Postprocessing cochlea segmentation using erosion

[schilling40]

Functions for post-processing the SGN segmentation of cochleas using erosion with spatial statistics.

WIP

• edit script for post-processing
• more flexible spatial statistics

[constantinpape]

Looks good for now, we should add the option for graph connected components, as you already marked down in the TODO

[constantinpape]

Also, I think it would make sense if we order the SGN components by size, so that 1 is always the largest component etc.

[schilling40]

The current implementation of the graph connected component analysis seems to have some error.
After post-processing the eroded subset, the connected components of the graph feature nodes, which appear to be further distanced from the main cochlea segmentation when visioned not along the x-axis.
I am still trying to find out the issue and resolve it.

[schilling40]

The current implementation of the graph connected component analysis seems to have some error. [...]

The indexing of the 'component_labels' was offset by 1 because of the 'label_id' starting at 1.

[constantinpape]

Looks good. I didn't fully check the graph component logic, but it makes sense on a quick glance. I pushed a commit to fix the test errors. Anything else to add here from your end @schilling40 ? Otherwise I will merge it.

[schilling40]

Looks good. I didn't fully check the graph component logic, but it makes sense on a quick glance. I pushed a commit to fix the test errors. Anything else to add here from your end @schilling40 ? Otherwise I will merge it.

Thanks! From my end, there is nothing to add. I am going ahead and merge it.