Hi Dr. Treeby and Dr. Cox
I am using kwave toolbox to generate my data set and use this data for image reconstruction.
I did it by analytical algorithm like back projection.
I want to use modelbased algorithm to reconstruct absorption so for each iteration of model based algorithm i need to simulate forward model to generate pressure signals from corresponding optical absorption.
is it possible to run the forward model for modelbased image reconstruction?
best
Moein Mozaffarzadeh
kWave
A MATLAB toolbox for the timedomain
simulation of acoustic wave fields
modelbased image reconstruction
(5 posts) (3 voices)
Posted 1 year ago #

Hi Moein,
Yes, you can use kWave as a forward model for modelbased image reconstruction, although the most general image reconstruction algorithm incorporated directly into kWave is time reversal. There are some examples describing that. 'Modelbased' image reconstruction is not a very specific term  all image reconstruction techniques are based on models of the acoustic propagation, although they are not always explicitly stated as such. Sometimes it is used to describe an approach where a matrix formulation of the forward model is used in an iterative minimisationbased reconstruction. Whether or not that is feasible depends on the size of the problem you are tackling. A related approach that doesn't use matrices is based on adjoint equations. Anastasio has done some good work on this recently. The adjoint for kWave's forward model is not yet incorporated into the code, but there is a paper that includes a description of how you can do it using the existing toolbox here: arxiv.org/abs/1602.02027
Hope that helps.
BenPosted 1 year ago # 
Dear Dr. Cox
thanks for your help. as you mentioned there are some cases that are working with matrix formulation of the forward model. this is my goal. I want to have the forward matrix model and use iterative minimization approach. how can I have the forward model matrix in kWave? is it even possible?
best
Moein MozaffarzadehPosted 1 year ago # 
Hi Moein,
You could in theory build the matrix laboriously by putting point sources at every point, oneafteranother, and measuring the response everywhere for all times, and putting those measured 'Green's functions' into a matrix. Not sure what the point of that would be though. The matrix would likely be huge. As I mentioned before, Mark Anastasio's group have written a paper on constructing a matrix version of the forward model that kWave is based. See Huang, Chao, et al. "Fullwave iterative image reconstruction in photoacoustic tomography with acoustically inhomogeneous media." IEEE transactions on medical imaging 32.6 (2013): 10971110. There has also been work in this area by the group of Daniel Razansky in Munich. They use a different discrete model but it is matrix based.
Best wishes,
Ben
Posted 1 year ago # 
Moein,
Another article which you might find helpful, also by Anastasio et al., is "An Imaging Model Incorporating Ultrasound Transducer Properties for ThreeDimensional Optoacoustic Tomography," IEEE Transactions on Medical Imaging, Vol. 30, No. 2, Feb 2011.
Best,
OsmanPosted 1 month ago #
Reply
You must log in to post.