| Title: | CRImage a package to classify cells and calculate tumour cellularity |
|---|---|
| Description: | CRImage provides functionality to process and analyze images, in particular to classify cells in biological images. Furthermore, in the context of tumor images, it provides functionality to calculate tumour cellularity. |
| Authors: | Henrik Failmezger <[email protected]>, Yinyin Yuan <[email protected]>, Oscar Rueda <[email protected]>, Florian Markowetz <[email protected]> |
| Maintainer: | Henrik Failmezger <[email protected]>, Yinyin Yuan <[email protected]> |
| License: | Artistic-2.0 |
| Version: | 1.55.0 |
| Built: | 2024-11-21 06:06:43 UTC |
| Source: | https://github.com/bioc/CRImage |
CRImage allows classification of cells in biological images. It offers methods to segment cells or cell nuclei in biological images for example HE stained images. It offers methods to create a classifier and to classify cells in these images. Furthermore it allows the calculation of tumour cellularity for large microscope images.
CRImage makes use of the image processing package EBImage, which uses the 'ImageMagick' library for image I/O operations and the 'GTK' library to display images.
| Package: | CRImage |
| Type: | Package |
| Version: | 1.0 |
| Date: | 2010-04-27 |
| License: | LGPL Version 2 or later |
| LazyLoad: | yes |
Image processing methods:
calculateThreshold
segmentImage
Classification:
createTrainingSet
createClassifier
classifyCells
Tumour cellularity
calculateCellularity
processAperio
Henrik Failmezger, <[email protected]>
Yinyin Yuan, <[email protected]>
Oscar Rueda, <[email protected]>
Florian Markowetz, <[email protected]>
CRI Cambridge
Li Ka Shing Centre
Robinson Way
Cambridge, CB2 0RE, UK
Ludwigs-Maximilians University of Munich
example(segmentImage) example(createClassifier) example(classifyImage)example(segmentImage) example(createClassifier) example(classifyImage)
The function calculates the tumour cellularity of an image by counting tumour and non tumour cells.
calculateCellularity(filename="",image=NA,classifier=NULL,cancerIdentifier=NA,KS=FALSE,maxShape=NA,minShape=NA,failureRegion=NA,colors=c(),threshold="otsu",classesToExclude=c(),numWindows=2,classifyStructures=FALSE,pixelClassifier=NA,ksToExclude=c(),densityToExclude=c(),numDensityWindows=4)calculateCellularity(filename="",image=NA,classifier=NULL,cancerIdentifier=NA,KS=FALSE,maxShape=NA,minShape=NA,failureRegion=NA,colors=c(),threshold="otsu",classesToExclude=c(),numWindows=2,classifyStructures=FALSE,pixelClassifier=NA,ksToExclude=c(),densityToExclude=c(),numDensityWindows=4)
filename |
A path to an image file. |
image |
If filename is undefined, an Image object |
classifier |
A SVM object, created with createClassifier or directly with the package e1071 |
cancerIdentifier |
A string which describes, how the cancer class is named. |
KS |
Apply kernel smoother? |
maxShape |
Maximum size of cell nuclei |
minShape |
Minimum size of cell nuclei |
failureRegion |
minimum size of failure regions |
colors |
Colors to paint the classes |
threshold |
Which threshold should be uses, "otsu" or "phansalkar" |
classesToExclude |
Should a class be excluded from cellularity calculation? |
numWindows |
Number of windows for the threshold. |
classifyStructures |
Use hierarchical classification. If yes a pixel classifier has to be defined. |
pixelClassifier |
A SVM to classify pixel based on their color values. Needed if hierarchical classification should be applied. |
ksToExclude |
These classes are excluded from kernel smoothing. |
densityToExclude |
This class is excluded from cellularity calculation. |
numDensityWindows |
Number of windows for the density plot. |
The method calculates tumour cellularity of an image. The cells of the image are classified and the cellularity is: numTumourCells/numPixel. Furthermore the number of cells of the different classes are counted. A heatmap of cellularity is created. The image is divided in 16 subwindows and cellularity is calculated for every subwindow. Green in the heatmaps indicates strong cellularity, white low cellularity.
