README.md

In muschellij2/NiftiArray: HDF5 Delayed Array for Nifti Objects

The NiftiArray package is under development please check back later!

NiftiArray

muschellij2 badges:

avalcarcel9 badges:

The goal of NiftiArray is to allow for memory efficient fast random access of NIfTI objects. NiftiArray is an R package that allows for convenient and memory-efficient containers for on-disk representation of NIfTI objects. We allow for DelayedArray extensions and support all operations supported by DelayedArray objects. These operations can be either delayed or block-processed.

You can find a package vignette here.

Installation

You can install the development version of NiftiArray from GitHub with:

# install.packages('remotes')
remotes::install_github("muschellij2/NiftiArray")

We are working to get a stable version on Neuroconductor.

Example

Here we use the example image from RNifti. We use the writeNiftiArray function to create a NiftiArray object:

library(NiftiArray)
nii_fname = system.file("extdata", "example.nii.gz", package = "RNifti")
res = writeNiftiArray(nii_fname)
class(res)
#> [1] "NiftiArray"
#> attr(,"package")
#> [1] "NiftiArray"
dim(res)
#> [1] 96 96 60
res
#> <96 x 96 x 60> NiftiArray object of type "integer":
#> ,,1
#>        [,1]  [,2]  [,3]  [,4] ... [,93] [,94] [,95] [,96]
#>  [1,]     0     0     0     0   .     0     0     0     0
#>  [2,]     0     0     0     0   .     0     0     0     0
#>   ...     .     .     .     .   .     .     .     .     .
#> [95,]     0     0     0     0   .     0     0     0     0
#> [96,]     0     0     0     0   .     0     0     0     0
#> 
#> ...
#> 
#> ,,60
#>        [,1]  [,2]  [,3]  [,4] ... [,93] [,94] [,95] [,96]
#>  [1,]     0     0     0     0   .     0     0     0     0
#>  [2,]     0     0     0     0   .     0     0     0     0
#>   ...     .     .     .     .   .     .     .     .     .
#> [95,]     0     0     0     0   .     0     0     0     0
#> [96,]     0     0     0     0   .     0     0     0     0

We can see the file on disk that was written out:

res@seed@filepath
#> [1] "/private/var/folders/sq/x3htb34928bfn79jk3qksqg56zmf39/T/Rtmp0WocfN/file1585a6fe4f86b.h5"

We see that the object is a low-memory DelayedArray:

object.size(res)
#> 8912 bytes

You can also simply use the NiftiArray function of the NIfTI filename to create the array:

res = NiftiArray(nii_fname)

We see the header information is encoded in the seed slot of the object, which can be accessed using the nifti_header function:

nifti_header(res)
#> NIfTI-1 header
#>     sizeof_hdr: 348
#>       dim_info: 0
#>            dim: 3  96  96  60  1  1  1  1
#>      intent_p1: 0
#>      intent_p2: 0
#>      intent_p3: 0
#>    intent_code: 0
#>       datatype: 8
#>         bitpix: 32
#>    slice_start: 0
#>         pixdim: -1.0  2.5  2.5  2.5  0.0  0.0  0.0  0.0
#>     vox_offset: 352
#>      scl_slope: 0
#>      scl_inter: 0
#>      slice_end: 0
#>     slice_code: 0
#>     xyzt_units: 10
#>        cal_max: 2503
#>        cal_min: 0
#> slice_duration: 0
#>        toffset: 0
#>        descrip: TractoR NIfTI writer v3.0.0
#>       aux_file: 
#>     qform_code: 2
#>     sform_code: 2
#>      quatern_b: 0
#>      quatern_c: 1
#>      quatern_d: 0
#>      qoffset_x: 122.0339
#>      qoffset_y: -95.18523
#>      qoffset_z: -55.03814
#>         srow_x: -2.5000  0.0000  0.0000  122.0339
#>         srow_y: 0.00000  2.50000  0.00000  -95.18523
#>         srow_z: 0.00000  0.00000  2.50000  -55.03814
#>    intent_name: 
#>          magic: n+1

