In order to check the acurracy of registration, it's useful to overlay the fixed and registed images in an image viewer such as 3D Slicer. We have created our own viewer VPV that we use for this purpose.
Setting the fixed image colormap to cyan and the moving image colormap to red, regions where theer is good registration appear more grey/white.
Fixed popualtion average overalid with registrated moving image
If you have many specimens to register, loading them all into a image vieer can be time-consuming. To facillate faster QC, LAMA has a couple of scripts:
This script will search all subdirectories for LAMA registration directories and put all the qc images it finds there into single html files.
todo: Put this script in the entry ponts in setup.py
It will generate the follwong folders (3 files in each - sagittal.html, coronal.html and axial.html
red_cyan_overlays: The moving image output of each stage overalid onto the fixed image.
inverted_labels: The propagated atlas labels overalid on the input images
greyscales: The moving image output of each stage. No overlay
Eaxmple usage:
python3 lama/qc/collate_qc.py -i lines/baseline/output -o lines/qc/baseline
Html output showing rigidly-aligned specimens on the left with final regisrtered images on the right. One specimen per
row
Html output showing propagated labels overlaid on input moving images. Once specimen per row. Each image is a 2D slice
from evenly sampled from the given orientation (sagittal in this case)
