Hemi brain orientations - saalfeldlab/flyem-alignment GitHub Wiki

This page describes the different orientations of the hemibrain dataset.

Dvid

The Dvid dataset will be the "standard" / "canonical" orientation of the dataset.

These synapses are in dvid physical space

• /groups/saalfeld/public/flyem_tbars/tbar_render_20190304.nrrd

Pixel

As described here, the dataset is contained in the pixel interval with

min = [0, 0, 0] and

max = [34427, 39725, 41394].

Physical

The volume has pixel spacing 8 x 8 x 8nm, and as a result, points from DvidPixel can be transformed to points in DvidPhysical(um) by applying:

0.008, 0.0, 0.0, 0.0
0.0, 0.008, 0.0, 0.0
0.0, 0.0, 0.008, 0.0
0.0, 0.0, 0.0, 1.0

N5

The n5 export of the hemibrain here:

/nrs/flyem/data/tmp/Z0115-22.export.n5/22-34

can be transformed to Dvid pixel space using the affine transform below. This transforms points from Dvid pixel space to N5 space:

0.0, 0.0, -1.0, 34427 
1.0, 0.0, 0.0, 0.0
0.0, 1.0, 0.0, 0.0
0.0, 0.0, 0.0, 1.0

To registration space

It was helpful to reorient the volume for registration so that the cardinal direction correspond approximately to the target template (JRC2018 in this case).

The reoriented volume is located here:

/groups/saalfeld/public/flyem_tbars/tbar_render_20190304_reslice.nrrd

I'll call this "registration space"

A transform from N5 pixel space to registration physical space is:

0.0, 0.0, -0.008, 275.416
0.008, 0.0, 0.0, 0.0,
0.0, 0.008, 0.0, 0.0

Affine to JRC2018F

An affine that does a decent job transporting DvidPhysical(um) to JRC2018F is :

1.1764000836511492, 0.055155343055786864, 0.03538023618835689, -170.52475089080767
-0.06646000404339514, -0.41394144685829215, -1.4953538748443649, 395.24695439807454 
0.07849688248531729, 1.4017401676798784, -0.4531492491797921, 41.30281469989984
0.0, 0.0, 0.0, 1.0

Specifically, it transforms points in JRC2018F to DvidPhysical(um).