A list containing
cellularity values |
a vector, the n first values indicate the n numbers of cells in the n classes, the n + 1th value indicates the tumour cellularity, The n + 2th value is the ratio of tumour cells by all cells |
cancerHeatmap |
Heatmap of cancer density |
Henrik Failmezger, [email protected]
t = system.file("extdata", "trainingData.txt", package="CRImage") #read training data trainingData=read.table(t,header=TRUE) #create classifier classifier=createClassifier(trainingData)[[1]] #calculation of cellularity f = system.file("extdata", "exImg.jpg", package="CRImage") exImg=readImage(f) cellularity=calculateCellularity(classifier=classifier,filename=f,KS=TRUE,maxShape=800,minShape=40,failureRegion=2000,classifyStructures=FALSE,cancerIdentifier="c",numDensityWindows=2,colors=c("green","red"))t = system.file("extdata", "trainingData.txt", package="CRImage") #read training data trainingData=read.table(t,header=TRUE) #create classifier classifier=createClassifier(trainingData)[[1]] #calculation of cellularity f = system.file("extdata", "exImg.jpg", package="CRImage") exImg=readImage(f) cellularity=calculateCellularity(classifier=classifier,filename=f,KS=TRUE,maxShape=800,minShape=40,failureRegion=2000,classifyStructures=FALSE,cancerIdentifier="c",numDensityWindows=2,colors=c("green","red"))
Mean and SD calculation
calculateMeanStdTarget(imgT)calculateMeanStdTarget(imgT)
imgT |
the Image to calculate. |
Mean and SD
Vector with mean and standard deviation.
Henrik Failmezger, [email protected]
#read the target image f1= system.file("extdata", "exImg2.jpg", package="CRImage") targetImage=readImage(f1) #read the image whose color values should be adapted f2= system.file("extdata", "exImg3.jpg", package="CRImage") imgToConvert=readImage(f2) #calculate mean and standard deviation of target color channels mst=calculateMeanStdTarget(targetImage) # create a white pixel mask whitePixelMask=imgToConvert[,,1]>0.85 & imgToConvert[,,2]>0.85 & imgToConvert[,,3]>0.85 #adapt color channels of image imgCorrected=colorCorrection(imgToConvert,mst,whitePixelMask)#read the target image f1= system.file("extdata", "exImg2.jpg", package="CRImage") targetImage=readImage(f1) #read the image whose color values should be adapted f2= system.file("extdata", "exImg3.jpg", package="CRImage") imgToConvert=readImage(f2) #calculate mean and standard deviation of target color channels mst=calculateMeanStdTarget(targetImage) # create a white pixel mask whitePixelMask=imgToConvert[,,1]>0.85 & imgToConvert[,,2]>0.85 & imgToConvert[,,3]>0.85 #adapt color channels of image imgCorrected=colorCorrection(imgToConvert,mst,whitePixelMask)
The function applies Otsu thresholding on the image.
calculateOtsu(allGreyValues)calculateOtsu(allGreyValues)
allGreyValues |
Vector of grey values. |
The function calculates a value which separates the grey value histogram the best in foreground and background.
the threshold
Henrik Failmezger, [email protected]
Nobuyuki Otsu: A threshold selection method from grey level histograms. In: IEEE Transactions on Systems, Man, and Cybernetics. New York 9.1979, S.62-66. ISSN 1083-4419
calculateThreshold localOtsuThreshold
f1= system.file("extdata", "exImg2.jpg", package="CRImage") print(f1) img=readImage(f1) print(img) #convert to grayscale imgG=EBImage::channel(img,'grey') #threshold value t=calculateOtsu(as.vector(imgG))f1= system.file("extdata", "exImg2.jpg", package="CRImage") print(f1) img=readImage(f1) print(img) #convert to grayscale imgG=EBImage::channel(img,'grey') #threshold value t=calculateOtsu(as.vector(imgG))
The function classifies cells and paints the different class types in the image.
