Identifying Agglomerates Using Automated Image Analysis

However much trouble it may cause, agglomeration is taken as part and parcel of the chemical and pharmaceuticals industry. Although it may be beneficial in modifying the properties of a powder, such as the flowability and packability, to make powder handling and processing easier, it can be a source of difficulty as well. Agglomeration can affect the surface area and homogeneity of a blend that may then negatively impact a product's performance. [1]

To track the changes that agglomeration may be causing in a product, a robust analytical method is crucial. This will identify the points in the process where problems arise, and a strategy can be developed in order to improve the manufacturing process for an optimal product.

Agglomerates are most comprehensively identified through microscopy. This technique can give great insight into the particles, and the level of agglomeration is determined through optical analysis. It does have its downsides though, as it can take a long time to perform and the results are only as valid as the number of samples analyzed [2]. A huge step forward in optical analysis of agglomerates then is the development of automated image analysis. This keeps the benefits of manual microscopy, while adding statistical significance to the analysis and remaining free from human bias.

Image Analysis Explained

Automated image analysis systems are configured to capture and record images of 1000s of individual particles from a sample. These instruments include the Malvern Morphologi G3 and Sysmex FPIA 3000, and come with analysis software that can calculate a range of morphological properties for each particle. These parameters are then used along with the particle images to identify and quantify any agglomerates. The most commonly applied morphological parameters are the Circular Equivalent Diameter (CED), circularity and Convexity.

Figure 1 shows how the CED is a circle of similar diameter to the 2D image of the particle.

CED is the diameter of the circle with same area as the 2D image of the particle.

Figure 1. CED is the diameter of the circle with same area as the 2D image of the particle.

Often, agglomerates will have a larger CED compared to other particles in the sample, so classification of the sample based on this parameter provides one method of identifying agglomerates.

The circularity is the ratio of the perimeter of a circle with the same area as the particle divided by the perimeter of the actual particle image. Circularity can vary from 0-1. A perfectly spherical particle has a circularity of 1, whereas non-spherical particles have a circularity which is less than 1.Due to consisting of several bunched particles, agglomerates are generally less circular than primary particles, which makes classification by circularity another technique in identifying agglomerates.

The convexity is a metric to describe the surface roughness of a particle. It is calculated by dividing the convex hull perimeter by the actual particle image perimeter. The convex hull perimeter may be explained as the length of an elastic band put around a particle, as shown in figure 2. Similarly to circularity, convexity can vary from 0-1, with a smooth shape having a convexity of 1 while a spiky or irregular object has a convexity closer to 0. The surface of an agglomerate is generally quite rough, so they can be expected have lower convexity values than a singular particle.

Convex Hull Perimeter is the length of the elastic band put around the particle.

Figure 2. Convex Hull Perimeter is the length of the elastic band put around the particle.

This article will describe, with examples, how agglomerates may be identified. CED, circularity and convexity can be used on their own or in combinations as core analytical tools, and may also be used with automated image analysis to reduce the overall workload when compared to manual microscopy. The article will also discuss the use of optical analysis alongside Raman spectroscopy to detect multicomponent agglomerates.

Example 1: Comparing particle size and circularity data to detect agglomerates

Although agglomerated particles are generally larger than the primary particles in a powder, there may be larger primary particles due to a quirk in the manufacturing process. If this is the case, then the agglomerates may still be identified by using a combination of particle size (CED) and circularity data.

An example data set can be seen in Figure 3a which shows a powder that was thought to contain agglomerates. The scattergram is a plot of CED against circularity: the point where the color depth is strongest indicates there is a greater particle concertation there. This confirms that most of the particles are small and spherical, but there is a region where the particles are large and have low circularity. The most likely reason for this is that they are agglomerates of these smaller particles.

To confirm this method of classifying agglomerates, the region can be selected and images of the particles in question obtained during measurement (3b) can be examined. This agrees with the high CED with low circularity method, and proves a reasonable way of detecting agglomerates.

