Skip to content

Latest commit

 

History

History
122 lines (86 loc) · 3.44 KB

README.md

File metadata and controls

122 lines (86 loc) · 3.44 KB
Raw

microns-coregistration

microns functional / EM data coregistration

WARNING:

Python native blobs get enabled for DataJoint when api is imported.

Install API

pip install git+https://github.com/cajal/microns-coregistration.git#subdirectory=python/microns-coregistration-api

Install API with specified tag

pip install git+https://github.com/cajal/microns-coregistration.git@TAG#subdirectory=python/microns-coregistration-api

Using the AIBS transform

adapted from https://github.com/alleninstitute/em_coregistration/tree/phase3

from microns_coregistration_api.methods import fetch_aibs_transform, aibs_coreg_transform

There are 8 transforms indexed by transform_id:

1. phase2 	2P->EM 	  spline  (deprecated)
2. phase2 	EM->2P 	  spline  (deprecated)
3. phase2 	2P->EM 	  linear  (deprecated)
4. phase2 	EM->2P 	  linear  (deprecated)
5. phase3 	2P->EM 	  spline
6. phase3 	EM->2P 	  spline
7. phase3 	2P->EM 	  linear
8. phase3 	EM->2P 	  linear

The "full" transform as designed by AIBS is type "spline" (transform_id's 5 and 6). A more rigid version of the transform is type "linear" (transform_id's 7 and 8)

Usage:

Inputs to the EM->2P direction are EM coordinates with voxel resolution (4nm x 4nm x 40nm), and outputs are units of microns in the 2P Stack space (1um x 1um x 1um)

Inputs to the 2P->EM direction are 2P coordinates in 2P stack space with voxel resolution (1um x 1um x 1um), and outputs are units of EM voxels (4nm x 4nm x 40nm)

em_coords_vx = aibs_coreg_transform(tp_coords_um, **fetch_aibs_transform(transform_id=5))
tp_coords_um = aibs_coreg_transform(em_coords_vx, **fetch_aibs_transform(transform_id=6))
em_coords_vx = aibs_coreg_transform(tp_coords_um, **fetch_aibs_transform(transform_id=7))
tp_coords_um = aibs_coreg_transform(em_coords_vx, **fetch_aibs_transform(transform_id=8))

Adjusting EM Coordinates

from microns_coregistration_api.methods import M65Adjust

The following transforms can be sequentially applied to adjust user-provided coordinates in units of nanometers:

1. Rotation

2. Normalization

The default transform performs the following:

1. apply a 3.5 degree rotation about the z-axis

2. solve a linear model for the rotated pia with an included set of 
    manually placed points

3. solve a linear model for the rotated wm with an included set of 
    manually placed points

4. solve for the pia and wm depth of each user-provided coordinate 
    via the models

5. normalize each user-provided y-value according to its pia and wm 
    depth, setting the new min and new max to the mean pia depth and 
    mean wm depth, respectively

To adjust with the default transform:

M65Adjust.default_adjust(points)

To adjust with rotation only, ignoring all defaults:

M65Adjust.rotate_points(points, rotate_z=3.5)

To adjust with user-specified params:

NOTE: Any params not provided by user will use default params

Override default params (Recommended):

tform = M65Adjust(**param_dict)
tform.adjust(points)

Overwrite default params:

M65Adjust.update_default_params(**param_dict)
M65Adjust.default_adjust(points)

Restore default params:

M65Adjust.restore_default_params()

To access the manually placed pia and wm points:

M65Adjust.pia_pts_nm
M65Adjust.wm_pts_nm

For advanced usage see M65Adjust class documentation