Using Image Analysis to Differentiate Particle Size and Shape in Drug Products with Multiple Components

Published on: 
In the Lab eNewsletter, Pharmaceutical Technology's In the Lab eNewsletter, December 2023, Volume 18, Issue 12

Image analysis is crucial for particle analysis of complex finished drug products.

A significant number of resources and guidance literature are available for particle size determination of raw materials used in the drug product manufacturing process. However, when multiple components, such as excipients and APIs are added to a finished product, the use of the most suggested techniques may not be effective in providing particle size information. This is especially true if assessment of component homogeneity or the effects of final processing on each material is required.

Popular ensemble techniques (such as laser diffraction) or particle counters (including single particle optical sensing [SPOS] instruments) cannot differentiate between API ingredients, or any other materials added to the final product during a process. With these common techniques, there may be inherent difficulty finding a suitable carrier medium or methodology to sufficiently work with all the materials present given the materials’ chemical composition. In addition, a single broad distribution is often produced by these tests making the resolution of individual subpopulations generally rather impossible (1,2).

To overcome the obstacles typically seen with multicomponent finished products, the use of a non-ensemble technique, such as automated static image analysis, can provide particle characterization solutions under certain circumstances by using software filters and light polarization to differentiate between distinct populations, whether based on size, shape, or other material characteristics.

Understanding how static image analysis works

Static image analysis technique typically consists of an automated microscope that uses a series of objectives of varying magnifications, a motorized stage, and a digital camera with accompanying software to capture images of particles.These instruments determine the size of an individual pixel for the chosen magnification and creates a projected two-dimensional image of the particle. They then convert the pixels of the two-dimensional image into the most commonly cited shape—a circle that would have the same pixel area as the two-dimensional image, thus reporting the circular equivalent (CE) diameter (3).

While this is useful and similar in respects to other particle size determination techniques, image analysis instruments are capable of obtaining further information regarding shape factors to provide a more adequate description of typical non-spherical particles that would be limited to simpler techniques, such as laser diffraction. Examples of useful classifications are length, width, circularity, convexity, and solidity with more parameters described in ISO chapter 13322-1 (3). Like many other particle characterization techniques, the typical values (D10, D50, D90, and mean) can be reported for any of these parameters and also on a number and volume basis.

In addition, the use of a polarizer during analysis as well as software filters (removing populations of particles from the distribution based on user-entered classification limits) can also help differentiate populations of materials which lends itself well to addressing the initial problem of being able to properly evaluate a material with multiple components (4).

Samples that can benefit from this technique can include topical creams and ointments, oral suspensions, gels, and injectables. The use of either classification filters and/or polarized light, can help to work around components within a product with different dissolution rates, varying concentrations, and particles with similar particle size ranges.

Differentiating populations in a real-world sample

The use of these advanced classifications can be demonstrated with a typical undiluted topical cream sample containing emulsion droplets and crystalline API. Figure 1 shows a general microscopic image of a topical cream containing emulsion droplets and crystalline API particles in low concentration. The matrix is thick, and the emulsion droplets appear to be touching. Sporadically, API crystals can be observed.

Classifying both populations for particle size in this product can be done, but two different preparations and analyses must be conducted in order to representatively capture them. This is due to the fact that even though the software offers many options for classifying multiple populations of particles, the results will only be as good as the microscope slide preparation and the ability to differentiate the particles from the background, or thresholding, which can be affected by the inherent shape, translucency, and size of objects present in the field of view. A single user-defined setting may not work for more than one type of particle, and the thickness of the matrix the particles are sitting in may also compound the imaging process.

Another thing to be aware of is that image analysis, like most techniques, will incorrectly assume two particles touching to be one particle; therefore, in this sample, dilution should be performed in order to obtain an accurate particle size distribution for the droplets. However, upon dilution, the occurrence of the API is significantly further reduced, potentially biasing the crystalline material population if the user was to proceed with the analysis (Figure 2). Care should be taken to ensure dilution will not adversely affect the particles or droplets.


