Homec4science
Diffusion BIOP Run Macro (master) Browse

ijp-biop-run-macro/tests/CostesSignificanceTest.java
7245282e9c63master

Last Change
Raw File

CostesSignificanceTest.java
View Options

package tests;
import static org.junit.Assert.assertTrue;
import gadgets.DataContainer;
import mpicbg.imglib.image.Image;
import mpicbg.imglib.type.numeric.real.FloatType;
import org.junit.Test;
import algorithms.AutoThresholdRegression;
import algorithms.MissingPreconditionException;
import algorithms.PearsonsCorrelation;
/**
* This class contains JUnit 4 test cases for the Costes
* statistical significance test.
*
* @author Dan White & Tom Kazimiers
*/
public class CostesSignificanceTest extends ColocalisationTest {
/**
* This test checks the Costes statistical significance test implementation
* by artificially disturbing known colocalisation. It simulates what Costes
* describes in figure 3 of section "Simulated data". An image representing
* colocalised data is generated. This is put onto two random Perlin noise
* images. A smoothing step is applied after the combination. With the two
* resulting images the Costes calculation and shuffling is done. These steps
* are done multiple times, every time with an increasing percentage of
* colocalised data in the images. As stated by Costes, colocalisation data
* percentages above three percent should be detected (P value > 0.95. This
* is the assertion of this test and checked with every iteration. Percentages
* to test are calculated as percentage = 10^x. Five iterations are done,
* increasing "x" in steps of 0.5, starting at 0. The test uses circles with
* a diameter of 7 as objects (similar to Costes' paper, he uses 7x7 squares).
*/
@Test
public void backgroundNoiseTest() throws MissingPreconditionException {
final int width = 512;
final int height = 512;
final double z = 2.178;
final double scale = 0.1;
final int psf = 3;
final int objectSize = 7;
final double[] sigma = new double[] {3.0,3.0};
for (double exp=0; exp < 2.5; exp=exp+0.5) {
double colocPercentage = Math.pow(10, exp);
Image<FloatType> ch1 = TestImageAccessor.producePerlinNoiseImage(
new FloatType(), width, height, z, scale);
Image<FloatType> ch2 = TestImageAccessor.producePerlinNoiseImage(
new FloatType(), width, height, z, scale);
/* calculate the number of colocalised pixels, based on the percentage and the
* space one noise point will take (here 9, because we use 3x3 dots)
*/
int nrColocPixels = (int) ( ( (width * height / 100.0) * colocPercentage ) / (objectSize * objectSize) );
// create non-smoothed coloc image. add it to the noise images and smooth them
Image<FloatType> colocImg = TestImageAccessor.produceNoiseImage(
width, height, objectSize, nrColocPixels);
TestImageAccessor.combineImages(ch1, colocImg);
ch1 = TestImageAccessor.gaussianSmooth(ch1, sigma);
TestImageAccessor.combineImages(ch2, colocImg);
ch2 = TestImageAccessor.gaussianSmooth(ch2, sigma);
DataContainer<FloatType> container
= new DataContainer<FloatType>(ch1, ch2, 1, 1);
PearsonsCorrelation<FloatType> pc
= new PearsonsCorrelation<FloatType>(PearsonsCorrelation.Implementation.Fast);
AutoThresholdRegression<FloatType> atr
= new AutoThresholdRegression<FloatType>(pc);
container.setAutoThreshold(atr);
atr.execute(container);
try {
pc.execute(container);
}
catch (MissingPreconditionException e) {
/* this can happen for random noise data in seldom cases,
* but we are not after this here. The cases that are
* important for Costes work well, but are again sanity
* checked here.
*/
if (pc.getPearsonsCorrelationValue() == Double.NaN)
throw e;
}
algorithms.CostesSignificanceTest<FloatType> costes
= new algorithms.CostesSignificanceTest<FloatType>(pc, psf, 10, false);
costes.execute(container);
// check if we can expect a high P
if (colocPercentage > 3.0) {
double pVal = costes.getCostesPValue();
assertTrue("Costes P value was " + pVal, pVal > 0.95);
}
}
}
}
c4science · Help