classifyCells(classifier,filename="",image=NA,segmentedImage=NA,featuresObjects=NA,paint=TRUE,KS=FALSE,cancerIdentifier=NA, maxShape=NA,minShape=NA,failureRegion=NA,colors=c(),classesToExclude=c(),threshold="otsu",numWindows=2,structures=NA,classifyStructures=FALSE,pixelClassifier=NA,ksToExclude=c())classifyCells(classifier,filename="",image=NA,segmentedImage=NA,featuresObjects=NA,paint=TRUE,KS=FALSE,cancerIdentifier=NA, maxShape=NA,minShape=NA,failureRegion=NA,colors=c(),classesToExclude=c(),threshold="otsu",numWindows=2,structures=NA,classifyStructures=FALSE,pixelClassifier=NA,ksToExclude=c())
classifier |
A Support Vector Machine created by createClassifier or directly by the package e1071 |
filename |
A path to an image file. |
image |
An 'Image' object or an array. |
segmentedImage |
An 'Image' object or an array.The corresponding segmented image (created by segmentImage) |
featuresObjects |
Cell feature file of the segmentedImage (created by segmentImage) |
paint |
If true, the classified cells are painted with different colors in the image |
KS |
Use Kernel Smoohter in classification? |
cancerIdentifier |
A string which describes, how the cancer class is named. |
maxShape |
Maximum size of cell nuclei |
minShape |
Minimum size of cell nuclei |
failureRegion |
minimum size of failure regions |
colors |
Colors to paint the classes |
classesToExclude |
Which class should be excluded? |
threshold |
Which thresholding method should be used, "otsu" or "phansalkar" |
numWindows |
Number of windows to use for thresholding. |
structures |
If the image is already segmented, structures can be inserted to enable hierarchical classification. |
classifyStructures |
Use hierarchical classification. If yes a pixel classifier has to be defined. |
pixelClassifier |
A SVM to classify pixel based on their color values. Needed if hierarchical classification should be applied. |
ksToExclude |
These classes are excluded from kernel smoothing. |
The kernels smoother improves the classification for cells which are likely to occur in clusters, like tumour cells. The kernel smoothing method can only be applied for two classes. If there are more classes only the normal svm without kernel smoothing is applied. Different classes are labeled with different colors in the image.
A list with
comp1 |
classes |
comp2 |
Classes, painted in the image, if paint was true |
Henrik Failmezger, [email protected]
t = system.file("extdata", "trainingData.txt", package="CRImage") #read training data trainingData=read.table(t,header=TRUE) #create classifier classifier=createClassifier(trainingData)[[1]] #classify cells f = system.file("extdata", "exImg.jpg", package="CRImage") classesValues=classifyCells(classifier,filename=f,KS=TRUE,maxShape=800,minShape=40,failureRegion=2000)t = system.file("extdata", "trainingData.txt", package="CRImage") #read training data trainingData=read.table(t,header=TRUE) #create classifier classifier=createClassifier(trainingData)[[1]] #classify cells f = system.file("extdata", "exImg.jpg", package="CRImage") classesValues=classifyCells(classifier,filename=f,KS=TRUE,maxShape=800,minShape=40,failureRegion=2000)
The colors of one image are adapted to the colors of a target image.
colorCorrection(imgO, meanStdTarget,whiteMask = c())colorCorrection(imgO, meanStdTarget,whiteMask = c())
imgO |
The image who's colors should be adapted |
meanStdTarget |
Array with mean and standard deviation of the target image. |
whiteMask |
Boolean mask of white pixel in the image. These pixels are excluded from color correction. |
Mean and standard deviation of the target image can be calculated using the function calculateMeanStdTarget.
The image with adapted colors.
Henrik Failmezger, [email protected]
Reinhard, E.; Adhikhmin, M.; Gooch, B.; Shirley, P.; , "Color transfer between images," Computer Graphics and Applications, IEEE , vol.21, no.5, pp.34-41, Sep/Oct 2001 doi: 10.1109/38.946629 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=946629&isnumber=20481
calculateMeanStandardTarget
#read the target image f1= system.file("extdata", "exImg2.jpg", package="CRImage") targetImage=readImage(f1) #read the image whose color values should be adapted f2= system.file("extdata", "exImg3.jpg", package="CRImage") imgToConvert=readImage(f2) #calculate mean and standard deviation of target color channels mst=calculateMeanStdTarget(targetImage) # create a white pixel mask whitePixelMask=imgToConvert[,,1]>0.85 & imgToConvert[,,2]>0.85 & imgToConvert[,,3]>0.85 #adapt color channels of image imgCorrected=colorCorrection(imgToConvert,mst,whitePixelMask)#read the target image f1= system.file("extdata", "exImg2.jpg", package="CRImage") targetImage=readImage(f1) #read the image whose color values should be adapted f2= system.file("extdata", "exImg3.jpg", package="CRImage") imgToConvert=readImage(f2) #calculate mean and standard deviation of target color channels mst=calculateMeanStdTarget(targetImage) # create a white pixel mask whitePixelMask=imgToConvert[,,1]>0.85 & imgToConvert[,,2]>0.85 & imgToConvert[,,3]>0.85 #adapt color channels of image imgCorrected=colorCorrection(imgToConvert,mst,whitePixelMask)
The function converts images in the HSV colour space to the RGB colour space.
convertHSVToRGB(imgHSV)convertHSVToRGB(imgHSV)
imgHSV |
An 'Image' object or an array in the HSV colour space. |
Standard colour space conversion.
An array in the RGB colour space.
