Manually apply rigid transformations to images

Opens an interactive viewer where images are interactive objects that can be rotated, moved, scaled and mirrored. It is often difficult to automate image registration for H&E images, in particular if tissue sections are damaged, distorted or folded. In such situations, manual registration can be very useful.

Usage

RunAlignment(object, ...)

# Default S3 method
RunAlignment(
  object,
  container_width = "800px",
  container_height = "650px",
  launch_browser = TRUE,
  ...
)

# S3 method for class 'Seurat'
RunAlignment(
  object,
  image_height = 400,
  container_width = "800px",
  container_height = "650px",
  launch_browser = TRUE,
  verbose = TRUE,
  ...
)

Arguments

object

An object

...

Arguments passed to other methods

container_width, container_height

Width and height of paper widget given in pixels.

launch_browser

Should the app be opened in the default browser? Default is TRUE. Set launch_browser = getOption("shiny.launch.browser", interactive()) to use the RStudio built-in browser.

image_height

An integer used for rescaling images

verbose

Print messages

Value

An object with aligned images

default method

Takes a list of images prepared with the prep_image function and send them to the interactive application. When the application stops (after clicking "quit & save") a tibble is returned with information about the rigid transformations applied to the images.

Seurat method

Takes a Seurat object with at least 2 tissue sections and opens an interactive shiny application where rigid transformations can be applied to the images. When the application stops (after clicking "quit & save"), the transformations are applied to the images using RigidTransformImages and a Seurat object is returned. If no transformations are supplied, the function will return the input Seurat object unmodified.

See also

Other image transformations: MaskImages(), RigidTransformImages()

Other transforms: CoordAndImageTransform(), CoordMirror(), CoordTransform(), ImageTranslate(), RigidTransformImages()

Author

Ludvig Larsson

Examples

library(semla)
library(magick)

im_mbrain <- system.file("extdata/mousebrain/spatial",
                         "tissue_lowres_image.jpg",
                         package = "semla")

img1 <- prep_image(im_mbrain |>
                     image_read(),
                   height = 256)
img2 <- prep_image(im_mbrain |>
                     image_read() |>
                     image_flip(),
                   height = 256)

# Align images and find transform
if (FALSE) { # \dontrun{
if (interactive()) {
  transforms <- RunAlignment(object = list(img1, img2))
}
} # }


se_mbrain <- readRDS(system.file("extdata/mousebrain", "se_mbrain", package = "semla"))
se_mcolon <- readRDS(system.file("extdata/mousecolon", "se_mcolon", package = "semla"))
se_merged <- MergeSTData(se_mbrain, se_mcolon) |> 
  LoadImages()
#> 
#> ── Loading H&E images ──
#> 
#> ℹ Loading image from /private/var/folders/z8/nrcst881607gn95xh3_4qvb00000gp/T/RtmpAdeL2K/temp_libpath109e41827971/semla/extdata/mousebrain/spatial/tissue_lowres_image.jpg
#> ℹ Scaled image from 600x565 to 400x377 pixels
#> ℹ Loading image from /private/var/folders/z8/nrcst881607gn95xh3_4qvb00000gp/T/RtmpAdeL2K/temp_libpath109e41827971/semla/extdata/mousecolon/spatial/tissue_lowres_image.jpg
#> ℹ Scaled image from 541x600 to 400x444 pixels
#> ℹ Saving loaded H&E images as 'rasters' in Seurat object

# Run alignment application on a Seurat object
if (FALSE) { # \dontrun{
if (interactive()) {
  se_merged <- RunAlignment(se_merged)
}
} # }