Publication Details
On the usage of the Sparse Fourier Transform in ultrasound propagation simulation
Fourier transform, Sparse Fourier transform, high performance
computing, k-Wave, ultrasound wave propagation
The Fourier transform is an algorithm for transforming the signal
from the space/time domain into the frequency domain. This algorithm is essential for applications like image processing, communication, medicine, differential equations solvers, and many others.
In some of these applications, most of the Fourier coefficients are
small or equal to zero. This property of the signals is used by the
Sparse Fourier transform which estimates significant coefficients of
the signal with a lower time complexity than the Fourier transform.
The goal of this paper is to evaluate available implementations of
the Sparse Fourier transform on a set of benchmarks solving the
ultrasound wave propagation in 1D, 2D, and 3D heterogeneous
media. The results show that the fastest available implementation
in 1D domains is MSFFT, however, it is not possible to use it in
our implementation of the 2D Sparse Fourier transform. Thus the
AAFFT 0.9 is selected for our implementation of the 2D Sparse
Fourier transform as the most stable and acceptably fast implementation. The results on 3D simulation data show, that by using the
SpFFT library it is possible to reduce the computation time of the
Fourier transform in ultrasound wave propagation simulation.
@inproceedings{BUT188187,
author="Ondřej {Olšák} and Jiří {Jaroš}",
title="On the usage of the Sparse Fourier Transform in ultrasound propagation simulation",
booktitle="ICBRA '23: Proceedings of the 10th International Conference on Bioinformatics Research and Applications",
year="2024",
pages="107--113",
publisher="Association for Computing Machinery",
address="New York",
doi="10.1145/3632047.3632064",
isbn="979-8-4007-0815-2",
url="https://dl.acm.org/doi/10.1145/3632047.3632064"
}