Pipeline for 3D Reconstruction of Tau Burden Maps from MRI and Histology
- 3D Printing Cutting Molds
- Step 1. Convert 7T scan to NIFTI
- Step 2. Segment the MTL from background in ITK-SNAP
- Step 3. Create a reference mold
- Step 4. Find the correct rotation of the 7T image
- Step 5. Carve the tissue segmentation out of the mold
- Step 6. Generate Surface Mesh of the Mold
- Step 7. Simplify the STL file for 3D printing
- Step 8. Visualize mold and tissue in ParaView
- Data Organization
- Data Organization (PATCH|Lab PMACS cluster)
- WildCat Training and Whole-Slide Analysis
- Patch download and organization
- Train WildCat classifier in Jupyter
- Evaluate WildCat classifier on test patches
- Evaluate WildCat classifier on whole-slide image
- Validate WildCat classifier vs. Manual Ratings and Object Counts
- Training WildCat network from the command line
- Applying WildCat models in batch mode
- Whole-Slide Histology Processing
- Managing Histology Tasks with `svs_to_cloud.sh`
- Checking for New Scanned Slides (
check_for_new_slides_all) - Updating Histology Manifest Files (
update_histo_match_manifest_all) - Preprocessing Histology Slides (
preprocess_slides_all) - Running WildCat in GCP (
density_map_all) - Extracting NISSL DeepCluster feature maps (
nissl_multichannel_all) - Importing histology-derived maps (
rsync_histo_all) - Troubleshooting:
- Checking for New Scanned Slides (
- Running 3D Reconstruction
- Check the inputs
- Generate masks for histology
- MRI processing (7T to 9.4T registration)
- Run blockface reconstruction
- Training a custom random forest classifier
- Compute initial blockface to MRI matching
- Run blockface to MRI registration
- Histology Mask Generation
- MRI-Guided Nissl 3D Reconstruction
- IHC to Nissl Registration
- Generating 3D WildCat density reconstructions
- Tips and Tricks