k-Wave

1. Srijeeta
   

   Member
   Joined: Apr '12
   Posts: 6

   Hi Dr. Treeby and Dr. Cox,

   I am trying to get back reconstructed images from time-domain pressure data acquired using a curved array sensor. In order to do so, I have modified 'example_pr_2D_TR_circular_sensor.m' to use the experimental data and respective sensor locations. Could you please help me to sort out the following error?

   Error in ==> kspaceFirstOrder_inputChecking at 227
   sensor.time_reversal_boundary_data(:, new_col_pos) = order_index;

   Error in ==> kspaceFirstOrder2D at 427
   kspaceFirstOrder_inputChecking;

   The data in 'sensor.time_reversal_boundary_data' can be accessed as (sensor,time) and I have specified the sensor locations in Cartesian coordinates. Could I be going wrong with the sensor mask then?

   Thanks,
   Srijeeta

   Posted 12 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Srijeeta,

   Without seeing your code, it's difficult to say exactly what might be going wrong. You could try double checking that your experimental data and sensor mask have the same number of points, and that they are indexed in the correct way (e.g., via whos) using the example file as reference. Another thing to check is whether your Cartesian sensor points are overlapping in discrete space. You could try plotting the sensor mask as k-Wave sees it using imagesc and cart2grid. If any of your sensor points are overlapping so they are not unique when mapped to the discrete grid, you'll need to use a finer grid spacing.

   Kind regards,

   Brad.

   Posted 12 years ago #

3. Srijeeta
   

   Member
   Joined: Apr '12
   Posts: 6

   Hi Dr. Treeby,

   Thank you very much for the suggestions. The sensor locations were indeed overlapping and using a finer grid solved the problem.

   However, I am still unable to get a reconstruction using the experimental data which otherwise gives a reconstruction with the backprojection algorithm. Could you please tell me the unit for pressure used in the toolbox codes?

   Also, could I use Cartesian sensor locations in the example for reconstruction from bandlimited sensors?

   Kind regards,
   Srijeeta

   Posted 12 years ago #

4. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Srijeeta,

   I am not sure what you mean when you say you don't get a reconstruction? Do you get an error or just something that doesn't look correct? Another thing to double check is that your array of time points (kgrid.t_array) is constructed to match your experimental acquisition settings. The pressure is in units of Pascals.

   You could change the bandlimited sensors example to use a Cartesian sensor mask. Note that this example just shows the effect of using a bandlimited sensor, it doesn't correct for this behaviour.

   Brad.

   Posted 12 years ago #

5. Srijeeta
   

   Member
   Joined: Apr '12
   Posts: 6

   Hi Dr. Treeby,

   The reconstructed image I get back after time reversal is solely noise. I believe I am going wrong somewhere.

   I used the experimental time points to create 'kgrid.t_array' instead of calling 'makeTime'. The data was acquired at a sampling frequency of 80 MHz, so I use dt = 1.25e-8 sec with 4096 time points. Based on the Courant stability criterion, dt = 1.5664e-8 sec for an area of 40.1x40.1 mm^2 with a grid size of 512x512 points.

   Thank you for clarifying the intent of the bandlimited sensors example. I suppose I could just concentrate on modifying 'example_pr_2D_TR_circular_sensor' for experimental reconstruction.

   I would be grateful if you could provide further insight that would help me obtain a reconstructed image from the noisy data.

   Srijeeta

   Posted 12 years ago #

6. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Srijeeta,

   I'm not sure what else to suggest. When you run the reconstruction, do you see the waves propagating from the sensor positions into the domain? (Try setting 'PlotScale' to 'auto' if you're unsure about the pressure magnitudes.) You could also try filtering out the high frequency information in your signals that is not supported by the grid (for example, using filterTimeSeries). Are you using attenuation compensation?

   Brad.

   Posted 12 years ago #

7. Srijeeta
   

   Member
   Joined: Apr '12
   Posts: 6

   Hi Dr. Treeby,

   Yes, I do see the waves propagating from the sensor positions in circles but the final image does not show any build up at the inclusion locations. Thank you for suggesting 'filterTimeSeries', the measurements appear much less noisier now. However, the reconstruction still does not show much promise. I am not using attenuation compensation presently. Would it be better to do so?

   Kind regards,
   Srijeeta

   Posted 12 years ago #

8. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Srijeeta,

   I am not sure what else to suggest other than double checking that the grid size is sufficient to support the frequencies you are interested in, that the signals are filtered to only contain frequencies supported by the grid, and that the sensor positions, time step size, and sound speed match those used in your experiment. Probably best not to worry about attenuation compensation until you've got it working. Are you using interpCartData?

   Brad.

   Posted 12 years ago #

9. Srijeeta
   

   Member
   Joined: Apr '12
   Posts: 6

   Hi Brad,

   The transducer used to record measurements has a centre frequency of 6.25 MHz with a bandwidth slightly higher than 80%. So I am using a grid that allows frequencies till 12-13 MHz. I have hard-coded the sensor positions and time step size into the code instead of using the functions to generate them, and also specified the sound speed. I am using 'interpCartData' to interpolate the Cartesian coordinates of the sensor positions. Is that something I should be worrying about?

   I am still hoping I can make the k-Wave toolbox work for my problem because it is difficult to find an alternative reconstruction algorithm for such high frequencies.

   Kind regards,
   Srijeeta

   Posted 12 years ago #

10. Bradley Treeby
    

    Developer
    Joined: Mar '10
    Posts: 1,643

    Hi Srijeeta,

    I am sure we can get it working. Perhaps try first without using interpCartData and just your actual sensor positions defined as a Cartesian sensor mask to see if you can get an image.

    Brad.

    Posted 12 years ago #

11. Srijeeta
    

    Member
    Joined: Apr '12
    Posts: 6

    Hi Brad,

    I am a little confused but I think I have been using the sensor positions defined as a Cartesian sensor mask to run the time-reversal reconstruction. 'interpCartData' is used to interpolate the time-reversal reconstruction data to remove the gaps. Is that where I should modify the mask?

    Srijeeta

    Posted 12 years ago #

12. Bradley Treeby
    

    Developer
    Joined: Mar '10
    Posts: 1,643

    Hi Srijeeta,

    I would start with just using the Cartesian sensor mask and your measured data (without using interpCartData to remove the gaps), and see if you can reconstruct an image.

    Brad.

    Posted 12 years ago #

RSS feed for this topic

Reply

You must log in to post.