Hi,
I am working in the field of photoacoustics and K-Wave has helped a lot. But I have some quires related to K-Wave. They are as follows:
1. What is the specification(center frequency and bandwidth)of the sensor used in example'2D Time Reversal Reconstruction For A Circular Sensor Example'?
2. How to choose the number of sensor points for reconstruction?
Thanks and regards,
Deblina
Hi Deblina,
1. In this example, the detectors are broadband point detectors, meaning no filtering is applied to the recorded signals. If you want to add this, you could use the function gaussianFilter to filter the recorded sensor data after the forward simulation.
2. Do you mean in a numerical simulation, or for experimental data? Numerically, this will depend on what it is you're trying to simulate. Experimentally, factors such as the physical element size, sensitivity, cost, frequency range, etc will also come into play.
Brad.
Hi Brad,
Thanks for the reply. But still I have some doubts.In case of sensor, does broadband corresponds to infinite bandwidth? Does it anyhow related to 'maximum supported frequency' or any other parameters? And for number of sensor points, if I want to simulate a circular disc (for example 1 mm diameter using 'makedisc') then how do I choose the number of sensor points?
Hi Deblina,
Regarding the bandwidth of the detector, it is broadband in the sense that no numerical filtering is applied to the recorded signals. However, for all numerical simulations, the range of frequencies will necessarily be band-limited due to the discretisation. Using k-Wave with the default CFL number, the range of recorded frequencies will be dictated by the spatial grid spacing, where f_max = c0 / (2 * dx). You might find it helpful to refer to the k-Wave Manual for some more detailed information.
As for the number of sensor points, it really does depends on what you are trying to achieve with your simulation. If you are trying to replicate an experimental system, you might want to use the same number in your simulation. If you want to record full-view data, then perhaps you want to use a continuous sensor mask (e.g., using makeCircle).
Good luck with your simulations,
Brad.
Hi Brad,
Thanks so much for the information. Its really helpful to understand certain things. And what I understand from your reply is the sensor has infinite bandwidth but the detected frequency is limited by the maximum supported frequency.
Regarding the number of sensor points, I am not trying to plug in any experimental data. I am just trying to simulate a circular target of 1 mm radius with different number of sensor points (37, 100, 150,200, 250, 300). For that I am using '2D Time Reversal Reconstruction For A Circular Sensor Example' example. The following are the parameters used for simulation as well as reconstruction:
Nx= Ny=700
dx = 0.1 mm
Number of sensor points = 37, 100, 150,200, 250, 300
Sensor_radius = 3e-2 (3 cm)
Sensor_angle = 4pi/2
and it was observed that the size of the reconstructed image increased with the increase in number of sensor points. And with 300 sensor points the size is smaller than the original target. Just wanted to know what could be the probable reasons for this?
Thanks
Deblina
Hi Deblina,
Within k-Wave, the sensor is simply storing the values of the acoustic pressure at particular grid points over time as the simulation progresses. Thus the sensor itself doesn't have a bandwidth per se, but the frequencies you detect are intrinsically dependent on your grid parameters.
Regarding your observation, if your sensor doesn't completely enclose the imaging region, not all of the acoustic energy will be recorded, so your reconstructed amplitude will be less than the original. In a simplified sense, the more sensor points you have, the greater the proportion of recorded energy, and thus the greater the reconstructed amplitude. For comparison, you could try creating a continuous binary sensor mask using makeCircle rather than makeCartCircle.
Brad.
Hi Brad,
Thanks for your suggestion. I will try and get back to you.
Thanks and regards
Deblina
Hello Brad,
I have some more doubts regarding 2D and 3D simulations
1. I have gone through the following link:
http://www.k-wave.org/documentation/example_ivp_photoacoustic_waveforms.php
I want to know that the waves from 2D simulation are spherical or cylindrical.
2. In our experiments, we are using a circular source which produces circular waveform. So for reconstruction,which one will be suitable among 2D program(2D Time Reversal Reconstruction For A Circular Sensor Example)and 3D program(3D Time Reversal Reconstruction For A Spherical Sensor Example)?
Thanks and regards
Deblina
Hi Deblina,
1. The waves from a 2D simulation will be cylindrical, not spherical. A 2D simulation in (x,y) with a point source will give the same result as a 3D simulation in (x,y,z) with a line source parallel to the z axis. In other words, if there is no variation in one of the dimensions, there is no need to include it in the simulation.
2. If you are measuring over a circle then the 2D reconstruction will be what you want.
Hope that helps,
Ben
Hi Ben and Brad,
This post is giving many useful information. Thanks a bunch!!
If am using a circular sample (in an experiment) and I am having a scanning sensor in circular fashion, then you suggest using a 2D reconstruction in the above post. I have a feeling that the PA waves generated are actually radially expanding outwards. In other words, the amplitude of the PA signal would reduce over 1/r (where r is the distance between the sample and the sensor) rather than 1/sqrt(r). Then even though the experimental situation represents a 2D system, using the 2D reconstruction would allow the waves to expand in a cylindrical fashion. Is it correct to use 2D reconstruction / 3D reconstruction?
Thanks for clarifying our doubts!! K-wave is an amazing tool and we are enjoying using it.
Hi srivdevan,
Actual experiments are, of course, taking place in 3D, so the waves are propagating in 3D, but there are two ways to make them approximately 2D:
(1) if there is symmetry in one dimension (ie there is no variation with z, say) then the propagation can be modelled as 2D (x,y), as the waves will look the same for all values of z.
(2) if the detectors are tightly focussed in the plane then 2D will also be approximately correct. ie. for detectors positioned around a circle, the detectors need to be focussed in the plane in which the circle lies.
Hope that helps. Glad you're enjoying k-Wave!
Ben
Hello Ben,
Thank you so much for your clarification regarding 2D and 3D wave propagation. I understand that for experimental conditions, even though the generated waves propagate in 3D but it can be approximated as 2D wave considering mentioned approximations.
Now my confusion is, in the '2D Time Reversal Reconstruction For A Circular Sensor Example' program the waves generated are cylindrical waves which are used for reconstruction. Now in the same program i.e. '2D Time Reversal Reconstruction For A Circular Sensor Example' if spherical waves are feed for reconstruction, will it give a proper reconstruction or some modifications should be made in the 2D program so that it gives proper reconstruction for spherical waves as well.
Best regards,
Deblina
Hi Deblina,
If you use a 2D reconstruction algorithm on 3D data it will not work, unless, as mentioned above, you have measured the 3D wavefield using a transducer focussed in the plane of interest, so it rejects out-of-plane parts of the signal.
In general , the 2D and 3D situations are quite different; the 3D Green's function for the wave equation is different - even in character - to the 2D Green's function.
Ben
Dear Ben,
Thanks so much for the clarification.
Regards,
Deblina
You must log in to post.