Preparation

setRepositories(ind=c(1,2))
install.packages(c("EBImage","devtools"))
library("devtools")
install_github("bioimaginggroup/bioimagetools")
install_github("bioimaginggroup/nucim")

code to classify one rgb image

library(bioimagetools)
library(nucim)

choose RGB file

img = readTIF(file.choose())

number of slices

slices = dim(img)[4]

we need the dimensions of the image in microns

x = attributes(img)$x.resolution
y = attributes(img)$y.resolution
z = as.numeric(attributes(img)$spacing) * slices

and the dimensions of each voxel

X = x/dim(img)[1]
Y = y/dim(img)[2]
Z = as.numeric(attributes(img)$spacing)
zscale=mean(c(X,Y))/Z

we assume that the third channel is blue, ie, DAPI

blue = img[,,3,] 

we mask the kernel

mask = dapimask(blue, c(x,y,z), thresh="auto")

classify the DAPI channel

classes = classify(blue, mask, 7, z=zscale)

count voxel per class

counts <- table.n(classes, 7)

percentages

perc <- print(counts/sum(counts)*100, 1)
barplot(perc, names.arg=1:7, xlab="DAPI intensity class", ylab="percentage")

code to classify a folder full of rgb images

library(bioimagetools)
library(nucim)

choose one of the files in a folder of RGB files

folder = file.choose()
f = unlist(gregexpr("/",folder))
folder = substr(folder,1,f[length(f)])

scripts can use parallel computing, if available (not under windows)

nr.cores=ifelse(.Platform$OS.type=="windows", 1, 4)

split channels

splitchannels.folder(folder, rgb.folder="./", cores=nr.cores)

masks

dapimask.folder(folder, cores=nr.cores)

classification

classify.folder(folder, 7, cores=nr.cores)

results will be in folders “class7” und “class7-n”