search

QIICR / QuantitativeReporting

Segmentation-based measurements with DICOM import and export of the results.
https://qiicr.gitbooks.io/quantitativereporting-guide
Other
23 stars 23 forks source link

ENH: Add AI result SR support #260

Open pieper opened 3 years ago

pieper commented 3 years ago

This parses specific TID 1500 AI results that contain a rectangle and a score for each detection.

It is fairly special-purpose but make this code a more general than the previous version that only handled lines. Longer term a more general approach will be needed, probably by incorporating highdicom as a Slicer dependency.

lassoan commented 3 years ago

It is great that you use markups line (and not rulers anymore)!

Are you creating a model node for ROIs? We have not markups ROI, which should work much better. It supports automatic and custom (static) measurements with proper DICOM units and terminology support, so with that you can do lossless roundtrip of DICOM import/export, etc. Let me know if you have any questions, or we can discuss at the next developer meeting.

pieper commented 3 years ago

Yes, we would prefer markup ROIs but there were 1029 boxes in the SR I was working with and that seemed unworkable until this issue is resolved. This is all due for a revamp once we start using highdicom instead of all this manual parsing.

lassoan commented 3 years ago

As a quick workaround, we could turn off interaction and always (or never) show labels, so that we don't need to keep querying the z buffer. I've asked @Sunderlandkyl to investigate. It would be really bad if after all the invested efforts, markups could not be used because of this performance problem.

fedorov commented 3 years ago

@Sunderlandkyl any updates?

fedorov commented 3 years ago

@pieper does it make sense to consider support of those planar annotation types in this PR as well: https://github.com/OHIF/Viewers/issues/1215#issuecomment-854237685