Utilizing software filters to accurately assess droplet populations

Once proper methodology is obtained using good microscopy and dispersion practices, further classification of each population can commence. A way to classify the emulsion droplets minimizing running into multiple particles being counted as one in addition to dilution, would be to utilize the software filters, such as aspect ratio and circularity. Aspect ratio provides the width to length ratio that describes the elongation of the particle. Values within the software range from 0 to 1 where a rod would have a value closer to 0 compared to a circular object. In comparison, circularity provides the ratio of the circumference of a circle equal to the object’s projected area to the perimeter of the object. Values within the software range are also from 0 to 1 where a perfect circle has a value of 1 (3).

By placing a restriction on what particles are detected based on these filters, particles that don’t meet these requirements are immediately removed or included from the distribution based on the set criteria. Figure 3 is an example of the types of particles detected after the application of user-defined filters.

Utilizing polarizers to accurately assess crystalline API populations

A second analysis quantifying the crystalline API can then be performed to obtain distribution statistics on only this population. As noted, further dilution of the sample would not be recommended as this API population is already low in number. It is possible to generated biased results on an analysis with so few particles counted as the sample size is not representative. The original undiluted sample is highly concentrated, so instead of relying on filters to provide classification of the target API population in this instance, the use of a polarizer built into the microscope can be utilized.

For simple definition purposes within the scope of this article, a polarizer is an optical filter which allows light waves of a specific polarization to pass through while blocking light waves of other polarization planes. The polarizers work by using the birefringent properties—defined as the optical property of a material having a refractive index that depends on the polarization and propagation direction of light (5)—of crystalline material to further illuminate material while ignoring non-crystalline material. An example of the undiluted sample under episcopic light (top light) in comparison to the same field of view under diascopic light (bottom light) with a polarizer can be seen in Figure 4.

While the “as-is” image looks congested, by placing it under polarized light, the droplets are no longer relevant and only the crystalline API is apparent. The distribution results are then only reflective of the API particulate (See Figure 5 for images of captured crystalline particles).

Distribution statistics can be provided for both analyses, and the distributions can be overlaid to give a better picture of the compositional nature of the materials present in the finished ointment product, which can be seen in Figure 6 with an overlay of droplet and API distributions.

The benefit of developing multiple methodology that classifies the entire
composition of a final product is that monitoring processes and the effects this has on components can often be done on a single instrument. As discussed, shape distinctions can be used to distinguish the types of materials, but determining particle size and shape based on other classifications can also be done. A short list could also include using variations in opacity, texture, and reflective intensity of the particles present.

Some image analysis systems also can be equipped with a Raman spectroscopy unit, and, thus, in certain cases, the use of morphologically directed Raman spectroscopy can be used to assist in spot-checking the captured image analysis results. Raman spectroscopy works by randomly selecting a subset of particles that have already been classified by image analysis and allowing the user to chemically identify multiple components and their relative proportions within the product (6). This can be very useful when trying to discern two or more material components that share similar shape characteristics and size.

There are some limitations to this technique, however, such as if the medium in which the final product is suspended is not suitable for Raman, then it would not produce background signal and may swamp out the signals coming from the particles. The spectra of the isolated raw materials are used as a comparison reference to identify the particles in the finished product.


Many options are available for determining the particle size distribution of a multi-component product; however, few are able to provide the statistics and particle size information on individual materials. Automated static image analysis is a technique that allows for the classification of individual populations under certain conditions with the use of software filters and light polarizers. The success rate of characterization is dependent on the medium in which the particles are suspended, in chemical composition of ingredients, and overall concentration of the preparation, but is a more viable option of the standard particle characterization techniques, such as laser diffraction and SPOS.


  1. ISO. ISO 13320:2020-01 Particle Size Analysis–Laser Diffraction Methods (2020).
  2. ISO. ISO 21501-3:2019-11 Determination of Particle Size Distribution–Single Particle Light Interaction Methods–Part 3: Light Extinction Liquid-Borne
    Particle Counter (2019).
  3. ISO. ISO 13322-1:2014-05-15. Particle Size Analysis–Image Analysis Methods–Part 1: Static Image Analysis Methods (2014).
  4. Malvern Panalytical. (2018). Morphologi 4 User Manual: MAN0595-01-EN-00 United Kingdom: Malvern Panalytical.
  5. Abramowitz, M; Davidson, M. Optical Microscopy. (accessed Sept. 27, 2023).
  6. Malvern Panalytical, Morphologically-Directed Raman Spectroscopy (MDRS). (accessed Sept. 27, 2023).

About the author

Jorie Kassel is Laboratory Division Manager at Particle Technology Labs.