Copenhagen, Denmark
Onsite/Online

ESTRO 2022

Session Item

Saturday
May 07
16:55 - 17:55
Room D2
FLASH
Charlotte Robert, France;
Fernanda Villegas-Navarro, Sweden
1530
Proffered Papers
Physics
17:25 - 17:35
Improved dose rate by scanning-pattern optimization in FLASH proton therapy of small lung lesions
Steven Habraken, The Netherlands
OC-0282

Abstract

Improved dose rate by scanning-pattern optimization in FLASH proton therapy of small lung lesions
Authors:

Rodrigo José Santo1, Steven Habraken1,2, Sebastiaan Breedveld1, Mischa Hoogeman1,2

1Erasmus University Medical Center, Radiotherapy, Rotterdam, The Netherlands; 2HollandPTC, Medical Physics & Informatics, Delft, The Netherlands

Show Affiliations
Purpose or Objective

FLASH dose rates >40Gy/s are readily available in proton therapy (PT) with cyclotron-accelerated beams and pencil-beam scanning (PBS). The overal highest is achieved with protons at maximum energy shooting through the patient (transmission beam, no Bragg peak). The local time structure of the PBS, quantified in terms of PBS dose rate (PBS-DR), is critical for FLASH. However, methods to optimize it are lacking. Our aim was to optimize patient-specific scanning patterns in stereotactic FLASH-PT of small lung lesions, maximizing the percentage irradiated with a PBS-DR>40 Gy/s of the OAR voxels irradiated to >8Gy (FLASH coverage).

Material and Methods

Plans to 54Gy/3 fractions with 3 equiangular coplanar 244MeV proton shoot-through beams for 20 patients were optimized in in-house developed software. PTV-based planning with a 5mm margin was used with a median PTV of 8.7cc (range: 4.4-84 cc). Beams avoided serial OARs and, for each direction, the shortest path length from beam entrance to PTV was used. To enable FLASH-enhanced single-beam per fraction delivery, a single-field uniform dose approach was used. Sequential scanning pattern optimization was performed with a Genetic Algorithm to optimize the PBS-DR, run in parallel for 20 independent populations (islands). Mapped crossover, inversion, swap, and shift operators were applied to achieve (local) optimality on each island. Migration between islands was implemented to approach global optimality. The cost function was chosen to maximize the FLASH coverage at >8 Gy, >40Gy/s. The PBS-DR was calculated for a beam current of 40 nA. The optimized scanning patterns were evaluated in terms of the PBS-DR distribution and population PBS-DR-volume histograms, compared to conventional line by line scanning (snake pattern).

Results

Optimized patterns for various PTVs are shown in figure 1. They have a snowflake-like structure with adjacent pencil beams being irradiated consecutively where possible. For larger PTVs the scanning pattern and the high-PBS-DR volume develop a swirl-like shape.



Figure 1: Optimized scanning patterns with initial (green) and final (red) positions, overlaid with the resulting PBS-DR distribution in a beam transverse plane through the PTV.


Population median and quartiles of the PBS-DR-volume histograms are displayed in figure 2. Optimized patterns result in a significantly (p<0.01) larger FLASH coverage at 8 Gy, 40Gy/s, with a median increase of 198% (86cc) relative to the snake pattern.



Figure 2: Median, 1st and 3rd quartiles of PBS-DR volume histograms for the ipsilateral lung, including the 5mm PTV margin, irradiated to >8Gy with snake and optimized scanning patterns.

Conclusion

Significant improvements on the PBS-DR and, hence on the FLASH coverage and the potential healthy-tissue sparing are obtained by sequential scanning-pattern optimization. The patterns depend on the FLASH dose and DR thresholds and on the PBS-DR threshold. Therefore, it is critical to determine the dose and delivery time structure for which FLASH occurs.