SpatialImage-class: The 'SpatialImage' class

In drighelli/VisiumExperiment: S4 Class for Spatially Resolved -omics Data

The SpatialImage class

Description

The SpatialImage class hierarchy provides representations of images from a variety of sources. It is used by the SpatialExperiment class to manage the loading of images across multiple studies.

Constructor

SpatialImage(x, is.url) will return a SpatialImage object. The class of the object depends on the type of x:

• If x is a raster object, a LoadedSpatialImage is returned. This represents an image that is fully realized into memory, where the raster representation is stored inside the output object.

• If x is a string and is.url=TRUE or it starts with "http://", "http://" or "ftp://", a RemoteSpatialImage is returned. This represents an image that is remotely hosted and retrieved only on request.

• If x is a string and is.url=TRUE or it does not start with a URL-like prefix, a StoredSpatialImage is returned. This represents an image that is stored in a local file and is loaded into memory only on request.

Getting the raster image

For a SpatialImage object x, imgRaster(x, ...) will return a raster object (see ?as.raster). This is effectively a matrix of RGB colors for each pixel in the image.

For a StoredSpatialImage object x, additional arguments in ... are passed to image_read. This controls how the image is read into memory.

For a RemoteSpatialImage object x, the image file is first downloaded before the raster is returned. Here, ... may contain an extra cache argument, which should be a BiocFileCache object (from the BiocFileCache package) specifying the file cache location. The default location is determined by options("SpatialExperiment.remote.cache.path"), otherwise it defaults to a subdirectory in the R temporary directory. Any further named arguments in ... are passed to image_read.

as.raster(x, ...) is the same as imgRaster(x, ...).

In-memory caching

For StoredSpatialImage and RemoteSpatialImage objects, loading the image with imgRaster will automatically store the loaded raster object in an in-memory cache. Any subsequent imgRaster call will retrieve the raster from the cache, avoiding costly retrieval from the file system.

The cache policy is to evict the least recently used images when a new image would be added that exceeds the maximum cache size. If the new image by itself exceeds the maximum cache size, all images are evicted from the cache to trigger garbage collection and free up memory.

By default, the maximum size of the cache is 4 GB. This can be modified by setting options("SpatialExperiment.cache.size") to some number of bytes, e.g., 2^32.

Transformations

Two basic image transformations are currently supported for any SpatialImage x, namely, rotateImg(x, degrees) for clockwise (degrees > 0) and counter-clockwise (degrees < 0) rotation, and mirrorImg(x, axis) for horizontal (axis = "h") and vertical (axis = "v") mirroring.

Note that, both rotateImg() and mirrorImg() operate on the raster matrix of the input SpatialImage. Thus, any SpatialImage will automatically be coerced into a LoadedSpatialImage upon rotation/mirroring.

Other methods

dim(x) will return an integer vector of length 2, containing the width and height of the image in pixels. Note that this calls imgRaster under the hood and thus may interact with the file and memory caches as described above.

For any SpatialImage x, as(x, "LoadedSpatialImage") will create a LoadedSpatialImage containing an in-memory raster object.

For a RemoteSpatialImage x, as(x, "StoredSpatialImage") will create a StoredSpatialImage pointing to the file cache location.

Author(s)

Aaron Lun

Examples

path <- system.file(
  "extdata", "10xVisium", "section1", "outs", "spatial", 
  "tissue_lowres_image.png", package="SpatialExperiment")

spi <- SpatialImage(path)
plot(imgRaster(spi))

# the following operations all use the cache 
# so there is no need to reload the image
nrow(spi)
ncol(spi)
plot(as.raster(spi)) 

# coercing to an explicitly in-memory raster
spi <- as(spi, "LoadedSpatialImage")
plot(as.raster(spi))

###################
# transformations #
###################

# (counter-)clockwise rotation
spi1 <- rotateImg(spi, degrees = +90)
spi2 <- rotateImg(spi, degrees = -90)

par(mfrow = c(1, 3))
plot(as.raster(spi))
plot(as.raster(spi1))
plot(as.raster(spi2))

# horizontal/vertical mirroring
spi1 <- mirrorImg(spi, axis = "h")
spi2 <- mirrorImg(spi, axis = "v")

par(mfrow = c(1, 3))
plot(as.raster(spi))
plot(as.raster(spi1))
plot(as.raster(spi2))

drighelli/VisiumExperiment documentation built on April 10, 2024, 8:01 a.m.