NUMBER THREE: BRINGING BIG DATA TO RADIATION THERAPY

It might seem hard to imagine how computers could assist radiation oncologists more than they already do. Throughout the world, clinicians routinely use sophisticated digital tools to analyze patients’ tumors, to calculate radiation doses, and to operate the machines that actually deliver the radiation.

That might sound very state of the art. But a team of Johns Hopkins engineers and physicians believes that another wave of digital transformation in radiation oncology is just starting to take shape. What would happen, the Johns Hopkins team wonders, if the computers in radiation clinics around the world began to learn from one another?

“What we have tried to do,” says computer scientist Russell Taylor ’70, the John C. Malone Professor at the Whiting School, “is use all of the data from patients who have been previously treated to improve planning for future patients. If you build a learning system, you can use technology to make radiation therapy safer, more accurate, and more effective.”

The central challenge of radiation therapy is doing maximum damage to tumors while minimizing damage to nearby healthy tissue. In typical practice, radiation dosimetrists must go through a long trial-and-error process to calculate the optimal plan for each patient. But by statistically analyzing thousands of past patients’ plans, Taylor and his colleagues have built computer tools that can make the dosimetry-planning process much faster and more efficient.

“We can analyze outcomes data, toxicity data, and the three-dimensional relationships between target tissues and the tissues we want to spare,” says the project’s leader, Todd McNutt, an associate professor of radiation oncology physics at the School of Medicine. As the database—known as Oncospace— gets larger over time, its accuracy and predictive power will only grow, McNutt adds.

The system has already begun to improve clinical practice for head-and-neck cancer patients at Johns Hopkins. There is early evidence, McNutt says, that Oncospace-enhanced planning has reduced Johns Hopkins patients’ rates of xerostomia, the severe dry mouth that often occurs when radiation damages the salivary glands.

“What has made this successful,” McNutt says, “is that we have a few physicians who are very sold on the concept. We’ve been able to demonstrate to them that there’s value in these processes—value to their patients, and also academic value.”

This kind of partnership between engineers and clinicians is indispensable, says Taylor, a surgical-robotics pioneer who arrived at Johns Hopkins in 1995 after a long career in industry. “I came to Hopkins,” he says, “because I realized that if I wanted to do this kind of work, it made a lot more sense to be in the same institution as the physicians who will use these projects.”