Integrate SegmentEditor

[che85]

This is an attempt to get rid of the "old" Editor dependency and adding a dependency to the "new" SegmentEditor.

@fedorov let me know if you have anything on your mind, that I might be missing while thinking about how to integrate it with mpReview.

What do you think? Of which functionalities can we get rid with implementing this?

[fedorov]

For now, I would just get rid of the functionalities that would duplicate what SegmentEditor already provides (if there are any). The main goal of this update is to improve maintainability of the code, and harmonize with SliceTracker. It will also be easier to add DICOM export functionality, since we will have segment terminology support, and the process of exporting can be factored out between QuantitativeReporting and this module.

The only thing I can think of that we will lose is window/level effect with the rectangle ROI. But maybe we can add that to the w/l functionality you added in SliceTracker, and reuse that as a common component.

[che85]

We are not using that anyway anymore.

window/level effect with the rectangle ROI

[che85]

I meant. We are using the WL effect, but not with the rectangle ROI. The one that we are using in SliceTracker is implemented in SlicerProstate https://github.com/SlicerProstate/SlicerProstate/blob/master/SlicerProstate/SlicerProstateUtils/buttons.py#L195

Would be extremely simple to integrate

[fedorov]

Very good! Yes, I know we are not using it, but when we add PK maps, we will need it back!

[che85]

Oh I see. Well maybe we can have something similar to the button I referenced with the rectangle ROI functionality

[mehrtash]

While the new SegmentEditor's functionalities are much superior to the old one I experienced some bugs during our last project. we tried to use it in RSNA project (50 segmentations) and for some cases on Windows, it failed to segment so we were forced to switch back to the old one. Also, I'm not sure if (Simple)ITK readers support the segmentation objects (filename.seg.nrrd). If not, you probably need to create a copy of labelmaps when saving them.

[fedorov]

I experienced some bugs during our last project

@mehrtash was that issue with SegmentEditor reproducible? Did you look at the log? Did you use the standard or fractional segmentation when this was happening? It would be good to provide this feedback to the Queen's team. We did not experience this kind of issues, but also we did not yet use it on a large dataset.

I'm not sure if (Simple)ITK readers support the segmentation objects

We can save each segment into a separate file, as we do in QuantitativeReporting, in addition to saving the mult-segment seg.nrrd.

[mehrtash]

@fedorov It was reproducible for some cases in windows. I didn't check the log. It was using the standard segmentation. I'll take a look at into those cases again and see if I can find anything in log.

[fedorov]

Thank you Alireza! I think @lassoan and @cpinter would appreciate if you had a reproducible way to crash SegmentEditor, so that they can fix it.

[lassoan]

some cases on Windows, it failed to segment

Please send us the application log (you can access logs of previous sessions in menu: Help/Report a bug) and step-by-step instructions how to reproduce the problem, preferably using sample data sets included in Slicer.

I'm not sure if (Simple)ITK readers support the segmentation objects (filename.seg.nrrd).

Yes, both standard ITK and VTK readers can read the saved segmentation file. It's a simple 4D image, each 3D volume is one segment. If segments are not overlapping you can convert them to a merged labelmap without losing any information and save as a 3D volume.