R/classificationAperio.R

In CRImage: CRImage a package to classify cells and calculate tumour cellularity

classificationAperio <-
function(fileLocation,filename,pathToOutputFolderImgDir,classifier,pathToOutputFolderImgDirFiles,pathToOutputFolderImgDirImages,blockSlice,sliceColors,sizeO,index,cancerIdentifier,maxShape=800,minShape=40,failureRegion=2000,KS=TRUE,colors=c(),classesToExclude=c(),threshold="otsu",numWindows=2,classOther=NA,pathToOutputFolderImgDirStructures,colorCorrection=colorCorrection,classifyStructures=FALSE,pathToOutputFolderImgDirCells,ksToExclude=ksToExclude,pixelClassifier=pixelClassifier,densityToExclude=densityToExclude,numDensityWindows=32,plotCellTypeDensity=TRUE,greyscaleImage=greyscaleImage,penClassifier=NULL,referenceHist=NULL){
	widthO=as.numeric(as.character(sizeO[1]))
	heightO=as.numeric(as.character(sizeO[2]))
	pathToFile=file.path(fileLocation,filename)
#name of the image
	nameFile=strsplit(filename,"\\.")[[1]][1]
	message(paste("Start: ", nameFile))

# segmentation of the image
	imageData=try(segmentImage(filename=pathToFile,maxShape=maxShape,minShape=minShape,failureRegion=NA,threshold=threshold,numWindows=numWindows,colorCorrection=FALSE,classifyStructures=classifyStructures,pixelClassifier=pixelClassifier,greyscaleImage=greyscaleImage,penClassifier=penClassifier,referenceHist=referenceHist))

#original image
	img=imageData$image
#segmented image
	imgW=imageData$segmentedImage
#features of the nulcei
	cellData=imageData$features
#indices of all white pixel in the image
	indexWhitePixel=imageData$indexWhitePixel
#classify the nuclei, or is the image white?
	classify=imageData$classify

#is the image a failure image(no classification)
#section of the image
	slice=blockSlice[as.character(blockSlice$block)==nameFile,2]
	if(classify==TRUE){
		if(classifyStructures==TRUE){
#Structure Image
			structureImages=imageData$listStructures
			structures=imageData$structures
			writeImage(structureImages[[1]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img1.jpg",sep="")))
			writeImage(structureImages[[2]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img2.jpg",sep="")))
			writeImage(structureImages[[3]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img3.jpg",sep="")))
			writeImage(structureImages[[4]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img4.jpg",sep="")))
			writeImage(structureImages[[5]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img5_n.jpg",sep="")))
			writeImage(structureImages[[6]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img6_n.jpg",sep="")))
#print(structureImages[[5]])	
#						plot(structureImages[[5]])
		}else{
			structures=NA
		}
#classify the cells
		classValues=classifyCells(classifier,image=img,segmentedImage=imgW,featuresObjects=cellData,paint=TRUE,KS=KS,cancerIdentifier=cancerIdentifier,colors=colors,classesToExclude=classesToExclude,structures=structures,classifyStructures=classifyStructures,ksToExclude=ksToExclude)
		classes=classValues[[1]]
		
		writeImage(classValues[[2]],file.path(pathToOutputFolderImgDir,file.path(paste("section",slice,sep="_"),filename)))
		if(classifyStructures==TRUE){
			writeImage(classValues[[3]],file.path(pathToOutputFolderImgDirStructures,paste(nameFile,"_img5.jpg",sep="")))
			writeImage(classValues[[4]],file.path(pathToOutputFolderImgDir,file.path(paste("section",slice,sep="_"),paste(nameFile,"_structure.jpg",sep=""))))
			classProbs=classValues[[5]]
		}else{
			classProbs=classValues[[3]]
		}
		cellData=merge(classes,cellData)
		paintedCellsFinal=paintCells(imgW,img,as.character(cellData[,"classCell"]),cellData[,"index"],classifier$levels,colors=colors)
#writeImage(paintedCellsFinal,file.path(pathToOutputFolderImgDir,file.path(paste("section",slice,sep="_"),paste(nameFile,"_Final.jpg"))))
		
#calculation of cellularity, drawing of the heatmap
		cellValues=determineCellularity(cellData[,"classCell"],cellData,dim(imgW),img,imgW,indexWhitePixel,cancerIdentifier,classifier$levels,densityToExclude=densityToExclude,numDensityWindows=numDensityWindows,plotCellTypeDensity=plotCellTypeDensity)
		cellsSubImages=cellValues[[1]]
		cellsDensityImage=cellValues[[2]]
		
		cellValueTable=data.frame(t(cellsSubImages),stringsAsFactors=FALSE)
		colnames(cellValueTable)=names(cellsSubImages)
		write.table(cellValueTable,file.path(pathToOutputFolderImgDirFiles,file.path(paste("section",slice,sep="_"),nameFile)),sep="\t",row.names = FALSE)
		cellDataProbs=data.frame(cellData,classProbs,stringsAsFactors=FALSE)	
		write.table(cellDataProbs,file.path(pathToOutputFolderImgDirCells,paste(nameFile,".txt",sep="")),sep="\t",row.names=FALSE,quote = FALSE)
		######### READ WRITE qualityTable #############
		qualityTable=c(nameFile,NA,NA)
		qualityTableNames=c("Image","pen","distToRef")
		if(!is.null(penClassifier)){
			qualityTable[2]=as.character(imageData$penLabel)
		}
		if(!is.null(referenceHist)){
			qualityTable[3]=imageData$distToRef
		}
		existingQualityTable=read.table(file.path(pathToOutputFolderImgDir,"subimageQuality.txt"),header=TRUE)
		#check if any subimage was written in quality table yet
		if(colnames(existingQualityTable)[1] != "Image"){
			qualityTable=t(qualityTable)
			colnames(qualityTable)=t(qualityTableNames)
			write.table(qualityTable,file=file.path(pathToOutputFolderImgDir,"subimageQuality.txt"),sep="\t",row.names=FALSE,quote=FALSE)
		}else{
			wholeQualityTable=rbind(existingQualityTable,qualityTable)
			write.table(wholeQualityTable,file=file.path(pathToOutputFolderImgDir,"subimageQuality.txt"),sep="\t",row.names=FALSE,quote=FALSE)
		}
		
		message("Cell values written")	
		message("Write tumour density heatmap")
		writeDensityImage(pathToOutputFolderImgDir,cellsDensityImage,sizeO,blockSlice,"smallDensityImage.jpg",nameFile)
		
		if(plotCellTypeDensity==TRUE){
			cellsTypeImage=cellValues[[3]]
			writeDensityImage(pathToOutputFolderImgDir,cellsTypeImage,sizeO,blockSlice,"cellTypeImage.jpg",nameFile)
		}
###########################################################################################################################			
		
		message(paste(nameFile," finished"))
		cellData
	}else{
		
###########################################################################################################################			
		message(paste(nameFile," finished"))
		classifiedCells=data.frame(stringsAsFactors=FALSE)
	}
}