Time Reversal Query

[elma16]

Hi,

I was having a look at the time reversal script in time_reversal.py and I wanted to clarify my understanding. From what I can gather, the measurement data is flipped in time, and fed in as dirichlet sources. Isn't this the adjoint PAT operator, rather than the time reversal PAT operator [1]? It's not totally clear to me what the difference means in physical sense, but from what I can tell, this amounts to using the time_reversal_boundary_data attribute of the kSensor - but correct me if I'm wrong!

Thanks for all the amazing work in this project!

[1] Arridge, S.R., Betcke, M.M., Cox, B.T., Lucka, F. and Treeby, B.E., 2016. On the adjoint operator in photoacoustic tomography. Inverse Problems, 32(11), p.115012.

[djps]

I have had a quick look at the paper, and yes, technically what you describe is correct: by using only the recorded sensor data, rather than a closed sufficiently smooth surface, it is like the adjoint, rather than the time reversal operator.

I am not an expert in photoacoustics, but I think often time reversal is used to describe both operators. But note:

If you want to see the difference between the adjoint operator and the time reversal operator, see
https://github.com/djps/k-wave-python/tree/expand_interp_cart_grid_testing/examples/pr_2D_circular_sensor
in which a discrete approximation of the time reversal operator is created (the third simulation), and compared against the adjoint operator (the second simulation). However, both use the TimeReversal class as the simulating backwards in time method.

Hope this clarifies things somewhat.