The ultrahigh dose rates used in FLASH radiotherapy may increase the therapeutic window by protecting normal tissues against radiation damage. In addition, some researchers believe that FLASH proton beams might be available with commercially available cyclotron-accelerated proton beams. But when FLASH is combined with the most advanced type of proton therapy, lateral pencil-beam scanning (PBS), the very PBS proton deliveries used to treat complex cancers with unparalleled precision also impact the local dose rates critical to achieving the FLASH effect.
Researchers at Erasmus University Medical Center, Instituto Superior Técnico and HollandPTC set out to account for the local variations in dose rate resulting from PBS proton delivery. Their recent study, reported in International Journal of Radiation Oncology Biology Physics, maximizes FLASH coverage by optimizing the PBS scan pattern with voxel-based metrics.
“We were trying to optimize FLASH through optimization of dose rate, without compromising plan quality in terms of radiation dose,” says lead author Rodrigo José Santo. “We were trying to set up a pipeline that would consistently optimize FLASH coverage for different tumour shapes and sizes, without re-optimizing the treatment plan and considering FLASH as a local effect dependent on the pencil-beam delivery pattern.”
The result: optimizing FLASH proton therapy treatment plans without compromising dose rate.
PBS as a travelling salesperson
The travelling salesman problem poses the following question: “Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city exactly once and returns to the origin city?”
This problem, long studied by combinational optimization researchers, is a barometer for genetic algorithms used in computer science and operations research. José Santo, who is currently a doctoral student at UMC Utrecht but was a master’s student when the work was performed, realized that genetic algorithms could be used to solve his own problem – optimizing the order in which proton pencil beams are irradiated to maximize FLASH coverage.
The researchers’ resulting approach uses a voxel-based metric defined by fixed-dose thresholds to determine when irradiation of that voxel starts and ends. The algorithm evaluates dose rate for each pencil beam separately and assumes that FLASH is a local effect and that total irradiation time is a critical FLASH parameter.
The algorithm is run on different solutions in parallel, though it occasionally shares information between them. The average distance between pencil beams is included as a cost function to minimize the total distance travelled in the plane transverse to the beam direction. The algorithm is applied sequentially after pencil-beam positions and weights are optimized and without compromising the plan quality in terms of (nominal) absorbed dose.
The researchers tested their algorithm on treatment plans using transmission proton pencil beams for 20 patients with early-stage lung cancer and lung metastases. (Lung lesions are ideal sites for FLASH, the researchers say – current FLASH proton treatments involve high-energy beams that pass through the patient rather than the Bragg-peak beams harnessed for conventional proton therapy.)
Median FLASH coverage improved from 6.9% for standard line-by-line scan patterns to 29% with PBS optimization. The researchers observed that PBS-optimized plans have a whorl-like appearance. The FLASH window changed only slightly for marginally different beam currents.
FLASH proton therapy: uncovering the optimal delivery technique
Since other research groups are primarily working to optimize FLASH at the treatment planning level, the researchers say that it’s challenging to compare their own PBS-optimized results to other FLASH proton therapy studies – to their knowledge, this study is the first to perform pencil-beam delivery pattern optimization for FLASH proton therapy. They are now focusing on optimizing PBS delivery for larger targets and integrating dose-rate optimization into their existing dose optimization pipeline.
“Radiation therapy is still [being] continuously improved, and the FLASH effect is a promising path to better treatment outcomes for patients. Proton therapy, combined with optimization algorithms such as the one we have developed, is an important step towards achieving exactly that,” José Santo says. “Our manuscript underlines that there is a lot of room for further optimization of FLASH proton therapy as a treatment modality, even with current beam hardware.”
