Hi everyone,
I wanted some help or sample codes that could help me out in setting up raw data for use in reconstruction algorithms like time reversal or full wave inversion. I specifically want to know how I should go about choosing the imaging region size. I am using the k-Wave transducer defining function for defining the transducer.
Any help would be hugely appreciated. Thanks!
Just to add to this, I have the data in the time series form where each sensor element has one time series set of data to it. in the kgrid struct, I have hard-coded the values of Nt and dt as per the RF data and the sampling time interval of the data acquisition. The problem is that if I use the exact same size of imaging region used in the data acquisition, the reconstruction becomes incredibly CPU intensive and doesn't even begin to run on my laptop. I tried using a region which was the same as the prime factors of my original region but the data I seem to get largely just seems to be noise. I do see the wave propagating out back from the transducer elements but it just keeps going on and hits the wall
Hi Mahaan,
There are lots of examples of running simulations included in the toolbox (see the MATLAB help browser). Unfortunately, if your domain is large and you want to model higher frequencies, the grid size will be large and the computations will be intensive. Are you using the C++/CUDA codes? These can help a lot.
Brad.
Hi Brad,
Thanks so much for the reply! Yes, I have worked with the simulations before and I have no problem working with those. Since I am working with experimentally obtained RF data now, I was wondering if I need to have the exact same grid size for me to get a comparable result since now I see the wave propagating on and essentially just leave the imaging grid.
Would it help me if I choose my t_end in such a way that I know my time reversed time series wave is able to make it back to the location it propagated out from? I am choosing the delta t variable as the sampling time of the data acquisition system.
I just re-read my initial post and realised it is easy to be confused thinking that I am asking for help with simulated time series data. I am sorry about that, I am actually using experimentally acquired data and trying to reconstruct that with time reversal on k-Wave.
the grid spacing the inbuilt reconstruction algorithm our DAQ uses, and the imaging region it’s choosing is too large for k-wave to work with. However, I thought that since we are using time reversal all we would need is the right time sample step, the speed of sound and the an adequate number of time steps with respect to the imaging region being large enough to ensure that the wave propagates out to the correct location by the end of the time reversal, regardless of the imaging region of the DAQ
You will need to make the simulation domain large enough to capture both the position of the transducer elements (from which the signals will be emitted during time reversal), and the region of interest. Assuming you are solving an inverse initial-value problem (like photoacoustics), the simulation t_end should be exactly the length of time you recorded the signals for (e.g., 0 should be when the laser pulse was emitted).
You must log in to post.