Transform section pattern for some cases of non-auto axis image saver

[yousefmoazzam]

Fixes https://jira.diamond.ac.uk/browse/IMGDA-606

For cases where the axis parameter of the image saver is configured to:

• not be auto
• and the given integer value is not the slicing dim represented by the pattern associated with the section the image saver method is in

then the images saved are likely not what is intended.

For example, consider the following sequence of methods:

• normaliser
• stats calculator (produces side output)
• rescaler (depends on side output of global stats)
• image saver (with axis set to 1)

The fact that the rescaler requires the side output produced by the stats calculator means that a new section is started after the stats calculator. Thus, where will be two sections made from this sequence of methods.

The normaliser will have the projection pattern, and this means that the first section has the projection pattern (which has slicing dimension 0).

The rescaler has the "all" pattern and so does the image saver. This will ultimately result in the second section having the projection pattern, which means that a chunk will be split into blocks of projections.

However, the image saver's setting of axis=1 will result in saving images sliced along axis 1 of each of the blocks of projections, which are something like "partial sinograms" within each block of projections. This is unlikely to be what the user intended, and instead what was probably intended was to save the output of the rescaler as "full sinograms". In order to accomplish this, a reslice operation is needed between the first and second section, and this can be achieved by transforming the pattern for the second section from projection to sinogram.

The changes made in this commit perform such a transformation where relevant.

Note that the above example is one specific case, but more generally, this transformation will be applied to a section containing an image saver method if the image saver is configured with an axis value that is not auto and this value is not the slicing dim represented by the pattern of the image saver's section.

Checklist

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have made corresponding changes to the documentation

[dkazanc]

thanks! So, if I understand correctly, the reslice operation happens only if the user select the right axis in the image saver method. For instance axis=1 will trigger the reslice?

[yousefmoazzam]

thanks! So, if I understand correctly, the reslice operation happens only if the user select the right axis in the image saver method. For instance axis=1 will trigger the reslice?

I'm not sure what "the right axis" means, so I'll defer to enumerating the options in the hope that it answers the question 😅; if the image saver is in a section whose pattern is projection:

• if axis=0, the section pattern stays as projection, so no reslice is triggered
• if axis=1, the section pattern is changed to sinogram, so a reslice to sinograms is triggered

Perhaps a succinct way to describe the behaviour is the following: if the axis value doesn't match the pattern of the section that the image saver is in, a reslice will be triggered.

[dkazanc]

thanks, yes this axis 0,1 is slightly less intuitive as the string-based pattern. But I understand some inconveniences when moving to defining quite a lot of different strings for different cases, like volume_xz, volume_xy, etc. Happy with the changes.

[yousefmoazzam]

The IRIS GPU runners are still experiencing issues, but the framework and small data tests pass locally, so I'll go ahead and merge this.