DeepSSM Test Systems

[jadie1]

Create datasets that can be used for consistent DeepSSM benchmarking.

Training data must include:

• Original and groomed images
• Original and groomed shapes (meshes or binary segmentations)
• Cutting planes and landmarks if available
• Transforms (grooming/alignment as well as procrustes if used)
• Local and world particles (where local are in the original coordinate system and world are in the aligned coordinate system)
• PCA data and sample-wise scores/loadings
• Augmented data (generated groomed images, world particles, and PCA loadings)
• Save reference used for alignment and bounding box used for image cropping

Testing and Validation data includes:

• Original images
• Groomed images using image-to-image registration
• Ground truth particles from fixed domain optimization

Datasets to be added include: In-House:

• Left Atrium
• Femur
• Shoulder Public:
• Med seg pancreas

We can add to this list as we go.

[jadie1]

I have created a folder for this here: /usr/sci/datanew/SSM-Data/DeepSSM/ and added the left atrium