Finds and counts white spots in a stack and measures volume and area of them
This package contains utilities to count and measure lipid droplets or any other bright spots in a 3D stack.
It also contains:
Copy the file Droplet_Counter.jar into your Imagej plugins directory. The jar file also contains the GPL-licensed source code. You can open the jar file with any archive manager, e.g. 7zip.
Author: Samuel Moll (email@example.com)
Source Code GitHub
write bug reports, feature requests and questions to: firstname.lastname@example.org
What you need:
What the Filterstacker actually does:
The “Filterstacker” plugin actually only does a 3D-bandpass on your stack. It will first 3D-blur the input image with a filter size of “maximum feature size” and substract the blurred image from the original. It will then eliminate small features (like noise) by blurring again with a small filter the size of “minimum feature size”. The “Z/X aspect ratio” compensates for different lateral and vertical resolutions.
It will then call the “Watershed 3D” plugin that finds all local maxima (white spots) and grows regions around them, so that each region only contains one maximum. It will output a new stack where each region is labeled with a unique color. The radii control how far apart the maxima must be until they are regarded as only one maximum. Setting this to higher values makes the watershed transform more resistant to noise while increasing the running time (higher radii than 3 take VERY long) and the probability that two spots that are close together are combined into one region. Big radii also make the edges of the regions more fuzzy. Setting all radii to 2 is a good compromise. If the “Invert” option is unchecked, the plugin will find minima (black spots) instead of maxima.
Finally the Filterstacker calls the “Segment Analyzer” plugin that takes as input an image and a mask (=the watershed transform). The Segment Analyzer decides which regions contain a droplet by thresholding (explained below). It then does a FWHM threshold on all regions that passed the previous thresholding test and measures the volume, position and surface area of each droplet. The region thresholding is very simple at the moment. It only considers the maximal and summed (over the whole region) brightness (the “maximum threshold” and “area threshold” parameters). If both values are above the respective thresholds, the region is considered to contain a droplet. The Segment Analyzer has a preview. Because of ImageJ design restrictions, the stack has to be navigated with the “Slice” slider. Red marks particles, the borders between particles are green. If there is a green line through one of your particles, lower the “connect threshold”.
If you publish a paper relying on results obtained with ImageJ and this plugin, and want to add a citation to your paper, you could e.g. follow these recommendations for citing.
DOI for citing: DOI 10.5281/zenodo.2581434
For example: Lets say you measure 30 images in the z-plane with a resolution of 512×512 each. The imaged region is 10um x 10um big, and the 30 z-planes span a distance of 2um, i.e. the lowest image of the z-stack (number 1) is right on your glass substrate and the highest image (number 30) is 2um above the glass substrate. Then your pixels/voxels have the following dimensions:
x: 10um / 512 = 19.5nm y: 10um / 512 = 19.5nm z: 2um / 30 = 66.7nm
Normally, your imaging system gives you the width of a pixel (x and y values) and the stack step size (z value) directly. In our example the z-resolution is considerably worse than the x- and y-resolutions. (This is very common with CLSM images) The z-ratio is then just z divided by x, i.e. 66.7nm/19.5nm = 3.4. This is what should be put in the confusingly-named “Z/X aspect ratio” field. That said, you can experiment a bit with this setting (set it to higher/lower values than the theoretically calculated ones) to get optimal object separation.