Publication Details
Evolutionary Optimization of a Focused Ultrasound Propagation Predictor Neural Network
Evolutionary Optimisation, Evolutionary Design, Ultrasound Propagation Predictor,
Cartesian Genetic Programming
The search for the optimal treatment plan of a focused ultrasound based procedure
is a complex multi-modal problem, trying to deliver the solution in clinically
relevant time while not sacrificing the precision bellow a critical threshold. To
test a solution, a multitude of computationally expensive simulations need to be
evaluated, often thousands of times. Recent renaissance of machine learning could
provide a solution to this. Indeed, a state-of-the-art neural predictor of the
Acoustic Propagation through a human skull was published recently, speeding up
the simulation significantly. The utilized architecture, however, could use some
improvements in precision. To explore their design more deeply, we made an
attempt to improve the solver by use of an evolutionary algorithm, challenging
the importance of different building blocks. Utilizing Genetic Programming, we
managed to improve their solution significantly, resulting in a solver with
approximately an order of magnitude better RMSE of the predictor, while still
delivering solutions in reasonable time frame. Furthermore, a second study was
conducted to gauge the effects of the multi-resolution encoding on precision of
the network, providing interesting topics for further research on the effects of
the memory blocks and convolution kernel sizes for PDE RCNN solvers.
@misc{BUT185147,
author="Jakub {Chlebík} and Jiří {Jaroš}",
title="Evolutionary Optimization of a Focused Ultrasound Propagation Predictor Neural Network",
booktitle="GECCO 2023 Companion - Proceedings of the 2023 Genetic and Evolutionary Computation Conference Companion",
year="2023",
pages="635--638",
publisher="Association for Computing Machinery",
address="Lisbon",
doi="10.1145/3583133.3590661",
isbn="979-8-4007-0120-7",
url="https://www.fit.vut.cz/research/publication/12949/",
note="presentation, poster"
}