Reorient Image to Canonical (RAS+) Orientation

Reorients a neuroimaging volume or vector to the canonical RAS+ (Right-Anterior-Superior) orientation by permuting and flipping axes. This is equivalent to nibabel's as_closest_canonical().

Usage

as_canonical(x, target = c("R", "A", "S"))

Arguments

x

A NeuroVol or NeuroVec object.

target

Character vector of length 3 giving the desired orientation. Default is c("R", "A", "S") (RAS+).

Value

A reoriented object of the same class as x.

Details

Reorienting to a target that differs only in which anatomical direction each axis runs is a relabelling of the grid, not a resampling: the voxels are the same voxels, permuted and flipped. This function does exactly that, so it is exact, works on images of any size, and costs a single aperm().

It used to build the target space and hand it to resample, which had three consequences: values were interpolated where they should merely have moved, the registration backend refused images smaller than four voxels on any axis, and – because the target space kept the source's dimensions – an axis-permuting reorientation wrote into a grid that did not contain the data and silently lost about a quarter of it.

See also

axcodes, reorient, apply_orientation

Examples

sp <- NeuroSpace(c(10L, 10L, 10L), c(2, 2, 2))
vol <- DenseNeuroVol(array(rnorm(1000), c(10,10,10)), sp)
ras_vol <- as_canonical(vol)
axcodes(ras_vol)  # "R" "A" "S"
#> [1] "R" "A" "S"