Saving time with AI-generated treatment plans for breast cancer

Drawing in the organs of individual breast cancer patients and then creating precise radiation plans appears to be faster by using artificial intelligence (AI) models. That way, it remains just as reliable and accurate. It saves considerable time per patient—a pleasant conclusion with current health care developments in mind.

This is the net result of the scientific research Nienke Bakx conducted in recent years at the Catharina Hospital. She received her doctorate from Eindhoven University of Technology (TU/e) on September 19. Her research deserves further explanation. Thanks to newly developed AI models, breast cancer patients can have their tumors and surrounding organs drawn in largely automatically.

Manual drawing

Until two years ago, however, this was done with “old-fashioned” manual work. Once the anatomy of the individual patient was imaged, the radiation plan—how best to administer radiation to destroy cancer cells while sparing healthy organs—could be made. That, too, was done manually.

“If you express it in minutes, it saves half an hour per patient,” says Bakx

After a successful pilot, it was put into practice in May 2022. With the caveat that radiotherapists and lab technicians check everything, make necessary adjustments, and give the “go.”

Bakx says that things are going well. “You see the biggest time savings when you automatically draw in the organs, about 60% of the time. For drawing in the tumor, about 40%. If you express it in minutes, it saves half an hour per patient.”

Radiation plans

In addition to drawing in tumors and organs, AI models can also be used to create the radiation plan. Here, too, using artificial intelligence proved successful: in 74% of cases, the AI model could create a radiation plan that was immediately usable. With minor adjustments, this percentage increased to 86%.

The time gained is crucial for the long term, Bakx says. “Health care in the Netherlands is under pressure due to the aging population, the associated increasing number of cancer patients, and staff shortages. In Eindhoven, we are still in a good position thanks to the medical imaging and radiotherapy techniques (MBRT) study at Fontys.

“Still, at many hospitals, you see vacancies that are not being filled. Therefore, we must start doing something about this because otherwise, the work pressure will become too great.”

Relevant innovation

She has a feel for relevance. “That’s what’s basically happening now with AI: we need more people to develop these kinds of models, but in the long run we need fewer medical people as a result. I see this period and all the work we are doing in it as investing in a sustainable health care solution.”

With this research, Bakx has done what she had wanted to do before. “I knew as a student that studying medicine wasn’t for me, but I did find the medical world very interesting. I wanted to improve it through technology. The great thing about this course is that it also led to implementing an innovative method. That is certainly not always the case with scientific research.”

A fitting conclusion

The Eindhoven native concluded her time at Catharina Hospital with the defense of her dissertation. “It makes me very proud. I started here during the master’s part of my medical engineering studies, specializing in medical imaging. Then I continued at the ‘Cathrine’ for the postmaster’s program Qualified Medical Engineer (QME).”

“This is practice-oriented follow-up research where I developed and evaluated the AI models. That this would eventually result in scientific research and a Ph.D., I would never have believed back then. That always seemed far too theoretical to me.

“The great thing is that I collaborated this entire period with my supervisor and clinical physicist Coen Hurkmans, who became a professor last year.”

The success of AI models in breast cancer has led to expansion into other types of cancer. For example, automated entries are now being applied to esophageal and lung cancer patients. Although these applications were not directly part of Bakx’s research, they have been made possible by its results.

“In practice, we see that it works well in different target areas. It saves time everywhere,” says Bakx, who has also been project leader at the AI Expertise Center at Catharina Hospital for the past two years.

Her work is continued by her colleague Niels van Acht, who is investigating how these AI models can be used sustainably. “It is important that the models continue to work even if something changes in medical guidelines or workflows. The goal is to make this process as efficient as possible and ensure that the models will still be relevant in a few years.”

The thesis is titled “Automation of radiotherapy treatment planning for breast cancer—Utilizing artificial intelligence for automated segmentation and planning,” and the supervisors were Coen Hurkmans and Hanneke Bluemink.