Validation

Every solver in SonoSK is compared with a closed-form result or with an independent method. These comparisons are part of the test suite in the tests folder and run on every change of the code, on Linux and Windows.

Quantity	Reference	Agreement in the tests
On-axis field of a flat piston, Rayleigh integral	closed form (Kinsler et al. 2000)	better than 0.5 % beyond 20 mm
On-axis field of a focused bowl, Rayleigh integral	O’Neil (1949)	better than 0.5 % of the peak
Attenuation in the Rayleigh integral	\(\alpha_0 f z\) in dB	1 %
Angular spectrum propagation	Rayleigh integral on the target plane	4 % of the peak
Area of discretised piston and bowl	\(\pi a^2\), \(2\pi R h\)	exact
Focusing delays	equal arrival time at the focus	exact
Axial resolution of a simulated point target	\(-6\) dB width of the two-way Gaussian pulse	5 %
Position of a simulated point target	true position	within \(\lambda/8\) axially
Cyst contrast in a simulated image	anechoic region	below \(-15\) dB, gCNR above 0.8
Speckle of a Rayleigh envelope	\(\sqrt{\pi/(4-\pi)} = 1.913\) (Burckhardt 1978)	1 %
Heating without conduction or perfusion	\(\Delta T = Q t / (\rho c_p)\)	exact
Heating with perfusion only	\(\Delta T = \frac{Q}{w_b c_b}\left(1 - e^{-t w_b c_b / \rho c_p}\right)\)	0.1 %
Conduction of a Gaussian temperature profile	analytic diffusion	0.1 % of the peak
CEM43 at constant temperature	43 °C for 60 min and 44 °C for 30 min both give 60 min; 42 °C for 60 min gives 15 min	exact
Electrical impedance of an unloaded plate, Mason model	closed form	\(10^{-6}\)
Energy balance of a loaded stack	electrical power = acoustic power	\(10^{-9}\)
Matching layer	Desilets et al. (1978) formula; wider band than without
Pulse duration, PII, ISPPA, MI	IEC 62359 and FDA definitions on test signals	exact
Water sound speed and density	Marczak (1997), Kell (1975)	published values

To run the tests yourself, from a copy of the repository:

git clone https://github.com/sonosk/sonosk.git
cd sonosk
pip install -e ".[test]"
pytest

Comparison with other simulators

Comparisons with established tools on shared test cases are planned as a benchmark set (see the Roadmap). They will use the published plane-wave imaging challenge data (PICMUS) for beamforming and the transcranial benchmark problems of Aubry et al. (2022) for the full-wave solver.

References

Aubry J.-F. et al. Benchmark problems for transcranial ultrasound simulation: intercomparison of compressional wave models. J. Acoust. Soc. Am. 152(2):1003-1019, 2022.

Burckhardt C.B. Speckle in ultrasound B-mode scans. IEEE Trans. Sonics Ultrason. 25(1):1-6, 1978.