Convert 3D x,y,z locations in brainmaps volumes to segmentation ids

Usage

brainmaps_xyz2id(
  xyz,
  volume = getOption("fafbseg.skeletonuri"),
  rawcoords = FALSE,
  rawvoxdims = c(8, 8, 40),
  chunksize = getOption("fafbseg.brainmaps_xyz2id.chunksize", 200),
  ...
)

Arguments

xyz: N x 3 matrix of points or an object containing vertex data that is compatible with xyzmatrix. These should be in physical space (i.e. nm) unless voxdims=NULL.
volume: character vector identifier string for the volume containing segmentation data - by default it uses the value of the fafbseg.skeletonuri option. Any input that can be parsed by brainmaps_volume is acceptable.
rawcoords: Whether the coordinates are voxel indices (when rawcoords=TRUE) or physical units (nm, the default).
rawvoxdims: the implied voxel dimensions for the volume. If rawcoords=TRUE then this will be used to convert the raw coordinates into physical units. The default value matches the normal voxel size in neuroglancer. If rawvoxdims=NULL then no attempt is made to scale the coordinates whatsoever. See details.
chunksize: send queries in batches each of which has at most chunksize points. The default is chosen since the brainmaps API can time out if the points take too long to map (more likely if they are spread out across the brain). There is also a maximum number of points per call (10,000 was the recommended upper limit at one point).
...: Additional arguments passed to brainmaps_fetch. This can include setting the retry argument to >0. See Details and Examples.

Value

A numeric vector of Google segment ids

Details

The underlying brainmaps API expects raw coordinates i.e. voxel indices. This is slightly complicated by the fact that different segmentation, skeleton etc volumes may have different associated voxel dimensions. brainmaps_xyz2id automatically looks up this voxel dimension.

However it may be that you want to pass in raw coordinates from neuroglancer. These will generally be associated with a the resolution of the image data not e.g. skeletons which may have a larger (coarser) voxel size. As a convenience in this situation you can set rawcoords=TRUE.

If you have problems with requests failing sporadically, it may be helpful to know

Google does not keep the underlying data live so there may be some spin-up time.
there is an arbitrary timeout at the Google end (currently ~ 5s)
batching nearby points helps

Using the chunksize argument or the retry argument passed on to brainmaps_fetch can overcome these issues. See the examples.

Examples

if (FALSE) {
# Physical location in nm
brainmaps_xyz2id(c(433368, 168208, 128480))
# Same location as displayed in neuroglancer
brainmaps_xyz2id(c(54171, 21026, 3212), rawcoords=TRUE)

# Raw coodinates for the brainmaps volume in question - don't touch
brainmaps_xyz2id(c(433368, 168208, 128480)/c(32,32,40), rawvoxdims=NULL)

library(elmr)
# get a manually traced neuron (just keep first and only entry in neuronlist)
dl4=read.neurons.catmaid('glomerulus DL4 right')[[1]]
# map every node location to segmentation ids
dl4.segs=brainmaps_xyz2id(dl4)
# remove unmapped locations which get id 0
dl4.segs=setdiff(dl4.segs, 0)
# read in corresponding skeletons
dl4.skels=read_segments2(dl4.segs)
# read in corresponding skeletons after including agglomeration merge groups
dl4.allskels=read_segments2(find_merged_segments(dl4.segs))

## retries / cache / chunksize issues
# set small chunk size
dl4.segs=brainmaps_xyz2id(dl4, chunksize=500)
# use retries in case of failure
dl4.segs=brainmaps_xyz2id(dl4, chunksize=500, retry=3)
# cache successful requests (if you might need to repeat)
dl4.segs=brainmaps_xyz2id(dl4, chunksize=500, retry=3, cache=TRUE)
}