CCF_Registration_Pipeline

Using Elastix

• Profiling:
  • Environment: AMD Ryzen Threadripper 1950X 16-Core Processor@2.2GHZ
  • System: Ubuntu 21.04
  • Peak Memory Usage: 70000MB (~69GB)
  • Total Time: 3014s (~1h)
  • Experiment Brian:
    • Experiment brain id: 192341_SLA
    • Physical spacing: 0.01mmx0.01mmx0.01mm
    • Image size: 1128x1029x590 (ZYX)
    • Numerical dtype: uint8
  • CCF template:
    • Physical spacing: 0.01mmx0.01mmx0.01mm
    • Image size: 1140x800x1320 (ZYX)
    • Numerical dtype: uint16 for average template, uint32 for annotation
• To run the program:
  • Please use absolute path
  • python pipeline_elastix.py --ref_img_path /path/to/ccf_template.tif --ref_ann_path /path/to/ccf_template_annotation.tif --sub_img_path /path/to/experimental_brain.tif--result_path /path/to/save_results/ --suffix .tif

Using AntsPy

• Profiling:
  • Environment: AMD Ryzen Threadripper 1950X 16-Core Processor@2.2GHZ
  • System: Ubuntu 21.04
  • Peak Memory Usage: 36578MB (~36GB)
  • Total Time: 10902s (~3h)
  • Experiment Brian:
    • Experiment brain id: 192341_SLA
    • Physical spacing: 0.01mmx0.01mmx0.01mm
    • Image size: 1128x1029x590 (ZYX)
    • Numerical dtype: uint8
  • CCF template:
    • Physical spacing: 0.025mmx0.025mmx0.025mm
    • Image size: 456x320x528 (ZYX)
    • Numerical dtype: uint16 for average template, uint32 for annotation
• To run the program:
  • Please use absolute path
  • python pipeline_ants.py --ref_img_path /path/to/ccf_template.nii.gz --ref_ann_path /path/to/ccf_template_annotation.nii.gz --sub_img_path /path/to/experimental_brain.nii.gz --result_path /path/to/save_results/ --suffix .nii.gz

[Results]: Warped ccf_template and ccf_template_annotation images are stored in /path/to/save_results/subject/ants/ccf_affined/

[Important Note]: If you want faster (actually much faster) registration (~20-25mins), you can reduce the iterations of non-linear deformation. This although may suffer from inferior registration performance. Unfortunately, AntsPy doesn't provide an api to modify these paprameters (it has been hardcoded), what you can do is following:

1. first indentify this piece of code elif type_of_transform == "SyNCC": (it's in Line 816 currently for antspyx 0.3.3) in ants.registration.interface.py. You will need to look up from site-packages if you are using anaconda.
2. replace synits = "2100x1200x1200x20" smoothingsigmas = "3x2x1x0" shrinkfactors = "4x3x2x1" with synits = "50x50x0x0" smoothingsigmas = "4x2x1x0" shrinkfactors = "8x4x2x1", or simply replace the file with misc/interface.py (This may subject to version update, be careful).