Radiotherapy Plans Based on Machine Learning Being Used to Treat Localized PC Patients

Radiotherapy Plans Based on Machine Learning Being Used to Treat Localized PC Patients

Radiotherapy plans generated with machine learning in the RayStation system are now being used at the Princess Margaret Cancer Centre, in Canada, to treat men with localized prostate cancer.

RayStation is a treatment planning system (TPS) to optimize care for cancer patients. It combines features such as adaptive therapy, multi-criteria optimization, and algorithms to improve intensity-modulated radiation therapy (IMRT) — a type of radiotherapy — and volumetric modulated arc therapy (a newer type of IMRT), with accurate dose engines for photon, electron, proton and carbon ion therapy.

The technology supports various treatment machines, working as an integrated control center and maximizing value from existing equipment. It also integrates with RayCare, an oncology information system designed to coordinate radiation therapy, chemotherapy, and surgery.

According to RaySearch Laboratories, RayStation represents the first use of such a TPS in the radiation oncology field. It generates high-quality radiation treatment plans in minutes, with no need for human intervention.

The system was developed by the machine learning department at RaySearch, in partnership with Princess Margaret, and at the Techna Institute, both in Toronto. Since May, all patients with localized prostate cancer treated at Princess Margaret have been taking part in a prospective study led by Alejandro Berlin, a radiation oncologist. This study was launched after a 2018 study showed that machine learning plans were preferred, or at least deemed equivalent to manual plans in 94% of cases.

The research contains two blinded treatment plans: one is manually generated, while the other is a machine-learning plan. The selected plan then undergoes peer review and quality control before being used with patients. The scientist said this study will help quantify the performance of machine learning plans and whether they are preferred in real-world practice.

“It has been really exciting for the team to help materialize this machine learning advancement in the radiation oncology field, including deployment into the clinical realm,” Berlin said in a press release. “Our positive results to date validate our observations about the robustness of this planning solution.”

Johan Löf, RaySearch’s founder and CEO, said he was “thrilled” that patients were being treated with machine learning generated plans from RayStation.

“These functionalities are the first of its kind and now it is proven that they function just as we anticipated,” he said.

“Our collaboration with Princess Margaret on this project has been fortunate and will lead to better cancer care for more patients,” Löf added.