(a) CED Vs Circularity scatter gram (left); (b) images of particles from the selected region in the scatter gram (right).

Figure 3. (a) CED Vs Circularity scatter gram (left); (b) images of particles from the selected region in the scatter gram (right).

Example 2: Comparing particle size and convexity data to detect agglomerates

Figure 4 shows an example of a common mode of agglomeration that produces fractal-like agglomerates. These ‘strings’ of small particles are often seen within suspension or emulsion samples where the force of adhesion between the particles is very small.

Images obtained during the measurement of a sample containing small primary particles and open fractal-like agglomerates.

Figure 4. Images obtained during the measurement of a sample containing small primary particles and open fractal-like agglomerates.

For samples similar to the ones shown in figure 4, agglomerates may be classified by identifying particles with a large size and low convexity. This route is preferred as the primary particles are both small (low CED) and have a smooth surface (a common feature for emulsions or milled materials. For these materials, agglomerates may be classified as those particles with a CED of >5 µm and a convexity of less than 0.993 than a solid classification can be obtained, with the degree of agglomeration between the samples being able to be asses, which is shown in figure 5. A short glance at each of the images of the particles within in each class affirms that the agglomeration classification method has been successful.

Classification of two samples, where agglomerates are classified using both size (CED) and shape (convexity) parameters. This enables the degree of agglomeration in two samples to be compared. The images of the particles classed as either individual particles or agglomerates confirm that the classification has been successful.

Figure 5. Classification of two samples, where agglomerates are classified using both size (CED) and shape (convexity) parameters. This enables the degree of agglomeration in two samples to be compared. The images of the particles classed as either individual particles or agglomerates confirm that the classification has been successful.

Example 3: Comparing convexity and circularity data to detect agglomerates

Some samples may consist of primary particles that aren’t spherical. This may cause a problem with detecting agglomerates, as size and a single shape factor aren’t a robust enough classification. A way around this is by combining multiple size or shape parameters to allow comprehensive identification of agglomerates.

Figure 6 shows a basic example of this method. Although the sample is primarily made up of smooth spherical particles that have are small in size (low CED, high circularity and high convexity), there are some slightly larger particles that are misshapen through improper processing (higher CED, lower circularity and high convexity). On top of this, there are also fractal-like agglomerates (higher CED, lower circularity and lower convexity).

Example image obtained for a sample containing (a) spherical primary particles (left), (b) misshapen primary particles (middle) and (c) agglomerates (right).

Figure 6. Example image obtained for a sample containing (a) spherical primary particles (left), (b) misshapen primary particles (middle) and (c) agglomerates (right).

Due to the multiple types of particles present in the sample, each had to be detected and classified using a combination of the circularity and convexity values for each particle. Table 1 shows the final classification scheme that was applied to the particles. Following the procedure established by the two previous examples, the success of the classification scheme was confirmed by checking the particle images.

Table 1. Circularity and Convexity values selected for identification of agglomerates.

Particle Type Classification Example particle
Primary Particle Circularity > 0.96
Misshapen Primary Particle Circularity ≥ 0.896 < 0.96
Agglomerate Circularity < 0.896
Convexity < 0.873

 

Example 4: Sorting particles based on size to aid manual identification of agglomerates

Some powders may consist of primary particles that don’t have a uniform shape or size. This makes it a challenge to tell a primary particle and agglomerate apart based on size and shape alone. In this case, by sorting the particles by their particle size will aid in manually identifying and classifying agglomerates. This method relies on the human eye being a more effective tool than mathematical analysis for these varying dimensions.

Figure 7 shows a series of particle images arranged according to the CED of the particle, with those with the largest CED on the top and smallest on the bottom. The images were captured using the Morphologi G3 automated imaging system. Ordering the particles in such a way allows the user to quickly check over the particles and identify those which may be agglomerates, which can then be tagged and classified manually.