Henrik Failmezger, [email protected]
convertRGBToHSV convertRGBToLAB convertLABToRGB
f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to RGB color space imgRGB=convertHSVToRGB(img)f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to RGB color space imgRGB=convertHSVToRGB(img)
Color space conversion.
convertLABToRGB(imgLAB)convertLABToRGB(imgLAB)
imgLAB |
LAB channel vectors. |
Color space conversion
RGB channel vectors.
Henrik Failmezger, [email protected]
f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to HSV color space imgRGB=convertLABToRGB(img)f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to HSV color space imgRGB=convertLABToRGB(img)
The RGB Image is converted to an HSV image.
convertRGBToHSV(img)convertRGBToHSV(img)
img |
The RGB image |
The entries of the array are Hue, Saturation and Value.
The image in HSV color space.
Henrik Failmezger, [email protected]
convertHSVToRGB convertRGBToLAB convertLABToRGB
f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to HSV color space imgHSV=convertRGBToHSV(img)f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to HSV color space imgHSV=convertRGBToHSV(img)
Conversion of Color spaces.
convertRGBToLAB(imgT)convertRGBToLAB(imgT)
imgT |
The RGB image. |
Color space conversion
The image in LAB color space.
Henrik Failmezger, [email protected]
f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to LAB color space imgLAB=convertRGBToLAB(img)f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to LAB color space imgLAB=convertRGBToLAB(img)
This function segments copy number and corrects log-ratios (LRR) and beta allele frequencies (BAF) values for cellularity.
correctCopyNumber(arr="Sample1", chr=NULL, p=NULL, z=NULL, min.value=-5)correctCopyNumber(arr="Sample1", chr=NULL, p=NULL, z=NULL, min.value=-5)
arr |
Name of the array. |
chr |
Chromosome to run. If |
p |
Percentage of tumoural cells |
.
z |
Copy Number Data. Must be a dataframe with the following columns: Name (id of the probe), Chr (chromosome), Pos (position), LRR (log ratios) and BAF (beta allele frequencies). |
min.value |
Value assigned to the probes that have 0 copies after correction. |
The data.frame z must contain only SNP probes, that is probes
with both LRR and BAF values. It is recommended that all replicated
probes are merged so the positions are unique.
This function calls DNAcopy to segment the LRR and then correct
the segmented profiles for normal contamination according to the
method described in the reference below (see for details).
A list with 2 components:
y |
a data.frame with as many rows as probes containing the following variables: Chrom (chromosome), Pos (position), Orig.LRR (LRR before correction) Orig.BAF (BAF before correction), Corr.LRR (LRR after cellularity correction) and Corr.BAF (BAF after correction) |
seg |
a data.frame with the segmented data. Contains the following columns: ID (name of the array), chrom (chromosome), loc.start (start of the region), loc.end (end of the region), num.mark (number of probes in the region), seg.mean (LRR of the region), BAF (BAF of the regions), num.BAF (number of SNP probes in the region), Sa (estimated absolute copy number for the first allele), Sb (estimated absolute copy number for the first allele), LRR.tum (corrected LRR for the region), BAF.tum (corrected BAF for the region). |
Includes an adaptation of aCGH mergeLevels function to fix
a problem with ansari.test.
Oscar M. Rueda, [email protected]
Yuan, Y et al. Quantitative image analysis of cellular heterogeneity in primary breast tumors enriches genomic assays. In prep.
LRR <- c(rnorm(100, 0, 1), rnorm(10, -2, 1), rnorm(20, 3, 1), rnorm(100,0, 1)) BAF <- c(rnorm(100, 0.5, 0.1), rnorm(5, 0.2, 0.01), rnorm(5, 0.8, 0.01), rnorm(10, 0.25, 0.1), rnorm(10, 0.75, 0.1), rnorm(100,0.5, 0.1)) Pos <- sample(x=1:500, size=230, replace=TRUE) Pos <- cumsum(Pos) Chrom <- rep(1, length(LRR)) z <- data.frame(Name=1:length(LRR), Chrom=Chrom, Pos=Pos, LRR=LRR, BAF=BAF) res <- correctCopyNumber(arr="Sample1", chr=1, p=0.75, z=z)LRR <- c(rnorm(100, 0, 1), rnorm(10, -2, 1), rnorm(20, 3, 1), rnorm(100,0, 1)) BAF <- c(rnorm(100, 0.5, 0.1), rnorm(5, 0.2, 0.01), rnorm(5, 0.8, 0.01), rnorm(10, 0.25, 0.1), rnorm(10, 0.75, 0.1), rnorm(100,0.5, 0.1)) Pos <- sample(x=1:500, size=230, replace=TRUE) Pos <- cumsum(Pos) Chrom <- rep(1, length(LRR)) z <- data.frame(Name=1:length(LRR), Chrom=Chrom, Pos=Pos, LRR=LRR, BAF=BAF) res <- correctCopyNumber(arr="Sample1", chr=1, p=0.75, z=z)
Creates a binary image from a grayscale image by thresholding.
