Measure overlap

Authors: Robert Haase, Daniela Vorkel, April 2020

Source

This macro shows how to measure an overlap of two binary images in the GPU.

Get test data:

run("Blobs (25K)");
// open("C:/structure/data/blobs.gif");
input = getTitle();

Init GPU

and push image data to GPU memory:

run("CLIJ2 Macro Extensions", "cl_device=");
Ext.CLIJ2_clear();

// push data to GPU
Ext.CLIJ2_push(input);

// clean up ImageJ
run("Close All");

Create a mask by using two different threshold methods

Ext.CLIJ2_automaticThreshold(input, mask1, "Otsu");
Ext.CLIJ2_automaticThreshold(input, mask2, "MinError");

Ext.CLIJ2_pullBinary(mask1);
Ext.CLIJ2_pullBinary(mask2);

Measure the overlap between two masks

Ext.CLIJ2_getJaccardIndex(mask1, mask2, jaccardIndex);
Ext.CLIJ2_getSorensenDiceCoefficient(mask1, mask2, diceIndex);

// output result
IJ.log("Overlap (Jaccard index): " + (jaccardIndex*100) + "%");
IJ.log("Overlap (Sorensen / Dice coefficient): " + (diceIndex*100) + "%");

> Overlap (Jaccard index): 81.9294%
> Overlap (Sorensen / Dice coefficient): 90.0672%

At the end of the macro, clean up:

Ext.CLIJ2_clear();