation planning at an earlier stage by automatically outlining organs of interest in each image of a CT scan, which is another time-consuming manual step.

The research is supported by the National Cancer Institute and the Center for Subsurface Sensing and Imaging Systems at Rensselaer, which is funded by the National Science Foundation.

Renzhi Lu, a graduate student in electrical engineering at Rensselaer, also contributed to the research.

Rensselaer Polytechnic offers bachelor's, master's, and doctoral degrees in engineering, the sciences, information technology, architecture, management, and the humanities and social sciences.