createBinaryImage(imgG,img=NULL,method="otsu",threshold=NULL,numWindows=1,whitePixelMask=c())createBinaryImage(imgG,img=NULL,method="otsu",threshold=NULL,numWindows=1,whitePixelMask=c())
img |
An Image object or an array. |
imgG |
The grey valued Image object. |
method |
Either "otsu" or "phansalkar" |
threshold |
Fixed threshold |
numWindows |
Number of windows to use for threshold calculation. |
whitePixelMask |
Boolean mask of white pixels, if they should be excluded from thresholding |
The functions returns the binary image resulting from the thresholding. If threshold is defined, all pixels smaller than this value will be fixed to 1 all other values will be set to 0. If threshold is undefined, the thresholding value is calculated automatically using 'otsu' or 'phansalkar' thresholding.
The function 'otsu' does Otsu thresholding on the grey level histograms of the image. The function 'phansalkar' does thresholding using the mean and standard deviation of a specified window. The thresholding is done on the RGB as well as the LAP color space and the results are ORed. The window size is dim(img)/numWindows. White pixel can be excluded from thresholding (e.g. if white is background) by defining a whitePixelMask
The binary image.
Henrik Failmezger, [email protected]
Neerad Phansalkar, Sumit More, Ashish Sabale, Dr. Madhuri Joshi, "Adaptive Local Thresholding for Detection of Nuclei in Diversly Stained Cytology Images," 2011 IEEE International Conference in Communications and Signal Processing (ICCSP), pp. 218, 10 Feb. 2011
Nobuyuki Otsu: A threshold selection method from grey level histograms. In: IEEE Transactions on Systems, Man, and Cybernetics. New York 9.1979, S.62-66. ISSN 1083-4419
f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to grayscale imgG=EBImage::channel(img,"gray") imgB=createBinaryImage(imgG,img=img,method="otsu",numWindows=4) #white pixel mask whitePixelMask=img[,,1]>0.85 & img[,,2]>0.85 & img[,,3]>0.85 #exclude white pixels from thresholding imgB=createBinaryImage(imgG,img=img,method="otsu",numWindows=4,whitePixelMask) #phansalkar threshold imgB=createBinaryImage(imgG,img=img,method="phansalkar",numWindows=4)f= system.file("extdata", "exImg.jpg", package="CRImage") img=readImage(f) #conversion to grayscale imgG=EBImage::channel(img,"gray") imgB=createBinaryImage(imgG,img=img,method="otsu",numWindows=4) #white pixel mask whitePixelMask=img[,,1]>0.85 & img[,,2]>0.85 & img[,,3]>0.85 #exclude white pixels from thresholding imgB=createBinaryImage(imgG,img=img,method="otsu",numWindows=4,whitePixelMask) #phansalkar threshold imgB=createBinaryImage(imgG,img=img,method="phansalkar",numWindows=4)
Creates a classifier for a training set.
createClassifier(trainingData, cross = FALSE)createClassifier(trainingData, cross = FALSE)
trainingData |
A table, created by segmentImage with manually added classes. |
cross |
Does 10-fold cross validation to test the classifiers performance. |
Topological features include the density of cells and the size of the surrounding cytoplasma of a cell. These features depend on the size of the image. If training image and the image to classify have different size, these features can fool the classification and should not be enabled.
A List containing:
classifier |
The classifier |
performance |
cross validation performance |
Henrik Failmezger, [email protected]
'createTrainingSet','classifyCells'
f = system.file("extdata", "trainingData.txt", package="CRImage") #read training data trainingData=read.table(f,header=TRUE) #create classifier classifier=createClassifier(trainingData)[[1]]f = system.file("extdata", "trainingData.txt", package="CRImage") #read training data trainingData=read.table(f,header=TRUE) #create classifier classifier=createClassifier(trainingData)[[1]]
The functions creates an interactive session in order to label cells with their classes. The labeled cells can be used as training set for the classifier. Note!! This is until now only tested for MacOsX.
