In ellispatrick/AFremover: A method to identify and remove autofluorescent signals from multi-channel images post acquisition

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-"
)

AFremover

Autofluorescence is a long-standing problem that has hindered fluorescence microscopy image analysis. To address this, we have developed a method that identifies and removes autofluorescent signals from multi-channel images post acquisition.

Installation

library(devtools)
devtools::install_github("ellispatrick/AFremover")

A quick example

library(AFremover)
set.seed(51773)
## Read in images.
imageFile1 = system.file("extdata","ImageB.CD3.tif", package = "AFremover")
imageFile2 = system.file("extdata","ImageB.CD11c.tif", package = "AFremover")
im1 <- EBImage::readImage(imageFile1)
im2 <- EBImage::readImage(imageFile2)

## Rescale the images.
im1 = im1/max(im1)
im2 = im2/max(im2)

combined <- EBImage::rgbImage(green=sqrt(im1), red=sqrt(im2))
EBImage::display(combined, all = TRUE, method = 'raster')


## Create masks using EBImage.

# Find tissue area
tissue1 = im1 > 2*min(im1)
tissue2 = im2 > 2*min(im2)

# Calculate thresholds
imThreshold1 <- mean(im1[tissue1]) + 2*sd(im1[tissue1])
imThreshold2 <- mean(im2[tissue2]) + 2*sd(im2[tissue2])

# Calculate masks.
mask1 <- EBImage::bwlabel(im1 > imThreshold1)
mask2 <- EBImage::bwlabel(im2 > imThreshold2)

## Calculate intersection mask
mask <- intMask(mask1,mask2)


## Calculate textural features.
df <- afMeasure(im1, im2, mask)

## Alternatively
## Correlation only
# afMask <- afIdentify(mask, df, minSize = 100, maxSize = Inf, corr = 0.6)

## Clustering with given k
# afMask <- afIdentify(mask, df, minSize = 100, maxSize = Inf, k = 6)

## Clustering with estimated k.
afMask <- afIdentify(mask, df, minSize = 100, maxSize = Inf, k = 20, kAuto = TRUE)


## Remove autofluorescence from images (Use ImageJ for halo removal)
im1AFRemoved <- im1
im2AFRemoved <- im2
im1AFRemoved[afMask != 0] <- 0
im2AFRemoved[afMask != 0] <- 0

combinedRemoved <- EBImage::rgbImage(green = sqrt(im1AFRemoved), red = sqrt(im2AFRemoved))
img_comb = EBImage::combine(combined, combinedRemoved)
EBImage::display(img_comb, all = TRUE, method = 'raster',nx = 2)


## Or
## Exclude AF ROIs

exclude1 = unique(mask1[afMask>0])
mask1Removed = mask1
mask1Removed[mask1Removed%in%exclude1] = 0
exclude2 = unique(mask2[afMask>0])
mask2Removed = mask2
mask2Removed[mask2Removed%in%exclude2] = 0


## OR
## A more agressive strategy to remove autofluorescence from images as an alternative to the halo removal implemented in ImageJ
im1AFRemovedv2 <- im1
im2AFRemovedv2 <- im2
im1AFRemovedv2[!mask1%in%mask1Removed|!mask2%in%mask2Removed] <- 0
im2AFRemovedv2[!mask1%in%mask1Removed|!mask2%in%mask2Removed] <- 0

combinedRemovedv2 <- EBImage::rgbImage(green = sqrt(im1AFRemovedv2), red = sqrt(im2AFRemovedv2))
img_combv2 = EBImage::combine(combined, combinedRemovedv2)
EBImage::display(img_combv2, all = TRUE, method = 'raster',nx = 2)

ellispatrick/AFremover documentation built on June 11, 2019, 7:49 p.m.