macro "Calculate by Image" {

	DefCats = newArray("Default", "Axon", "AIS", "Distal Axon", "Dendrite", "Synapse1", "Synapse2", "Axon (NT)", "AIS (NT)", "Distal Axon (NT)", "Dendrite (NT)", "Synapse 1 (NT)", "Synapse 2 (NT)", "Primary", "Secondary", "Tertiary", "Cat0", "Cat1", "Cat2", "Cat3", "Cat4", "Cat5", "Cat6", "Cat7");
	Path = File.openDialog("Choose Results table");
	Results = File.openAsString(Path);
	RName = File.getNameWithoutExtension(Path);

	// Retrieve the Labels column, the mean intenisty column and the area column
	IMAGENAMES = getColumn(Results,"Slice");
	TYPES = getColumn(Results,"ROI type#");
	CATS = getColumn(Results,"ROI type");
	if (IMAGENAMES[0] == -1 || TYPES[0] == -1 || CATS[0] == -1) exit("One of the required column doesn't exist!");
	// Mean intensity used is the background-corrected mean (output by the "Measure Intensities" macro) or if no corrected mean column is found, the raw mean intensity.
	CORRMEAN = getColumn(Results,"Corr Mean");
	if (CORRMEAN[0] == -1) CORRMEAN = getColumn(Results,"Raw Mean");
	if (CORRMEAN[0] == -1) exit ("Mean intensity column doesn't exist!");
	AREA = getColumn(Results, "Area");
	if (AREA[0] == -1) exit ("Area column doesn't exist!");

	// gets the number of images (defines the per-image arrays length)
	U=getUnique(IMAGENAMES);


	UT = getUnique(CATS);
	UCATS = newArray(UT);
	UCATS[0] = CATS[0];
	u = 0;
	for (i = 1; i < CATS.length; i++) {
		if (CATS[i] != CATS[i-1]) {
			u++;
			UCATS[u] = CATS[i];
		}
	}

	

//	Dialog.create("Calculate Ratios Options");
//	Dialog.addChoice("Numerator (N)", UCATS, UCATS[0]);
//	Dialog.addChoice("Denominator (D)", UCATS, UCATS[UCATS.length-1]);
//	Dialog.show();

//	CatNum = Dialog.getChoice();
//	CatDen = Dialog.getChoice();

//	CatNumN = getIndex(DefCats, CatNum);
//	CatDenN = getIndex(DefCats, CatDen);

	// Define all per-image arrays
	// Names : name of the image
	// SumType: summed integrated intensities
	// LenType: summed areas
	// MoyType: SumType / LenType, i.e. overall mean intensity for all ROIs pooled
	// MoyType does not depend on the spatial scale as it is present as a multiplicative factor in SumType and LenType
	// Type: summed mean intensities
	// NType: number of ROIs
	// NMoy: Type / NType, i.e. average of mean intensities for each ROI as individual mean intensities
	// NType does not depend on the spatial scale
	// Ratio: ratio of MoyTypes, i.e. ratio of overall mean intensity for all ROIs pooled
	// NRatio: ratio og NMoy, i.e. ratio of averages for each ROI as individual mean intensities


	// Generate the Results table
	title1 = RName + "_byIm";
	title2 = "[" + title1 + "]";
	f = title2;
	if (isOpen(title1))
		print(f, "\\Clear");
	else
		run("New... ", "name="+title2+" type=Table");
	Headings = "\\Headings:n\tImage\tROI Type #\tRoi Type\tSummed ID\tSummed Area\tOverall MI\tSummed MI\tNumber\tAverage MI";
	print(f, Headings);

	t = 0;


	// loop on all unique categories
	for (c= 0; c < UT; c++ ) {
			
		CatNum = UCATS[c];
		CatNumN = getIndex(DefCats, CatNum);


		Names = newArray(U);	
		SumTypeA = newArray(U);
		LenTypeA = newArray(U);
		MoyA = newArray(U);
		TypeA = newArray(U);
		NTypeA = newArray(U);
		NMoyA = newArray(U);


		// Initialization for i = 0 (first Results Table line)
		Names[0] = IMAGENAMES[0];
		// Stores values in the first per-image slot if the ROI in the current line is #A
		if (TYPES[0] == CatNumN){
			LenTypeA[0] = 0 + AREA[0];
			SumTypeA[0] = 0 + CORRMEAN[0] * AREA[0];
			TypeA[0] = 0 + CORRMEAN[0];
			NTypeA[0] = 1;
			typeflag = 1;
			}
	
		// Loop on all following Results Table lines
		UIndex = 0;
		for (j = 1; j < IMAGENAMES.length; j++) {
			// This detects if it's a new image (new averages) and iterates UIndex to store in the next slot in the per-images array
			if (IMAGENAMES[j] != IMAGENAMES[j - 1]) {
				UIndex = UIndex+1;
				Names[UIndex] = IMAGENAMES[j];
			}
			// Stores values in the current per-image slot if the ROI in the current line is #A
			if (TYPES[j] == CatNumN) {
				LenTypeA[UIndex] = LenTypeA[UIndex] + AREA[j];
				SumTypeA[UIndex] = SumTypeA[UIndex] + CORRMEAN[j] * AREA[j];
				TypeA[UIndex] = TypeA[UIndex] + CORRMEAN[j];
				NTypeA[UIndex] = NTypeA[UIndex] + 1;
			}
			
		}
	
		// Loops on all per-image slots
		for (j=0; j<Names.length; j++) {
	
			// Calculate the overlall MIs and the ratio of overall MI
			if (LenTypeA[j] != 0) MoyA[j] = SumTypeA[j] / LenTypeA[j];
			else MoyA[j] = NaN;
	
			// Calculate the average MIs and the ratio of average MIs
			if (NTypeA[j] != 0) NMoyA[j] = TypeA[j] / NTypeA[j];
			else NMoyA[j] = NaN;
		}


		for (n = 0; n < Names.length; n++) {
			ResultsLine = d2s(t + 1, 0) + "\t" + Names[n] + "\t" + CatNumN + "\t" + CatNum + "\t" + SumTypeA[n] + "\t" + LenTypeA[n] + "\t" + MoyA[n] + "\t" + TypeA[n] + "\t" + NTypeA[n] + "\t" + NMoyA[n];
			print(f, ResultsLine);
			t++;
		}

	}

// This function returns all values in a Results Table column as an array (or -1 if the Header is not found)
function getColumn(TableString, Header) {
	Rows = split(TableString, "\n");
	Labels = split(Rows[0], "\t");
	IndexCol = -1;
	for (i = 0; i < Labels.length; i++) {
		if (Labels[i] == Header) {
			IndexCol = i;
		}
	}
	if (IndexCol == -1) return -1;
	Column = newArray(Rows.length - 1);
	for (i = 1; i < Rows.length; i++) {
		CurrentLine = split(Rows[i], "\t");
		Column[i-1] = CurrentLine[IndexCol];
	}
	return Column;
}


// This function returns the number of unique elements in a sorted array
function getUnique(SortedArray) {
	Array.sort(SortedArray);
	DiffNumber = 1;
	for (i = 1; i < SortedArray.length; i++) {
		if (SortedArray[i] != SortedArray[i - 1]) {
			DiffNumber = DiffNumber + 1;
		}
	}
	return DiffNumber;
}

function getIndex(array, el) {
	for (i = 0; i < array.length; i++) {
		if (array[i] == el) return i;
	}
	return -1;
}