In a radiation therapy room, a patient remains under anesthesia while the team manually adjusts a treatment plan that can take more than an hour. In
brachytherapy of
cervical cancerevery millimeter of dose counts. Doctors must place the radioactive source as close as possible to the tumor while sparing the bladder and rectum, which requires a long series of calculations and checks. Researchers from the University of California at San Diego today presented a tool forartificial intelligence in one click, capable of greatly reducing the time spent on this step.
Brachytherapy: current planning slows down brachytherapy
According to the World Health Organization, the cervical cancer affects approximately 600,000 women and causes 340,000 deaths per year worldwide. In France, there are an estimated 3,100 new cases annually, for nearly 1,000 deaths. There
brachytherapy is however a reference treatment, but rarely offered because the planning is cumbersome.
There treatment planning in brachytherapy remains largely manual, often with more than an hour of work even for simple cases without needles. Much of this time takes place while the patient is sedated or in discomfort. The quality of the plan can vary between centers and clinicians, exposing some patients to less effective treatments.
The first self-planning scripts remained external to the clinical software and required manually exporting and then reimporting DICOM files ( Digital Imaging and Communication in Medicine – the standard for digital files created during medical imaging examinations), a time-consuming step and a potential source of errors. The San Diego team wanted to eliminate this break by running the algorithm directly in the planning system used on a daily basis by radiotherapy teams.
An AI tool allows treatment plans to be established in less than 4 minutes
Lance Moore sums it up: “The new tool uses AI to automate and accelerate treatment planning and, with one click, the system analyzes a patient’s medical images and creates a high-quality personalized plan in less than four minutes – potentially reducing both patient discomfort and the risk of human error“, explained the researcher from the Department of Radiation Medicine and Applied Sciences at UC San Diego.
This new technology combines advanced deep learning with optimized data processing. Tested on hundreds of patient files, it allowed researchers to note that the treatment plans developed by the tool were of equivalent quality to those established by experienced doctors, but much faster. In detail, the difference between the dose calculated by the AI and that of the clinical plans was 3.8%, with 10.3 seconds difference in the source times, for an execution time of approximately 3.5 minutes.
The authors judge the quality obtained to be similar to that of expert plans. “This approach could help standardize care, particularly in clinics with fewer resources or less specialized staff, and could allow for more focus on improving the quality of the plan rather than rushing to complete” said Lance Moore.
What are the benefits for patients?
By reducing planning to minutes, this tool can limit time spent under anesthesia, reduce the risk of human error and help under-resourced centers. It should be able to be adapted to other cancers, such as breast or prostate cancer.