In aoles/RBioFormats: R interface to Bio-Formats

Introduction

r Biocpkg('RBioFormats') provides an interface from R to the OME Bio-Formats Java library. Bio-Formats is a solution for reading data of various image types, including many popular in life sciences as well as proprietary microscopy image formats. It supports over 150 file formats from domains such as High Content Screening, time lapse imaging, digital pathology and other complex multidimensional image formats.

Image pixel data is typically complemented by image metadata containing, for example, technical and temporal parameters of the acquisition in the case of microscopy images. Such annotation can be an invaluable source of additional insight helpful during postprocessing or analyzing of the image data.

The package builds on top of the infrastructure provided by r Biocpkg('EBImage') by extending its class abstracting image data. The primary motivation behind developing r Biocpkg('RBioFormats') was to fill the gap between data acquisition and analysis by providing a tool which allows to directly read the acquired images without the need of any tedious image format conversion in between.

The following chapters provide some practical examples illustrating the use of the package. Along the way the classes used for representing image data and metadata are described too.

Getting started

r Biocpkg("RBioFormats") is an R package distributed as part of the Bioconductor project. To install the package, start R (version 4.2 or higher) and enter:

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("RBioFormats")

Once r Biocpkg("RBioFormats") is installed, it can be loaded by the following command.

library("RBioFormats")

Reading images

Images can be loaded into R with the help of the package function read.image. The following examples illustrates how to load a sample grayscale image

f <- system.file("images", "sample.png", package = "EBImage")

img <- read.image(f)
img

or an RGB image.

f <- system.file("images", "sample-color.png", package = "EBImage")

img <- read.image(f)
print(img, short = TRUE)

Note the use of short = TRUE argument to print in the example above for displaying object summary without the image data preview. There is also a convenience function to query just for the order of dimensions.

dimorder(img)

The AnnotatedImage class

r Rpackage('RBioFormats') stores image data in an AnnotatedImage class which extends the Image class from r Biocpkg('EBImage').

getClassDef("AnnotatedImage")

Compared to the original Image class the AnnotatedImage class features an additional metadata slot containing image metadata.

meta <- metadata(img)
meta

To alter the length of the printed output use the list.len attribute to print.

print(meta, list.len = 99L)

Image metadata

Image metadata is represented by an ImageMetadata class structured as a named list of coreMetadata, globalMetadata and seriesMetadata.

names(meta)
cMeta <- meta$coreMetadata
names( cMeta )

coreMetadata stores information which is guaranteed to exist for any image type, whereas the latter two metadata types are format-specific and can be empty.

Each of these metadata sublists has an corresponding accessor function, e.g.,

identical( coreMetadata(meta), cMeta)

and similarly for globalMetadata and seriesMetadata. These accessors are useful for extracting the corresponding metadata directly from an AnnotatedImage object

identical( coreMetadata(img), cMeta )

Working with large data sets

The read.metadata function allows to access image metadata without loading the corresponding pixel data.

f <- system.file("images", "nuclei.tif", package = "EBImage")
metadata <- read.metadata(f)

metadata

This approach is especially useful when working with image series and/or stacks which have high memory requirements. Information from the metadata can be used as input to functions which read and process the data chunk-wise rather than loading it all at once. For this purpose the subset argument to read.image comes in handy. Just to give you an idea the following toy example iterates over individual time frames. Similarly a region if interest from within individual image frames could be extracted by providing ranges on the X and Y planar dimensions. To subset image series specify them in the series argument.

for(t in seq_len(coreMetadata(metadata)$sizeT)) {
  frame <- read.image(f, subset = list(T = t))
  # perform some operations on each `frame`
}

OME-XML representation

The OME-XML DOM tree representation of the metadata can be accessed using tools from the r CRANpkg('XML') or r CRANpkg('xml2') package. For details on working with XML data in R see the corresponding package's documentation.

library("xml2")

omexml <- read.omexml(f)
read_xml(omexml)

Session info {.unnumbered}

Here is the output of sessionInfo() on the system on which this document was compiled:

sessionInfo()

Appendix A: Working with test images {.unnumbered}

For development purposes it is useful to have images of a specific size or pixel type for testing. Mock files containing gradient images can be generated with

f <- mockFile(sizeX = 256, sizeY = 256)
img <- read.image(f)

library("EBImage")
display(img, method = "raster")

Note that the native image data range is different depending on pixel type.

types <- c("int8", "uint8", "int16", "uint16", "int32", "uint32", "float", "double")

ranges <- sapply(types, function(t) {
  minmax <- FormatTools$defaultMinMax(FormatTools$pixelTypeFromString(t))
  setNames(minmax, c("min", "max"))
  })
knitr::kable(ranges)

Image data returned by r Biocpkg('RBioFormats') is by default scaled to the [0:1] range. This behavior can be controlled using the normalize argument to read.image.

sapply(types, function(t) {
  img <- read.image(mockFile(sizeX = 65536, sizeY = 11, pixelType = t), normalize = FALSE)
  if (typeof(img)=="raw") 
    img <- readBin(img, what = "int", n = length(img), size = 1L)
  setNames(range(img), c("min", "max"))  
})

Appendix B: Compared to EBImage {.unnumbered}

Loading images using r Biocpkg('RBioFormats') should give the same results as using the r Biocpkg('EBImage') package.

library("EBImage")
f <- system.file("images", "sample-color.png", package = "EBImage")
identical(readImage(f), as.Image(read.image(f)))

aoles/RBioFormats documentation built on May 3, 2024, 7:46 a.m.