In figure 7, the particles with the blue background have been tagged as agglomerates by the user after in depth observation of the whole series of particles. The Morphologi G3 is capable of returning to the particle of interest on the sample slide, allowing the user to view the particle using different image magnification or light levels making agglomerate identification easier.

Manual identification and selection of agglomerates.

Figure 7. Manual identification and selection of agglomerates.

After the user is satisfied that all the agglomerates have been tagged, the size and shape parameters can be analyzed to see if there are any similarities. These similar parameters can then be used in the future to help automate the classification process.

Example 5: Combining imaging and spectroscopy for the detection of multicomponent agglomerates

Figure 8 is the Raman spectrum of a particle (in black) obtained from the Morphologi G3-ID MDRS system. This has been overlaid with the library spectra for two pure components (Active Pharmaceutical Ingredient (API) 1 and API 2) the sample is known to contain. This particle was chosen for Raman analysis due to its size and shape parameters identifying it as potential agglomerate. By comparing the particle spectrum with the spectra for API1 and API2 shows that it contains features relating to both pure components, confirming it as Multi-Component Agglomerate (MCA).

Raman spectrum (black) obtained for a particle using the Morphologi G3-ID MDRS system. This spectrum contains features relating to two of the components of the sample (API 1 and API 2), confirming that it is an MCA.

Figure 8. Raman spectrum (black) obtained for a particle using the Morphologi G3-ID MDRS system. This spectrum contains features relating to two of the components of the sample (API 1 and API 2), confirming that it is an MCA.

These automated techniques, including, MDRS, allow a statistically-significant amount of particles to be imaged and measured for particle size, shape and spectral information. The particle spectra can then be compared to those of pure components and the whole sample can be classified, similar to the way in figure 9. A measure of degree of agglomeration can also be obtained.

Classification of a sample containing 3 pure components (API 1, API 2 and a Lactose excipient). The application of MDRS enables the identification and classification of MCAs within the sample.

Figure 9. Classification of a sample containing 3 pure components (API 1, API 2 and a Lactose excipient). The application of MDRS enables the identification and classification of MCAs within the sample.

Figure 9 also shows how difficult it is to identify agglomerates for cases like this. Both the pure components and MCAs have a similar particle size and shape, so it is vital that Raman spectroscopy is used alongside size and shape parameters to enable successful agglomeration classification.

Conclusion

Many industries face challenges in identifying agglomerates. Although sometimes necessary, agglomeration can also affect the properties of a powder, such as flowability, surface area and dissolution rate that may negatively impact the final product. In order to prevent this, a robust agglomerate analysis method is required.

Automated image analysis systems are an excellent method for classification of parameters. By including spectroscopic measurements alongside imaging, in the case of MDRS, chemical composition of particles can be identified, aiding understanding of the type of agglomeration occurring with the sample (cohesion or adhesion).

Together, these techniques enable quick analysis of statistically relevant particle population for effective identification and classification of agglomerates. These techniques can therefore be used to establish robust analytical methods for the analysis of agglomerates, so users may be sure that of consistent product quality.

References

  1. Maryam Maghsoodi, Katayoun Derakhshandeh and Zahra Yari, "On the mechanism of agglomeration in suspension", Advanced Pharmaceutical Bulletin, 2012, 2(1), 25-30
  2. Gary Nicolas, Stephan Bynard, Mark J. Bolxham, Joanne Botterill, Neil J. Dawson, Andrew Dennis, Valerie, Nigel C. North, John D. Sherwood, "A Review of the Terms Agglomerate and Aggregate with a Recommendation for Nomenclature Used in Powder and Particle Characterization", Journal of Pharmaceutical Sciences, VOL. 91, NO. 10, October 2012

This information has been sourced, reviewed and adapted from materials provided by Malvern Panalytical.

For more information on this source, please visit Malvern Panalytical.

Ask A Question

Do you have a question you'd like to ask regarding this article?

Leave your feedback
Submit