search

AIM-Harvard / pyradiomics

Open-source python package for the extraction of Radiomics features from 2D and 3D images and binary masks. Support: https://discourse.slicer.org/c/community/radiomics
http://pyradiomics.readthedocs.io/
BSD 3-Clause "New" or "Revised" License
1.11k stars 485 forks source link

ROI cropping and segmentation #821

Open nimayousefi75 opened 1 year ago

nimayousefi75 commented 1 year ago

Hi, In case of feature extraction from ct images and for any analysis such as prediction, should we crop then segment ROI or just segmentation is enough?

Thanks.

dorkylever commented 1 year ago

So the devs may correct me but I would think it depend on if you're using the z-normalisation (for intensity correction) within the pyradiomics package.

If you are, there would be a difference in the normalisation between the original image or the ROI. So then it would depend on which performs the better normalisation and you should probably test out both.

If you aren't (say you're using different normalisation methods), then the results should be the same, only thing would maybe improved run-time (its easier to perform the diagnostics on small images).

JoostJM commented 1 year ago

Segmentation-only is fine. If you are resampling the spacing using PyRadiomics, this auto-crops the image onto the ROI bounding box (for speed and memory optimization, this happens after any normalization). If no-resampling is enabled, you can achieve a similar performance boost by adding preCrop=true to the settings.