spatialtiffterra

Reading a segmented tif image (values 0 and 1) into r

I have just started to learn how to analyse immunohistochemitry images using R. I have a segmented image (segmentation performed by others) that identifies cells positive for one marker, and I need to read this into R for plotting and spatial analyses. I struggle with understanding how to read in this file using terra:rast(), or if other methods are more appropriate?

Link to the file: https://drive.google.com/file/d/1152QLPxF98GqI0IrbVy6VHTEa9-IpcXU/view?usp=sharing

When I read in this file I get a warning:

test <- rast("Test.tif")
Warning: [rast] unknown extent

Description of test:

test

class       : SpatRaster 
dimensions  : 8648, 7170, 1  (nrow, ncol, nlyr)
resolution  : 1, 1  (x, y)
extent      : 0, 7170, 0, 8648  (xmin, xmax, ymin, ymax)
coord. ref. :  
source      : Test.tif 
color table : 1 
name        : Test

QUESTIONS:

  1. What should I do with that warning?

  2. Why are the values changed so that white is 0 and black is 1 compared to other methods to read in the file?

  3. When I plot the image (plot(test)), it is both turned and mirrored. Why? And how to fix this?

Original file:

Plotted in r:

  1. I try to change the colors in the plot to: black = no color, white = brown using the following code, but nothing happens to the colors!
plot(test, col = c("brown", "NA"))

If I read in using

test <- readTIFF("Test.tif", convert = TRUE)

and then convert to raster:

test <- rast(test)

Then the code for plotting with changed colors works. The description of test is somewhat different with min and max values specified:

test
class       : SpatRaster 
dimensions  : 8648, 7170, 1  (nrow, ncol, nlyr)
resolution  : 1, 1  (x, y)
extent      : 0, 7170, 0, 8648  (xmin, xmax, ymin, ymax)
coord. ref. :  
source(s)   : memory
name        : lyr.1 
min value   :     0 
max value   :     1

How to read in using terra:rast (have a lot of files so would like to omit storing the file information in memory) but be able to change the colors?

Some additional information:

terra::describe("Test.tif")

[1] "Driver: GTiff/GeoTIFF"                                  "Files: "Test.tif"
[3] "Size is 7170, 8648"                                     "Metadata:"                                             
[5] "  TIFFTAG_XRESOLUTION=72"                               "  TIFFTAG_YRESOLUTION=72"                              
[7] "  TIFFTAG_RESOLUTIONUNIT=2 (pixels/inch)"               "Image Structure Metadata:"                             
[9] "  COMPRESSION=CCITTRLE"                                 "  INTERLEAVE=BAND"                                     
[11] "  MINISWHITE=YES"                                       "Corner Coordinates:"                                   
[13] "Upper Left  (    0.0,    0.0)"                          "Lower Left  (    0.0, 8648.0)"                         
[15] "Upper Right ( 7170.0,    0.0)"                          "Lower Right ( 7170.0, 8648.0)"                         
[17] "Center      ( 3585.0, 4324.0)"                          "Band 1 Block=7170x135 Type=Byte, ColorInterp=Palette"  
[19] "  Image Structure Metadata:"                            "    NBITS=1"                                           
[21] "  Color Table (RGB with 2 entries)"                     "    0: 255,255,255,255"                                
[23] "    1: 0,0,0,255"
Solution

  • You can ignore the warning - it comes from read_gdal.cpp, I believe. Terra package is supposed to work with georeferenced data, which have extent and CRS specified and coded within data itself.

    r <- terra::rast("~/Downloads/Test.tif")
#> Warning: [rast] unknown extent

r
#> class       : SpatRaster 
#> dimensions  : 8648, 7170, 1  (nrow, ncol, nlyr)
#> resolution  : 1, 1  (x, y)
#> extent      : 0, 7170, 0, 8648  (xmin, xmax, ymin, ymax)
#> coord. ref. :  
#> source      : Test.tif 
#> color table : 1 
#> name        : Test

terra::plot(r)

    The black and white colors are forced by color table in tiff:

    terra::coltab(r)
#> [[1]]
#>   value red green blue alpha
#> 1     0 255   255  255   255
#> 2     1   0     0    0   255

    To plot it in different colors (without changing the values of raster), we have to define and change color table:

    ctb <- data.frame(value = 0:1, col = c("brown", "black"))
terra::coltab(r) <- ctb

terra::plot(r)

    If you wish to change the raster values, you might use something like:

    s <- r
s[s == 1] <- NA
terra::plot(s, col = data.frame(0, "brown"))

    Just for information the version of {terra} used:

    packageVersion("terra")
#> [1] '1.8.3'

    If your raster is flipped/rotated, please have a look on terra::flip() and/or rev() functions.

    Created on 2024-12-08 with reprex v2.1.1.9000

    PS. The "color map" in TIFF is given in bits per pixel and color.space attributes of file:

    tt <- tiff::readTIFF("~/Downloads/Test.tif", info = TRUE, convert = TRUE)
attributes(tt)$bits.per.sample
#> [1] 1
attributes(tt)$color.space
#> [1] "white is zero"