Scripts for converting TCIA CT-Lymph-Node dataset to dicom
Collection is here: https://wiki.cancerimagingarchive.net/display/Public/CT+Lymph+Nodes
Goal: convert masks for use as training set.
Status:
- Source CT data is in DICOM
- Masks (volumetric segmentations) exist in nifti format
- Some additional annotations are in MITK format
Conversion ideas:
- Convert all masks to DICOM SEG linked to corresponding series
- Load full collection into dedicated Google DICOM store for viewing in OHIF or Slicer
- Use dcmqi to do the nifti to dicom conversion
- Re-export the CTs with better window/level settings and experiment with compression
- Create MRB files preloaded with segmentations
Open issues:
- What colors should we use so the seg outlines are easily visible
- What naming conventions should we use in the SEG (picking the right dicom terminology)
- What anatomy labels should we use and can we determine them automatically, or do we need to manually assign them
- Are there smoothness or other geometric constratints we can apply to clean up the segmentations