Reading a NIFTI formatted image volume
The way to read an volumetric image file is to use
read_vol
:
library(neuroim2)
#> Loading required package: Matrix
#>
#> Attaching package: 'neuroim2'
#> The following object is masked from 'package:base':
#>
#> scale
file_name <- system.file("extdata", "global_mask2.nii.gz", package="neuroim2")
vol <- read_vol(file_name)
Working with image volumes
Information about the geometry of the image volume is shown here:
print(vol)
#>
#> === NeuroVol Object ===
#>
#> * Basic Information
#> Type: DenseNeuroVol
#> Dimensions: 64 x 64 x 25 (406.6 Kb)
#> Total Voxels: 102,400
#>
#> * Data Properties
#> Value Range: [0.00, 1.00]
#>
#> * Spatial Properties
#> Spacing: 3.50 x 3.50 x 3.70 mm
#> Origin: 112.0, -108.5, -46.2 mm
#> Axes: Right-to-Left x Posterior-to-Anterior x Inferior-to-Superior
#>
#> ======================================
#>
#> Access Methods:
#> . Get Slice: slice(object, zlevel=10)
#> . Get Value: object[i, j, k]
#> . Plot: plot(object) # shows multiple slices
read_vol
returns an object of class
NeuroVol
object which extends an R array
and
has 3 dimensions (x,y,z).
class(vol)
#> [1] "DenseNeuroVol"
#> attr(,"package")
#> [1] "neuroim2"
is.array(vol)
#> [1] TRUE
dim(vol)
#> [1] 64 64 25
vol[1,1,1]
#> [1] 0
vol[64,64,24]
#> [1] 0
Arithmetic can be performed on images as if they were ordinary
array
s:
vol2 <- vol + vol
sum(vol2) == 2 * sum(vol)
#> [1] TRUE
vol3 <- vol2 - 2*vol
all(vol3 == 0)
#> [1] TRUE
A numeric image volume can be converted to a binary image as follows:
vol2 <- as.logical(vol)
class(vol2)
#> [1] "LogicalNeuroVol"
#> attr(,"package")
#> [1] "neuroim2"
print(vol2[1,1,1])
#> [1] FALSE
We can also create a NeuroVol
instance from an
array
or numeric
vector. First we consruct a
standard R array
:
Now we reate a NeuroSpace
instance that describes the
geometry of the image including, at minimum, its dimensions and voxel
spacing.
bspace <- NeuroSpace(dim=c(64,64,64), spacing=c(1,1,1))
vol <- NeuroVol(x, bspace)
vol
#>
#> === NeuroVol Object ===
#>
#> * Basic Information
#> Type: DenseNeuroVol
#> Dimensions: 64 x 64 x 64 (2 Mb)
#> Total Voxels: 262,144
#>
#> * Data Properties
#> Value Range: [0.00, 0.00]
#>
#> * Spatial Properties
#> Spacing: 1.00 x 1.00 x 1.00 mm
#> Origin: 0.0, 0.0, 0.0 mm
#> Axes: Left-to-Right x Posterior-to-Anterior x Inferior-to-Superior
#>
#> ======================================
#>
#> Access Methods:
#> . Get Slice: slice(object, zlevel=10)
#> . Get Value: object[i, j, k]
#> . Plot: plot(object) # shows multiple slices
We do not usually have to create NeuroSpace
objects,
because geometric information about an image is automatically determined
from information stored in the image file header. Thus,
NeuroSpace
objects are usually copied from existing images
using the space
extractor function when needed:
vol2 <- NeuroVol((vol+1)*25, space(vol))
max(vol2)
#> [1] 25
space(vol2)
#>
#> NeuroSpace Object
#>
#> >> Dimensions
#> Grid Size: 64 x 64 x 64
#> Memory: 5.9 KB
#>
#> >> Spatial Properties
#> Spacing: 1.00 x 1.00 x 1.00 mm
#> Origin: 0.00 x 0.00 x 0.00 mm
#>
#> >> Anatomical Orientation
#> X: Left-to-Right | Y: Posterior-to-Anterior | Z: Inferior-to-Superior
#>
#> >> World Transformation
#> Forward (Voxel to World):
#> 1.000 0.000 0.000 0.000
#> 0.000 1.000 0.000 0.000
#> 0.000 0.000 1.000 0.000
#> 0.000 0.000 0.000 1.000
#> Inverse (World to Voxel):
#> 1.000 0.000 0.000 0.000
#> 0.000 1.000 0.000 0.000
#> 0.000 0.000 1.000 0.000
#> 0.000 0.000 0.000 1.000
#>
#> >> Bounding Box
#> Min Corner: 0.0, 0.0, 0.0 mm
#> Max Corner: 63.0, 63.0, 63.0 mm
#>
#> ==================================================