http://www.k-wave.org/forum/topic/about-adjoint-operator-of-wave-propagation Support for the k-Wave MATLAB toolbox en-US Wed, 13 May 2026 00:22:12 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/about-adjoint-operator-of-wave-propagation#post-8412 Thu, 06 Jan 2022 17:22:37 +0000 bencox 8412@http://www.k-wave.org/forum/ <p>Hi Refik, </p> <p>Thanks for flagging that up. We'll put a corrected version in the next release of the k-Wave toolbox (thanks to Felix Lucka). I guess the missing scaling factor comes out in the wash if you are using a gradient descent-type method with a chosen step-size for the reconstruction.</p> <p>Best wishes<br /> Ben </p> http://www.k-wave.org/forum/topic/about-adjoint-operator-of-wave-propagation#post-8316 Sat, 25 Sep 2021 05:28:14 +0000 rcam2 8316@http://www.k-wave.org/forum/ <p>Hello,</p> <p>First of all, thank you for your efforts to develop and share the k-wave toolbox!</p> <p>I am working on an iterative reconstruction method for Photoacoustic tomography, and I would like to use k-wave for simulation of forward and adjoint operators. However, I am having trouble with the implementation of adjoint operator. I have read your paper named "On the adjoint operator in photoacoustic tomography" and also looked at the example Matlab file for Iterative Image Reconstruction Using The Adjoint. In the example file and the paper, you state that the adjoint model is just the forward model with the source suitably defined.</p> <p>I am literally using the code snippet below from the example file. However, when I apply the inner product test to validate the adjoint, the results of the inner products are very different than each others. One inner product is more than three times the other, so I am not talking about a precision error. Could you help me understand more about how to implement the adjoint operator?</p> <p>% set the source type to act as an adjoint source<br /> source.p_mode = 'additive';</p> <p>% calculate the difference between the measured and modelled data<br /> difference = modelled_time_series - sensor_data;</p> <p>% assign the difference time series as an adjoint source<br /> % (see Appendix B in Arridge et al. Inverse Problems 32, 115012 (2016))<br /> time_reversed_data = fliplr(difference);<br /> source.p = [time_reversed_data(:, 1), time_reversed_data(:, 1), time_reversed_data(:, 1:end-1)] + ...<br /> [zeros(size(time_reversed_data(:, 1))), time_reversed_data(:, 1:end-1), 2 * time_reversed_data(:, end)];</p> <p>Best regards,<br /> Refik Mert Cam </p>