Creating a Matrix

mat = as(res, "NiftiMatrix")
mat
#> <552960 x 1> NiftiMatrix object of type "integer":
#>           [,1]
#>      [1,]    0
#>      [2,]    0
#>      [3,]    0
#>      [4,]    0
#>      [5,]    0
#>       ...    .
#> [552956,]    0
#> [552957,]    0
#> [552958,]    0
#> [552959,]    0
#> [552960,]    0

Now that the image is a matrix, we can bind the columns together,

mat = DelayedArray::acbind(mat, mat, mat, mat)
testthat::expect_is(mat, "DelayedMatrix")
object.size(mat)
#> 34568 bytes

Now that we have the data in a DelayedMatrix class, we can use the package DelayedMatrixStats package that calls the matrixStats package for quick operations:

vec_result = DelayedMatrixStats::rowMedians(mat)
head(vec_result)
#> [1] 0 0 0 0 0 0

Turning the output back into a NiftiArray, we have to pass in the header argument, passing in the correct header information. We can either create the NiftiArray output by creating a matrix, then the NiftiMatrix, then the NiftiArray.

res_mat = matrix(vec_result, ncol = 1)
res_mat = as(res_mat, "NiftiMatrix")
hdr = nifti_header(res)
res_mat = writeNiftiArray(res_mat, header = hdr)
class(res_mat)
#> [1] "NiftiMatrix"
#> attr(,"package")
#> [1] "NiftiArray"
res_arr = as(res_mat, "NiftiArray")

Or we can create an array and then making the NiftiArray:

arr = array(vec_result, dim = dim(res) )
hdr = nifti_header(res)
res_arr = writeNiftiArray(arr, header = hdr)
res_arr
#> <96 x 96 x 60> NiftiArray object of type "double":
#> ,,1
#>        [,1]  [,2]  [,3] ... [,95] [,96]
#>  [1,]     0     0     0   .     0     0
#>  [2,]     0     0     0   .     0     0
#>   ...     .     .     .   .     .     .
#> [95,]     0     0     0   .     0     0
#> [96,]     0     0     0   .     0     0
#> 
#> ...
#> 
#> ,,60
#>        [,1]  [,2]  [,3] ... [,95] [,96]
#>  [1,]     0     0     0   .     0     0
#>  [2,]     0     0     0   .     0     0
#>   ...     .     .     .   .     .     .
#> [95,]     0     0     0   .     0     0
#> [96,]     0     0     0   .     0     0
nifti_header(res_arr)
#> NIfTI-1 header
#>     sizeof_hdr: 348
#>       dim_info: 0
#>            dim: 3  96  96  60  1  1  1  1
#>      intent_p1: 0
#>      intent_p2: 0
#>      intent_p3: 0
#>    intent_code: 0
#>       datatype: 8
#>         bitpix: 32
#>    slice_start: 0
#>         pixdim: -1.0  2.5  2.5  2.5  0.0  0.0  0.0  0.0
#>     vox_offset: 352
#>      scl_slope: 0
#>      scl_inter: 0
#>      slice_end: 0
#>     slice_code: 0
#>     xyzt_units: 10
#>        cal_max: 2503
#>        cal_min: 0
#> slice_duration: 0
#>        toffset: 0
#>        descrip: TractoR NIfTI writer v3.0.0
#>       aux_file: 
#>     qform_code: 2
#>     sform_code: 2
#>      quatern_b: 0
#>      quatern_c: 1
#>      quatern_d: 0
#>      qoffset_x: 122.0339
#>      qoffset_y: -95.18523
#>      qoffset_z: -55.03814
#>         srow_x: -2.5000  0.0000  0.0000  122.0339
#>         srow_y: 0.00000  2.50000  0.00000  -95.18523
#>         srow_z: 0.00000  0.00000  2.50000  -55.03814
#>    intent_name: 
#>          magic: n+1

Converting back to niftiImage

We can return a niftiImage from the NiftiArray object, as follows:

nii = as(res_arr, "niftiImage")
nii
#> Image array of mode "double" (4.2 Mb)
#> - 96 x 96 x 60 voxels
#> - 2.5 x 2.5 x 2.5 mm per voxel

muschellij2/NiftiArray documentation built on June 17, 2020, 1:39 a.m.