labelCells(img, segmentedImage, classes, classColours, nblocks = 3, labeledPoints = NULL, filename = NULL, filenameImage = NULL,transformCoordinates=FALSE)labelCells(img, segmentedImage, classes, classColours, nblocks = 3, labeledPoints = NULL, filename = NULL, filenameImage = NULL,transformCoordinates=FALSE)
img |
The image. |
segmentedImage |
The segmented image. |
classes |
The possible class labels. |
classColours |
The colors for the class labels. |
nblocks |
The image can be separated in several blocks, as zooming is not possible. |
labeledPoints |
Labeled cells from a previous training session. |
filename |
The table of labeled cells is saved at this location. |
filenameImage |
The image with the labeled cells is saved at this location. |
transformCoordinates |
deprecated |
Use the keys: a: In order to add a label to a cell. d: In order to delete a label from a cell. c: To switch between classes. q: To quit the interactive session. r: To refresh the session (labeled cells will be shown after refreshing)
A table with columns: index: the index of the cell in the segmented image. x: x-coordinate of the cell y: y-coordinate of the cell classCell: Label of the cell xLocal: Local x coordinate in the subimage(block) yLocal: Local y coordinate in the subimage(block) block: Block number in which the cell arises.
Henrik Failmezger
##---- Should be DIRECTLY executable !! ---- ##-- ==> Define data, use random, ##-- or do help(data=index) for the standard data sets. ## The function is currently defined as##---- Should be DIRECTLY executable !! ---- ##-- ==> Define data, use random, ##-- or do help(data=index) for the standard data sets. ## The function is currently defined as
This function takes the result of a call to correctCopyNumber and
plots the results.
plotCorrectedCN(CN, chr=NULL)plotCorrectedCN(CN, chr=NULL)
CN |
object result of a call to |
chr |
chromosome to plot. |
A panel with four plots is created. The top panel shows LRR (with DNAcopy segmentation overlayed) and BAF before correction and the bottom panel shows the plots after correction.
No value is returned.
Oscar M. Rueda, [email protected]
Yuan, Y et al. Quantitative image analysis of cellular heterogeneity in primary breast tumors enriches genomic assays. In prep.
LRR <- c(rnorm(100, 0, 1), rnorm(10, -2, 1), rnorm(20, 3, 1), rnorm(100,0, 1)) BAF <- c(rnorm(100, 0.5, 0.1), rnorm(5, 0.2, 0.01), rnorm(5, 0.8, 0.01), rnorm(10, 0.25, 0.1), rnorm(10, 0.75, 0.1), rnorm(100,0.5, 0.1)) Pos <- sample(x=1:500, size=230, replace=TRUE) Pos <- cumsum(Pos) Chrom <- rep(1, length(LRR)) z <- data.frame(Name=1:length(LRR), Chrom=Chrom, Pos=Pos, LRR=LRR, BAF=BAF) res <- correctCopyNumber(arr="Sample1", chr=1, p=0.75, z=z) plotCorrectedCN(res, chr=1)LRR <- c(rnorm(100, 0, 1), rnorm(10, -2, 1), rnorm(20, 3, 1), rnorm(100,0, 1)) BAF <- c(rnorm(100, 0.5, 0.1), rnorm(5, 0.2, 0.01), rnorm(5, 0.8, 0.01), rnorm(10, 0.25, 0.1), rnorm(10, 0.75, 0.1), rnorm(100,0.5, 0.1)) Pos <- sample(x=1:500, size=230, replace=TRUE) Pos <- cumsum(Pos) Chrom <- rep(1, length(LRR)) z <- data.frame(Name=1:length(LRR), Chrom=Chrom, Pos=Pos, LRR=LRR, BAF=BAF) res <- correctCopyNumber(arr="Sample1", chr=1, p=0.75, z=z) plotCorrectedCN(res, chr=1)
Procession of Aperio TX Slides.
