CytoImageList-manipulation: Manipulating CytoImageList objects

In BodenmillerGroup/cytomapper: Visualization of highly multiplexed imaging data in R

Manipulating CytoImageList objects

Description

Methods to change pixel values in CytoImageList objects. In the following sections, object is a CytoImageList object containing one or multiple channels.

Value

A CytoImageList object containing the manipulated Images.

Image scaling

In some cases, images need to be scaled by a constant (e.g. 2^16-1 = 65535) value to revert them back to the original pixel values after reading them in.

scaleImages(object, value)

Scales all images in the CytoImageList object by value. Here value needs to be a single numeric or a numeric vector of the same length as object.

Image normalization

Linear scaling of the intensity values of each Image contained in a CytoImageList object to a specific range. Images can either be scaled to the minimum/maximum value per channel or across all channels (default separateChannels = TRUE). Also, images can be scaled to the minimum/maximum value per image or across all images (default separateImages = FALSE). The latter allows the visual comparison of intensity values across images.

To clip the images before normalization, the inputRange can be set. The inputRange either takes NULL (default), a vector of length 2 specifying the clipping range for all channels or a list where each named entry contains a channel-specific clipping range.

Image normalization also works for images stored on disk. By default, the normalized images are stored as a second entry called "XYZ_norm" in the .h5 file. Here "XYZ" specifies the name of the original entry. By storing the normalized next to the original images on disk, space usage increases. To avoid storing duplicated data, one can specify overwrite = TRUE, therefore deleting the original images and only storing the normalized images. However, the original images cannot be accessed anymore after normalisation.

normalize(object, separateChannels = TRUE, separateImages = FALSE, ft = c(0, 1), inputRange = NULL, overwrite = FALSE):

object:

A CytoImageList object

separateChannels:

Logical if pixel values should be normalized per channel (default) or across all channels.

separateImages:

Logical if pixel values should be normalized per image or across all images (default).

ft:

Numeric vector of 2 values, target minimum and maximum intensity values after normalization (see normalize).

inputRange:

Numeric vector of 2 values, sets the absolute clipping range of the input intensity values (see normalize). Alternatively a names list where each entry corresponds to a channel-specific clipping range.

overwrite:

Only relevant when images are kept on disk. By specifying overwrite = TRUE, the normalized images will overwrite the original images in the .h5 file, therefore reducing space on disk. However, the original images cannot be accessed anymore after normalization. If overwrite = FALSE (default), the normalized images are added as a new entry called "XYZ_norm" to the .h5 file.

)

Author(s)

Nils Eling ([e-mail])

See Also

normalize for details on Image normalization

Examples

data(pancreasImages)

# Scale images to create segmentation masks
cur_files <- list.files(system.file("extdata", package = "cytomapper"),
                        pattern = "mask.tiff", full.names = TRUE)
x <- loadImages(cur_files)
# Error when running plotCells(x)
# Therefore scale to account for 16 bit encoding
x <- scaleImages(x, 2^16 - 1)
plotCells(x)

# Default normalization
x <- normalize(pancreasImages)
plotPixels(x, colour_by = c("H3", "CD99"))

# Setting the clipping range
x <- normalize(x, inputRange = c(0, 0.9))
plotPixels(x, colour_by = c("H3", "CD99"))

# Setting the clipping range per channel
x <- normalize(pancreasImages,
                inputRange = list(H3 = c(0, 70), CD99 = c(0, 100)))
plotPixels(x, colour_by = c("H3", "CD99"))

# Normalizing per image
x <- normalize(pancreasImages, separateImages = TRUE)
plotPixels(x, colour_by = c("H3", "CD99"))

BodenmillerGroup/cytomapper documentation built on March 24, 2024, 10:21 p.m.