Distributed Evolutionary Design of HIFU Treatment Plans

presentation, poster

English

High-Intensity Focused Ultrasound (HIFU) is a modern technique for non-invasive, non-ionising cancer treatment where malignant tissue is destroyed by thermal ablation. To perform such treatment, a plan consisting of a series of short sonications must be created. In recent years, a realistic thermal model using patient-specific materials was introduced. Thanks to this, an evolutionary strategy was employed to design such HIFU treatment plans. However, despite the improvements, the execution time of such a model was too prohibitive to allow routine use. In this paper, we present a new round of optimizations aimed at improving the performance of the thermal model and the evolutionary strategy. The experiments carried out in other works using an evolutionary strategy based on the island model are repeated with up to 4 times decrease in total computation time. This enables new experiments using a much higher number of sonications are introduced and evaluated. Lastly, a new real-life inspired benchmark is presented and the experiments repeated.

Evolutionary strategy, Island model, Farmer-Workers model, HIFU, Treatment planning, k-Wave

CHLEBÍK, J.; JAROŠ, J.

7. 7. 2021

Association for Computing Machinery

Lille

978-1-4503-8351-6

GECCO 2021 Companion - Proceedings of the 2021 Genetic and Evolutionary Computation Conference Companion

297

298

2

https://dl.acm.org/doi/10.1145/3449726.3459550

GECCO21-poster.pdf