Hi,
I am using k-wave 2D and I have a time varying pressure as a source. However k-wave changes the shape of my pressure source for linear acoustic which doesn't make sense at all. I have PML boundaries all around the domain.
Is there anything special with k-wave and source modelling?
Best regards,
Babak
Hi Babak,
Have you looked at the "Photoacoustic Waveforms in 1D, 2D and 3D Example"? This might answer your question.
Brad.
Hi,
Sorry for asking a lot of questions on this forum, but I definitely need help to success my simulations. So thanks again for answer.
I am trying to mimic the example described in the paper "A Discrete Source Model for Simulating Bowl-Shaped Focused Ultrasound Transducers on Regular Grids: Design and Experimental Validation". At the end of the simulation, I don't get the expected result. I obtain something, more or less, similar, so I guess that everything is not wrong, but I don't know where could be the error. I am confused regarding some points:
1) When you say "Simulations were repeated using grid discretisations from 2.2 points per wavelength", does it mean that dx=dy=dz=lambda/2.2 where lambda=c/f with c=1492m/s and f=1.1e6Hz?
2) For the description of the shaped-bowl. I don't understand the sentence "The grid points for which θp is greater than the half arc angle θa are then removed". When I look the drawing on the top of the first page, I don't understand the accordance with "the HALF arc angle". I would say "The grid points for which θp is greater than the arc angle θa are then removed".
3) In the paper, I don't see the information regarding the time. For the creation of the hdf5 file, I need to define dt and Nt. I have seen in the k-wave user manual:
dt=(cfl*dx)/1492
t_end=math.sqrt((Nx*dx)**2+(Ny*dy)**2+(Nz*dz)**2)/1492
Nt=round(t_end/dt)Does it correct to define time constant like that?
4) For the time varying input signal, it's written "the source was driven by a continuous wave sinusoid at 1.1 MHz".
For that, I created a time vector: t_array=np.linspace(0,Nt*dt,Nt) and I define:
p_source_input=np.sin(1.1e6*2.0*np.pi*t_array) . Is it a good way for creating the wanted input signal?
Thanks a lot,
lbruju
Hi lbruju,
There is a more extensive follow up to that paper here that might contain more details. You can find the exact process for forming the bowl in the makeBowl function. For time varying signals, I normally use createCWSignals in k-Wave. This generates a sine wave with a ramp at the beginning to avoid high-frequency transients at the beginning of the signal. Even if you can't use them directly, you should be able to follow the same steps in Python.
For the grid discretisation, yes, that's correct.
Brad.
Hi Dr Bradley Treeby,
Thanks for your answer.
My results are now really closed to yours in the paper. But I am confused regarding the amplitude of the final values I obtain.
The source is a continuous sinusoidal wave at 1.1 MHz and with an amplitude of 0.5 MPa. The computational time is: Nt*dt=0.1ms. At the end of the computation, I obtain a pressure peak at the focal zone around 1.5e7 Pa, corresponding to a volume rate of heat deposition of 2e9 W/m^3. I have implemented the temperature calculation based on the paper "non standard Fourier pseudospectral time Domain (PSTD) schemes for partial differential equations", and I obtain a final temperature at the focal zone around 61 degrees.
I am confused because I thought that these range of values should be obtained with an exposition time of some seconds and not, like in my simulation, with an exposition time of 0.1ms.
Do you have an idea of where the error could be?
Thanks a lot,
lbruju
Hi lbruju,
The values for pressure and volume rate of heat deposition are very high, especially the latter, so I'm not that surprised. To double check, you could try checking your implementation of the heat code against the closed form solution given in bioheatExact.m (note, this is only valid for homogeneous media).
Brad.
Hi,
Thanks for your answer. At the beginning, I have chosen an amplitude for the continuous sinusoidal wave of 0.5MPa because this is the value used in your example "Heating By A Focused Ultrasound Transducer". But my simulation is computed using a bowl-shaped transducer (in 3D) whereas in the example, the source is an arc (2D). I have finally found your sentence in the manual "the surface pressure is relatively high (0.5MPa) to achieve the required focal pressure to ablate the tissue. This is because the simulation is in 2D, and thus the focusing gain is much less than an equivalent simulation in 3D using a focused bowl transducer of the same radius and diameter" that makes me explain why my values for pressure and volume rate of heat deposition are very high.
Do you think that it can be the real explanation of my high values? Could you give me a range of usable initial amplitude to reach correct temperature after 10s of heating using a bowl-shaped transducer and to achieve the required focal pressure to ablate the tissue?
I don't use the function bioheatExact but the kWaveDiffusion one. Can you confirm that they give almost the same result?
Is the kWaveDiffusion function only valid for homogeneous media too?
Thanks,
lbruju
There is another confused point (in the case of the kWaveDiffusion function is valid for heterogeneous media).
My simulation grid is : Nx=220, Ny=250, Nz=250 and dx=dy=dz=0.0002.
When I compare temperature field after 10 seconds of heating with soft tissues in all the grid (homogeneous media) and temperature field after 10 seconds of heating with exactly the same parameters except a layer of fat (heterogeneous media), a difference around 10 degrees is observed. Indeed, the temperature reached at focal point in homogeneous media is around 10 degrees higher than the temperature reached in heterogeneous media. This difference between the two simulations seems to be very high.
Do you think that an error could explain this big gap? Or do you think that it is possible?
Maybe do you know if there is an article or a document which explains the impact of adding heterogeneity in k-Wave simulation?
Thanks for help,
lbruju
Hi lbruju,
The function kWaveDiffusion is valid for heterogeneous media. As with any simulation, you should always run a s convergence test to make sure the simulation result is accurate (e.g., keep reducing dt to see if you get the same answer).
Is the fat layer outside of the beam path, i.e., would you expect it not to have any effect from a heating perspective? If so, and a convergence test doesn't solve the problem, then it sounds suspicious.
The surface pressure to start with will depend on the exact properties of your transducer, in particular, the frequency and the focusing gain, and the properties of the medium, in particular, the absorption coefficient. I'd start with a focal pressure of a few MPa, and then adjust accordingly.
Brad.
Hi Bradley,
I have tested the simulation reducing the dt (from 4.2e-8s to 3e-8s) and I obtain the same results.
The layer of fat is inside the beam path. My computational grid is 220x250x250, and the layer of fat is defined by : soundspeed[50:120,:,:]=1478.0 (same definition for density, alpha_coeff, specific heat capacity and thermal conductivity). So, I expect that this layer has an effect from a heating perspective. But a loss of 10 degrees at focal point (only with this layer of fat) seems wrong...
(The transducer used as a radius of curvature of 35mm and a diameter of aperture of 48mm. It is composed of 16 ring elements. The time varying pressure source is a continuous sinusoidal wave with a frequency of 1.1MHz, and I have chosen an initial amplitude of 0.1MPa that seems to be acceptable.
Thanks for help,
lbruju
You must log in to post.