k-Wave

1. jrosenfield
   

   Member
   Joined: Jan '13
   Posts: 10

   Dear Brad,

   I am trying to run some high-frequency (~15 MHz) plane-wave simulations for a heterogeneous medium in 2D. There is substantial attenuation in the medium, on the order of 50 dB/cm. That probably seems very high, but it is consistent with the frequency dependence of attenuation for breast tissue reported in some of Stuart Foster's earlier papers.

   When I run one of these simulations using a grid point spacing of .025 mm (i.e., a Nyquist frequency of about 29 MHz), I am noticing that there appears to be some artifacts appearing in the simulation display. I think I saw someone refer to this as "ghosting" in the forum; the acoustic field appears in regions where it should not yet have propagated as the simulation runs.

   Do you think this is an issue with the PML being less effective due to the highly attenuating nature of the medium? I changed the PML settings from the default values of thickness = 20 grid points, attenuation = 2 Np/grid point, to a much thicker PML, but the artifacts can still be seen. I have also tried increasing the number of time steps per period. My CFL number is currently on the order of about 0.05.

   I was hoping you might be able to provide some suggestions to fix this issue. If it is just a rendering problem that won't produce inaccurate results, I won't worry about it. However, if there is something I need to do to suppress these artifacts and run accurate simulations, I would be very grateful for any help you can provide. I think this may be producing inaccurate results because the transmitted acoustic field intensity is several orders of magnitude higher than would be expected if the field was being properly attenuated.

   Best,

   Jon

   Posted 9 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,640

   Hi Jon,

   Do you mean the attenuation at 15 MHz is 50 dB/cm? If so, this seems ok. For example, from Szabo's book, the attenuation in breast tissue is given by 0.75 * 15 ^ 1.5 = 43.5 dB/cm.

   A couple of things to try:

   (1) Do you still see the artefacts if the PML is turned off completely, i.e., by setting PMLAlpha to 0?

   (2) Do you still see the artefacts if you make the simulation homogeneous? For a homogeneous medium, you could also try checking the results from kspaceSecondOrder, as this encodes the attenuation exactly.

   Brad.

   Posted 9 years ago #

3. jrosenfield
   

   Member
   Joined: Jan '13
   Posts: 10

   Hi Brad,

   I tried your second suggestion to make the simulation homogeneous and that seemed to get rid of the ghosting artifacts. The artifacts seemed to be most prominent when the acoustic field reached the boundary between two different media. Thank you for the suggestion.

   One other problem I am facing is that it appears there might be something wrong with the acoustic attenuation in my simulation. If I try to simulate a 15-MHz plane-wave source with a time-averaged intensity of 1 W/cm2 radiating into a homogeneous medium with 50.5 dB/cm of attenuation, the simulation calculates the correct average intensity at a depth of 1 cm but is several orders of magnitude off at a depth of 2 cm. My grid point spacing supports up to approximately 22 MHz and my CFL number is .13. Do you have any idea why this would be happening? I was thinking it might have something to do with how low the intensity is at 2 cm depth (~8e-11 W/cm2), and that maybe the pressure transmitted through the PML is adding to this value. I have been trying a PML of thickness 20 grid points and an alpha of 2 Np/grid point. Do you think this could be the source of the problem, since the efficacy of the PML is compromised by the high attenuation of the medium? One thing worth noting is that my detector is always immediately adjacent to the PML.

   Thanks again for your help,

   Jon

   Posted 9 years ago #

4. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,640

   Hi Jon,

   Just to confirm, what are the values of alpha_coeff and alpha_power that you are using? I could re-run the script to generate Fig 2.3 in the manual with attenuation set to see the effect (although I probably won't get to it until the new year).

   In a lossless medium, the PML at normal incidence will reduce the pressure magnitude by 60 to 80 dB, giving approximately 3 or 4 decimal places of accuracy. The intensity drop you mention is a reduction in pressure by a factor of ~10^5, so it's very likely that the accuracy of the PML is affecting your results.

   Brad.

   Posted 9 years ago #

RSS feed for this topic

Reply

You must log in to post.