processAperio(classifier=classifier,inputFolder=inputFolder,outputFolder=outputFolder,identifier=identifier,numSlides=numSlides,cancerIdentifier=cancerIdentifier,classOther = NA,maxShape=800,minShape=40,failureRegion=2000,slideToProcess=NA,KS=TRUE,colors=c(),classesToExclude=c(),threshold="otsu",numWindows=2,colorCorrection=FALSE,classifyStructures=FALSE,ksToExclude=c(),pixelClassifier=NA,densityToExclude=c(),numDensityWindows=32,resizeFactor=4,plotCellTypeDensity=TRUE,greyscaleImage=0,penClassifier=NULL,referenceHist=NULL,fontSize = 10)processAperio(classifier=classifier,inputFolder=inputFolder,outputFolder=outputFolder,identifier=identifier,numSlides=numSlides,cancerIdentifier=cancerIdentifier,classOther = NA,maxShape=800,minShape=40,failureRegion=2000,slideToProcess=NA,KS=TRUE,colors=c(),classesToExclude=c(),threshold="otsu",numWindows=2,colorCorrection=FALSE,classifyStructures=FALSE,ksToExclude=c(),pixelClassifier=NA,densityToExclude=c(),numDensityWindows=32,resizeFactor=4,plotCellTypeDensity=TRUE,greyscaleImage=0,penClassifier=NULL,referenceHist=NULL,fontSize = 10)
classifier |
The classifier. |
inputFolder |
The path to the image folder. |
outputFolder |
The path to the output folder. |
identifier |
The identifier of the files ("Ss" or "Da") |
numSlides |
The number of sections in the image. |
cancerIdentifier |
The identifier of the cancer class |
classOther |
deprecated |
maxShape |
Maximum size of cell nuclei |
minShape |
Minimum size of cell nuclei |
failureRegion |
minimum size of failure regions |
slideToProcess |
Set this parameter if only a certain slide should be processed |
KS |
Apply Kernel Smoother? |
colors |
Colors to paint the classes |
classesToExclude |
Which class should be excluded? |
threshold |
Which thresholding method should be used, "otsu" or "phansalkar" possible |
numWindows |
Number of windows to use for thresholding. |
colorCorrection |
deprecated |
classifyStructures |
Use hierarchical classification. If yes a pixel classifier has to be defined. |
ksToExclude |
These classes are excluded from kernel smoothing. |
pixelClassifier |
A SVM to classify pixel based on their color values. Needed if hierarchical classification should be applied. |
densityToExclude |
This class is excluded from cellularity calculation. |
numDensityWindows |
Number of windows for the density plot. |
resizeFactor |
Specifies the size of the cell density image. If this variable is not defined, the size of the thumbnail is used for the cell density image, else the size is calculated by size(thumbnail)*resizeFactor. The thumbnail is the small overview image, created by the Aperio software. |
plotCellTypeDensity |
Plot the density of different cell types? |
greyscaleImage |
Color channel of the RGB image that should be used for thresholding |
penClassifier |
Classifier to exclude low quality images (will be part of next release) |
referenceHist |
Colour Histogram of a reference image that can be used to calculate the quality of the recent image. (will be part of next release) |
fontSize |
will be part of next release |
The function processes images of Aperio TX scanners. The images have to be saved in the CWS format.
Four folders are created in the output folder.
Files |
Cellularity values and cell numbers are saved in the file |
classifiedImage |
Subimages with labeled tumour and non tumour cells |
tumourDensity |
Cancer heatmaps for every subimage |
cellCoordinates |
Coordinates and cell class for every cell in the subimage |
resizeFactor |
Size of the cellularity density image, calculated by size(thumbnail) * resizeFactor. Whereas the thumbnail is the small overview image produced by Aperio. |
Henrik Failmezger, [email protected]
#t = system.file("extdata", "trainingData.txt", package="CRImage") #read training data #trainingData=read.table(t,header=TRUE) #create classifier #classifier=createClassifier(trainingData,topo=FALSE)[[1]] #classify aperio #f = system.file("extdata", package="CRImage") #f=file.path(f,"8905") #dir.create("AperiOutput") #takes long time! f = system.file("extdata", package="CRImage") fc=file.path(f,"testClassifier") load(fc) fp=file.path(f,"pixelClassifier") load(fp) pixelClassifier=model pathToImage=file.path(f,"8905") pathToOutput="" #specify an output folder here #processAperio(classifier=classifier,inputFolder=pathToImage,outputFolder=pathToOutput,identifier="Da",numSlides=1,cancerIdentifier="c",maxShape=800,minShape=40,failureRegion=2000,slideToProcess=1,KS=FALSE,colors=c("white","green","blue","brown"),classesToExclude=c('a','l','nl'),threshold="otsu",ksToExclude=c('l','nl'),pixelClassifier=pixelClassifier,classifyStructures=TRUE,densityToExclude=c('a'),numDensityWindows=8,resizeFactor=1.5,plotCellTypeDensity=FALSE,greyscaleImage=1)#t = system.file("extdata", "trainingData.txt", package="CRImage") #read training data #trainingData=read.table(t,header=TRUE) #create classifier #classifier=createClassifier(trainingData,topo=FALSE)[[1]] #classify aperio #f = system.file("extdata", package="CRImage") #f=file.path(f,"8905") #dir.create("AperiOutput") #takes long time! f = system.file("extdata", package="CRImage") fc=file.path(f,"testClassifier") load(fc) fp=file.path(f,"pixelClassifier") load(fp) pixelClassifier=model pathToImage=file.path(f,"8905") pathToOutput="" #specify an output folder here #processAperio(classifier=classifier,inputFolder=pathToImage,outputFolder=pathToOutput,identifier="Da",numSlides=1,cancerIdentifier="c",maxShape=800,minShape=40,failureRegion=2000,slideToProcess=1,KS=FALSE,colors=c("white","green","blue","brown"),classesToExclude=c('a','l','nl'),threshold="otsu",ksToExclude=c('l','nl'),pixelClassifier=pixelClassifier,classifyStructures=TRUE,densityToExclude=c('a'),numDensityWindows=8,resizeFactor=1.5,plotCellTypeDensity=FALSE,greyscaleImage=1)
Thresholding method using mean and standard deviation.
