There's only so much you can do with limited information. If you know the attenuation (in dB/cm) the authors reported and the frequency (in MHz) of the ultrasound in their experiments, you could simply divide the attenuation by the frequency squared. There is no way to know if your power law exponent of 2 is correct, though.

If the authors provided multiple attenuation values at different frequencies, you can fit a power regression model to the data points. This essentially means fitting the parameters a, b in:

log(attenuation) = a + b * log(frequency)

via ordinary least squares. Once the parameters have been estimated, you can take an exponential to get:

exp(log(attenuation)) = exp(a + b * log(frequency))

attenuation = exp(a) * exp(log(frequency)) ^ b

attenuation = exp(a) * frequency ^ b

When using k-Wave, `exp(a)`

becomes `medium.alpha_coeff`

and `b`

becomes `medium.alpha_power`

.

Perhaps other people here might have different suggestions?