Artificial intelligence now is being used to support planning in radiation therapy, reducing the time needed to manually delineate structures on multiple images.
Radiotherapy treatment planning is using software that employs AI in two Finnish hospitals, with the hope that it will be more accurate and faster in planning procedures and plotting how therapy will be delivered.
The testing began last week, according to executives at MVision, which is a Finnish company focusing on the development of AI software in medical imaging.
The software is being used at the Department of Oncology and Radiotherapy at the Turku University Hospital and the Docrates Cancer Center in Helsinki. The initial deployment of the software is aimed at prostate cancer patients, but soon the software will accommodate other cancer sites as well, including head, neck and breast cancer.
“AI enables a more personalized care for cancer patients,” says adjunct professor and medical physicist Timo Kiljunen from Docrates Cancer Center. “As the software takes care of the core process of radiation treatment planning, the physician can focus more on the individual needs of the patient, such as taking the spread and aggressiveness of the cancer into consideration.”
For each cancer patient receiving radiotherapy, a detailed treatment plan is compiled to determine where and how radiotherapy is delivered. Factors to consider in a treatment plan include the size and location of the tumor, as well as possible metastases.
Until now, a physician has had to define the radiotherapy target by manually delineating structures onto multiple X-ray image slices of the patient. There can be hundreds of these images to manually process, and dozens of critical tissue structures may be plotted, resulting in a time-consuming process that is prone to error.
MVision’s AI application has been trained with image sets from hundreds of patients, enabling the software to delineate critical tissue structures automatically for a physician to review and approve.
“AI speeds up a physician’s job and allows for repeated and uniform treatment plans on a hospital level, not only on an individual physician’s level,” says Heikki Minn, professor and head of Tyks Department of Oncology and Radiotherapy. “With a high number of cancer patients, having quick treatment planning can also help with waiting times.”
The use of AI also could reduce treatment errors, the researchers contend. Currently, the most significant source of error in radiation treatment planning comes from the evaluation of the area of treatment, previous research has found.
With an AI-based software, uniform results are produced, decreasing the possibility of human error. The AI is also constantly learning to speed up and improve its delineations of structures. The result always requires a physician’s examination and, if necessary, enables the physician to adjust the target and organs-at-risk delineation. When the delineation task is complete, a medical physicist or radiotherapy technologist elaborates an optimized radiation treatment plan.
“In the future, AI is going to be part of radiotherapy dose calculation and even cancer diagnosis, e.g. MRI-targeted biopsy,” Kiljunen predicts.