SauvolaThreshold(allGreyValues)SauvolaThreshold(allGreyValues)
allGreyValues |
Vector of gray values. |
A threshold for the gray values is returned
The threshold.
Henrik Failmezger, [email protected]
J. Sauvola, M. Pietikainen, "Adaptive Document Image Binarization," Pattern Recognition, vol. 33, 225-236, 2000
createBinaryImage
f1= system.file("extdata", "exImg2.jpg", package="CRImage") print(f1) img=readImage(f1) print(img) #convert to grayscale imgG=EBImage::channel(img,'grey') #threshold value t=SauvolaThreshold(as.vector(imgG))f1= system.file("extdata", "exImg2.jpg", package="CRImage") print(f1) img=readImage(f1) print(img) #convert to grayscale imgG=EBImage::channel(img,'grey') #threshold value t=SauvolaThreshold(as.vector(imgG))
The function segments cells or cell nuclei in the image.
segmentImage(filename="",image=NA,maxShape=NA,minShape=NA,failureRegion=NA,threshold="otsu",numWindows=2, colorCorrection=FALSE, classifyStructures=FALSE,pixelClassifier=NULL,greyscaleImage=0,penClassifier=NULL,referenceHist=NULL)segmentImage(filename="",image=NA,maxShape=NA,minShape=NA,failureRegion=NA,threshold="otsu",numWindows=2, colorCorrection=FALSE, classifyStructures=FALSE,pixelClassifier=NULL,greyscaleImage=0,penClassifier=NULL,referenceHist=NULL)
filename |
A path to an image |
image |
An 'image' object, if no filename is specified. |
maxShape |
Maximum size of cell nuclei |
minShape |
Minimum size of cell nuclei |
failureRegion |
minimum size of failure regions |
threshold |
Thresholding method, "otsu" or "phansalkar" |
numWindows |
Number of windows to use for thresholding. |
colorCorrection |
deprecated |
classifyStructures |
Segment structures in the image, if yes a pixel classifier has to be defined |
pixelClassifier |
A SVM which classifies RGB color values in foreground and background. |
greyscaleImage |
Channel of the RGB image, to use for thresholding, if 0 use a joined greyscale image. |
penClassifier |
Classifier to exclude low quality images(will be part of next release) |
referenceHist |
Color histogram of a reference image, that can be used to estimate the quality of the recent image (will be part of next release) |
The image is converted to greyscale and thresholded. Clutter is deleted using morphological operations. Clustered objects are separated using watershed algorithm. Segmented Cell nuclei, which exceed the maximum size are thresholded and segmented again. Cell nuclei which fall below the minimum size are deleted. Dark regions which exceed the parameter failureRegion are considered as artefacts and deleted. If the parameters are not defined, the operations will not be executed. Features are generated for every segmented object.
A list is returned containing
image |
The original image |
segmented image |
The segmented image |
Henrik Failmezger, [email protected]
EBImage, 'http://www.bioconductor.org/packages/release/bioc/html/EBImage.html'
#segment image #f = system.file('extdata' ,'exImg.jpg',package='CRImage') #segmentationValues=segmentImage(f,maxShape=800,minShape=40,failureRegion=2000,threshold="otsu",numWindows=4) #image=segmentationValues[[1]] #segmentedImage=segmentationValues[[2]] #imageFeatures=segmentationValues[[3]]#segment image #f = system.file('extdata' ,'exImg.jpg',package='CRImage') #segmentationValues=segmentImage(f,maxShape=800,minShape=40,failureRegion=2000,threshold="otsu",numWindows=4) #image=segmentationValues[[1]] #segmentedImage=segmentationValues[[2]] #imageFeatures=segmentationValues[[3]]