Online Manual for the
WCIF-ImageJ collection

ImageJ
Image Processing and Analysis in Java

 
 

5. Particle Analysis

Particle counting can be done automatically if the specimen lends itself to it, i.e. the individual particles can touch – but not too much! If automatic particle counting cannot be done, ImageJ can facilitate manual counting with the “Point Picker” or “Cell counter” plugin.

5.1 Automatic Particle counting

The biggest issue is one referred to as “segmentation” which is to distinguish the object from the background. Once the objects has been successfully segmented, they can then be analysed.

 RAW Threshold Watershed “AnalyzeParticles” Merge

5.1.1 Setting a threshold

    5.1.1.1 Manual thresholding

Automatic particle analysis requires the image to be a “binary” image i.e. black or white. The software needs to know exactly where the edges are to perform morphology measurements. A “threshold” range is set and pixels in the image whose value lies in this range are converted to black; pixels with values outside this range are converted to white (or vice-versa depending on what the user requests).

There are several ways to set thresholds. For monochrome images it is most simply done via thee menu command “Image/Adjust/Threshold”. While the threshold is being set, the pixels within the threshold range are displayed in red. The user then changes then “Apply” the threshold and the image will be converted to a binary image.

 

For colour images, setting the threshold is a bit more complicated as each of the red, green and blue channels require different threshold ranges. The plugin “Plugins/Colour functions/Threshold Colour can be used to do this most easily. This opens a fairly complicated and extensive control window. The simplest way to perform a colour threshold is to use one of the selection tools to select one of the objects of interest, and then click the “Sample” button in the “Threshold Colour” control window. This should remove the pixels in the image that do not have pixels of the same colour as those in the selection. The thresholding can be manually tweaked with scroll bars in the histogram panes. It may be easier to conceptualise what you are doing at this point by switching the Control window from the default Hue, Saturation and Brightness colour model to red, green blue model by switching the option buttons at the bottom of the control window. The image can be displayed as a binary image with the “Threshold” button, and converted to a binary image via the menu command “Image/Type/8-bit” (?!).

 5.1.1.2 Automatic thresholding

One concern with setting a threshold manually is the potential to introduce user-bias. There are many algorithms described to automatically calculate the threshold. A evaluation of over 40 of these can be found in this paper: Sezgin & Sankur. 2004.  Survey over image thresholding techniques and quantitative performance evaluation. Journal of Electronic Imaging, 2004. (13:146-165). WCIF ImageJ has several plugins found in the menu "Plugins/Segmentation" for automatic calculation of an image threshold. These include Otsu's thresholding, maximum entropy threshold, mixture modelling thresholding. There is also k-means clustering which can segment an image in to multiple bands and can be used with colour images.

5.1.2 Watershed separation

Objects in a binary image that are slightly overlapping may be separated using the menu command “Process/Binary/Watershed”. This segmentation process is nicely illustrated with this figure from the ImageJ documentation web pages (http://rsb.info.nih.gov/ij/docs/index.html).

The image first needs to be converted to binary (set via thresholding). The black pixels are then replaced with grey pixels of an intensity proportional to their distance from a white pixel (i.e. black pixels close to the edge are light grey, those closer to the ‘middle) are nearer black. This is the Euclidian distance map (EDM). From this it calculated the centres of the objects the ultimate eroded points (UEPs), i.e. points that are equidistance from the edges. These points ere then dilated until the meet another black pixel, then a watershed line is drawn.

5.1.3 Analyse Particles

Once the image has been segmented, the menu command “Analyze/Analyze particles” can be used to obtain various information regarding particle size and numbers.

Set the minimum size and maximum size to exclude objects that appear in the binary image that are clearly not objects of interest. Select the “Show: Outlines” option to display an image of the detected objects. This can then be merged with the original for presentation purposes. Invert the outline image and use the menu command “Plugins/Colour functions/Colour merge” to add the original and outline image.

The particle analysis can be automated via plugins or macros once the correct threshold value and particle size range has been determined for your objects of interest. See the “Plugins/Particle Analysis/ Nucleus counter” plugin and its source code to customise it to your images.

 

5.1.4 Nucleus Counter

This plugin automates many of the steps discussed above. The user can:
1. Enter the size range which should be counted.
2. Select the automatic thresholding method. This can be either "Current", "Otsu", Maximum Entropy", "Mixture Modelling" or k-means" clustering. "Current" uses the threshold that has been set manually, see above.
3. Opt to perform a background correction.
4. Opt to perform a "Smooth" filter.
5. Opt to perform a watershed separation.
6. Opt to add the particles to ROI manager
7. Opt for a summary.

Other options can easily be added on request.

The count, area and average size are returned as a text window and the outlined particles are overlaid on a duplicate of the original image.

 

5.1.5 Particle tracking

The “MTrack2” plugin (“Plugins/Particle counting/MTrack2 ”) tracks particles and reports their coordinates at each time-point and their total path length.

   

 Raw Background corrected MultiTracker Motion paths
 Smoothed, Binary result of two objects

5.2 Manual Counting

You can use the ImageJ toolbar Crosshair (mark and count) tool to count particles. However, a greater degree of control can be accessed with the “Point Picker” or “Cell counter” plugin.

5.2.1 Cell Counter

When then “Plugins/Particle counting/Cell Counter” plugin) is run, a new Image window is generated with the original image plus four buttons along the bottom (Red type, Green type, Blue type and Yellow type). Clicking on a button changes the colour of the marking tool. The results table keeps a tally of the number of each of the cell types marked. Clicking the Results button will generate a summary of the data.

 

 

 

5.2.2 Point Picker

Plugins/Particle counting/PointPicker ”. Running this plugin will change the ImageJ toolbar to the PointPicker toolbar (below).

 

 

Crosses are reversibly overlaid on to the image, incrementally changing colour (this differs from the Crosshair tool which irreversible stamps a point in the image). With the Point Picker tool, crosses can be moved and deleted. Once marking is complete, the cross coordinates can be accessed via the button.

This produces the “Point List” dialog allowing you to Show, Save or Open the points’ coordinates. Clicking “Show” displays the coordinates in the Results window, where they can be saved or copied to the system clipboard. Once finished, you can return to the regular ImageJ toolbar by clicking the  button.