SEM Particle Size Analysis
The size distribution of microscopic objects may be determined through analysis of SEM images using readily available software. This method employs ImageJ and includes an ImageJ macro designed to analyze all of the images in a selected folder.
This procedure has been used with inorganic materials (fly ash) that are relatively easy to prepare and produce high contrast images. Powder was dispersed onto carbon tape placed upon 12mm stubs. The samples were carbon coated and BSE images were taken at the magnification and resolution required to discriminate particles of the desired minimum size, in this case, about 0.8 microns.
There is a trade off between magnification and image resolution: Lower magnification with higher resolution images (larger matrix) requires fewer images to achieve the desired number of particles for the analysis (in this case, a minimum of 10,000 particles). Higher magnification with lower image resolution results in a greater number of images required (doubling the magnification and halving the resolution = 4x more images required). It was found that lower magnification, high resolution images produced less sharp particles and the decision was made to image at higher magnification.
Contrast and brightness were set on microscope control to maximize the signal captured whereas contrast and brightness in image acquisition was set to slightly saturate particles and truncate low intensity pixels to black. This produced images that were not necessarily publication quality, but of sufficient contrast to ensure good thresholding during image analysis.
Particle Size Analysis with ImageJ
A macro intended for use with images captured by the EDAX Phoenix Image Acquisition and Mapping software installed on the ESEM analyzes every image in a selected folder. It interrogates the image parameter text file associated with an image and adjusts the image aspect ratio to produce a square pixel based upon the values found in the parameter file. It also calculates particle dimensions in microns based upon the pixel dimensions found in the parameter file. It therefore will produce results that are consistent and "correct" regardless of the magnification of individual images - a variety of magnifications may be used and consistent results will be produced.
The macro requires that the multithresholder plugin be installed. It produces an image showing the particle outlines overlaying the resized original image so that it's performance can be monitored. It also produces a comma separated values report listing shape parameters, Feret's diameters, fitted ellipse dimensions and volume for every particle found. It also allows for binning of particles based upon the minor axis dimension of a fitted ellipse, in microns, producing a histogram table of bin ranges, counts, cumulative volume and volume fraction for each bin range.
The macro will create a uniquely named set of images and report each time it is run if it finds previous results in the folder. This allows easier side-by-side comparison of results from changes in user-selectable analysis parameters.
Since the macro reads the image parameter text file produced by the EDAX image acquisition software, those text files must also be stored in the same folder as the images to be analyzed.
The macro presents the user with a dialog box having user-selectable values used during the analysis. Default values are those found to have worked best for the research the macro was written for and may not produce suitable results with your sample images.
Following is the step-by-step methodology used by the macro to count particles.
- Reduce the y dimension by the amount required to produce a square pixel having square micron dimensions.
- Invert the image.
- Subtract background (optional).
- Gaussian blur (optional).
- Threshold the image using one of the methods built into multithresholder.
- Fill holes.
- Watershed (optional).
- Analyze particles including or excluding particles intersecting the edge of the image.
- Bin results per a comma separated list of bins (optional).
- Create and save an image composite of the resized original image overlain by the particle outlines.
- Compile results in a single csv file in the same folder as selected for analysis.
Obtaining and Using the Macro
The macro is available as a compressed (zip) file here. Download the file, expand and copy into the macro folder in the ImageJ folder. Install or run the macro under the ImageJ Plugins menu item.
The macro produces an image of the resized original image with particle outlines overlaying it. It also produces a report in csv format that is readily opened and further analyzed in Microsoft Excel. All of the files are saved in the same folder as was selected for analysis. Results from each particle analyzed is saved in the report file along with binned results, if that option was selected.
- Determine the minimum number of particles to be analyzed and the minimum size of particle to analyze.
- Determine the appropriate magnification and image resolution (see above).
- Collect a number of images of the sample.
- Using ImageJ and the macro, analyze the images without binning the results.
- View each of the composite images (they will be named "Particles_X from image.bmp") to ensure that the particles are being properly discriminated. Make adjustments to the user-selectable parameters and rerun the macro, reviewing the results.
- Once the desired results are achieved, determine the approximate number of images required to meet the minimum number of particles for analysis: total particles required / (total particles/number of images).
- Collect the additional images required.
- Analyze the images with the macro this time selecting Bin Results and entering sequential bin values separated by commas.
- Perform post-analysis in MS Excel.
The Excel spreadsheet used for presenting the data is available here. Copying the binned results from the macro into the green area in the Data sheet will result in graphical and tabular representations of the data.
Notes and Tips
- With some images, subtracting the background may result in additional noise that isn't removed upon thresholding. The latest macro will disable background subtraction if a value of 0 is entered in the rolling ball diameter text box. This is a good first-thing-to-try if there are lines of noise being counted as particles.
- The thresholding methods that thus far have proven to be effective are 1) Triangle and 2) Minimum. It seems that minimum thresholding contracts the particles somewhat whereas the triangle method results in a slightly dilated particle and is also subject to not eliminating all the background texture from some images. Try different thresholding methods coupled with the above to tune the macro.
- Watershed works well with particles having a relatively square aspect ratio. Long, narrow particles tend to be broken up into a bunch of small, roughly square particles. It's really only effective with agglomerations of only a few (3-5) particles.