We have a number of OCT with segmentation data done with legacy systems. I would like to convert the OCT files are DICOM files and they load mostly alright. They are loaded by drag and drop and selecting volume. The .nrrd files that I created are also loaded with drag and drop with segmentation selected. The three surfaces contained in the .nrrd files are displayed. The three surfaces can be viewed and edited in the segmentation editor. The .nrrd file had 4 dimension. A volume for each surface.
We would like to have the segments loaded with a name for each surface and a defined color. I created a .seg.nrrd file. When this file is dragged and dropped into slicer, segmentation is now the default option. But the three surface are not displayed. The segmentation editor does show one row with a default name. The following is the header for the .seg.nrrd file I created. I would like to know what else is needed for the names and colors to load properly. Note, the only difference between the .nrrd header and the .seg.nrrd header are the lines that start with Segment
Thanks
One quick suggestion, might the tags be case-sensitive? On the documentation page it looks like the SegmentN_... tags all start with a capital “S”.
I have permission to send the .seg.nrrd file. But the upload button says that it is not an accepted type. How would you like me to get you this file?
Thanks
Mark
If not, then you can upload the data set anywhere (dropbox, onedrive, …) and post the download link here. You don’t need to upload real data, it can be some scribbles that you draw or some public/sample data.
Note that you can use the slicerio Python package to read segmentation files outside Slicer and I’ve been meaning to add segmentation writing as well. Since you are already developing a script that creates segmentation files, it would be great if you could contribute your script to this package. It does not have to be perfect to be useful, it just has to work and should be simple enough to use.
Hi @lassoan,
I have many nrrd files (2000+) that are all the same 3D spatial area, with binary representations for the presence or absence of different features stored in their 4th dimension. I am hoping to create a single seg.nrrd file with each of the unique features stored in a segment. Is this possible and was there ever anything created to help write segmentation files?
I’d be grateful for your thoughts and guidance on this.
You can write a simple Python script (probably ChatGPT or Copilot can write it for you) that iterates through all the files uses pynrrd to read the header and add the appropriate metadata so that Slicer can read them correctly as segmentation. Slicer supports storing segmentations in a 4D volume. The volume index is referred to as “layer” and within the layer you can specify the meaning of each label value. The full specification of the file format is in the links above.
Hi @lassoan,
Thank you for your prompt reply. I should have prefaced that I am python novice, so I am struggle a bit with the R/python syntax differences already. I’ve been able to successfully load all the volumes within the Slicer app using the python attachment, but I am not sure how to export these loaded volumes as a single nrrd file. I also tried to use ChatGPT to write a script that rewrites the headers of all the nrrd files for segmentation, but it keeps providing a means to rewrite the header of a single nrrd file. I am stuck at the point of taking two distinct nrrd files and saving them as a single object. Sorry to bother you with something trivial and novice, I just am unsure of how to proceed.
If all your volumes have the same geometry (origin, spacing, axis directions, and extents) then you can easily combine multiple volumes into one, using numpy. You can find examples in the Slicer script repository for accessing voxels of a volume as a numpy array.