Open ckolbPTB opened 5 months ago
Why does it fail? I thought it would just be slow as it would use the nufft if the data is unsorted?
Did we break something there?
True, it won't fail but it will reconstruct an incorrect image. If the data is unsorted our logic will still recognise that the points lie on a Cartesian grid and apply a fft but as the points are not sorted into the grid properly the output will not be correct.
This is related to #228 but so far I could not come up with any idea of how to detect when the warning message suggest in #228 needs to be provided.
We could think of always creating and applying a CartesianSamplingOp in FourierOp. Or creating a new Operator which combines both? Open to suggestions...
Currently the DirectReconstruction does not utilise the CartesianSamplingOp