New standards in visible particle detection in

MICROCRYSTALLINE
CELLULOSE (MCC)
Gerald A. Hebbink, Alberto Berardi
DFE Pharma, Goch, Germany
Abstract
Visible particles in pharmaceutical powder excipients are typically counted manually. Given
that the ability of the human eye to detect particles is highly subjective, manual counting is
inherently prone to operator-dependent bias and potential under- or overestimation of the
presence of visible particles. In this study, we present an automated system that eliminates
the disadvantages of manual counting. Once the color, contrast, and size thresholds for the
detection of non-white particles are properly set, particles can be counted without bias.
In addition to simply counting particles, this method enables the classification of detected
particles based on color, size, and shape. This method can therefore be easily used to discern
technically unavoidable particles (TUPs) of excipients from atypical visible particles (AVPs).
Moreover, this method can facilitate the detection of deviations in the production process
and guide the optimization of excipient production processes. In this study, an automat-
ed detection system was used to compare the levels of detectable and visible particles of
various microcrystalline cellulose (MCC) grades produced by three global manufacturers. It
was found that the number and type of detectable particles are different across different
brands of MCC; for example, one brand contained more particles than the other two, the
majority of which were 50-100 μm black particles. The automated method presented here
can provide an unprecedented level of reliability and detail in the counting of visible particles
in powders.

1
1. Introduction
An important aspect of excipient composition is the presence or absence of visible particles
[1,2]. There are many definitions for visible particles [1]. Herein, we refer to particles that are
potentially visible to the human eye. Therefore, they are large (typically larger than 50 μm)
and have a color contrast with the bulk of the material, for example, a black particle in a
white powder. In many cases, visible particles are technically unavoidable and originate from
processing and raw materials. Many excipients have a natural origin, such as wood pulp [3]
as the raw material for microcrystalline cellulose and bovine milk for lactose [4], which can
potentially be a source of visible particles. Although modern production processes are de-
signed to minimize discoloration, the formation of particles with unusual colors cannot be
completely avoided. For example, excessive local heat during drying may result in discolor-
ation. The excipient industry has been proactively working to reduce the formation of these
particles, and current excipient products can be regarded as clean with a minimal number of
discolored particles.

The International Pharmaceutical Excipients Council (IPEC) is driving an initiative to provide

a technically unavoidable particle profile1 (TUPP) for excipients [2]. The TUPP describes the
potential of unavoidable particles that can be expected in an excipient. An important aspect
in the definition of TUPP is the characterization of particle morphology and dimensions, for
example, through digital photographs. In addition to characterization, the IPEC suggests
that the quantification of visible particles can be used to verify consistency with the manu-
facturer’s historical levels. Finally, in the early phase of excipient development, the quanti-
fication of visible particles can be used to detect potential issues and optimize processes to
reduce their number.

Despite the importance of visible particle quantification, as outlined by the TUPP guidelines,
the actual counting of these particles is still based on subjective methods such as manu-
al counting by a quality control analyst. Even if proper training is implemented and tools,
such as magnifying glasses or lamps, are used, manual counting remains prone to bias. The
different visual abilities of the operators, and the typical low frequency of visible particles,
hinder the reliability of manual counting. To understand how infrequently the occurrence of
visible particles in powdered excipients is, a theoretical calculation is provided hereafter. In
the case of microcrystalline cellulose (MCC), the typical frequency for visible particles that
are manually observed is of the order of 1-10 particles per 100 g of MCC. A manual count
in the order of 10-20 particles in 100 g of MCC is regarded as an acceptable proportion.
Considering that the poured bulk density of MCC is 0.3 g/ml, 100 g of the product occupies
approximately 333 ml. An assumed visible cubic particle with dimensions of approximately
100 μm has a 10-6 ml volume. Therefore, 10 particles occupy one fraction of 10-7-10-8 of the
total volume of 100 g of powder. If the surface of a tablet accounts for approximately 10%
of the volume of the dosage form, the probability that a colored particle will visibly end up
at the surface of a pharmaceutical dosage form is approximately one per billion. The de-
velopment of automated and unbiased methods for the detection of visible particles would
enable manual counting of such small numbers of particles.

1
https://www.gmp-compliance.org/files/guidemgr/tuppguide2015-1539615522.pdf
2
This paper presents the development of an innovative automated system for visible particle
detection. Compared to previous manual approaches, this method has the following advan-
tages: it provides standardized and objective morphological characterization and quantifi-
cation of visible particles. Therefore, it can facilitate the development of a TUPP and provide
information on atypical visible particles that are not present in the TUPP.

3
2. Manual versus automated counting of visible particles
The need to develop an automated system for visible particle detection is driven by the poor
reproducibility of particle detection by the human eye. For a particle to be distinguished
from the bulk, it must have a certain difference in color and contrast with the surrounding
material. However, the concept of a “visible” particle is qualitative and arbitrary, as detec-
tion limits, i.e., thresholds, in both the contrast and color between the visible particle and the
background are not fixed in the human eye. For example, many individuals are color-blind
to some extent and cannot distinguish certain colors [5]; therefore, they may miss the count
of specific particles. This influences the reproducibility of counting. This subjectivity leads
to different observations and conclusions regarding particle counts in the same powder. A
system based on automated image analysis does not have these limitations. All color and
contrast parameters can be standardized, and the thresholds of the detection system can
be fixed. In addition to color, particle size is an issue in manual observations because small
particles cannot be detected by the human eye. Each human eye is different, and there is no
fixed size threshold. Tools such as lamps or magnifying glasses enhance the detection abil-
ity for both color and particle size. The values of particle size thresholds for visible particles
have been defined in several ways. The United States Pharmacopeia (USP) does not define
a sharp boundary between visible and subvisible and identifies a visible gray zone between
50-150 μm particle size (USP<1788>, Figure 1) [6]. In situations where the color contrast is
remarkably high, for example, a fully black particle in a white powder, particles smaller than
50 μm are expected to be identified by most individuals. In an automated system, detect-
able particle size can be standardized and calibrated using fixed lenses and a fixed digital
camera system.

Size domains considered in USP <1>, <788>, <789>

Sub-visible Visible

Sub-micrometer
Visible
1 μm gray zone
10 μm
25 μm
50 μm 150 μm

Figure 1. USP<1788> definition of visible particles [6]

In summary, the accuracy of manual particle counting is hindered by subjective and highly
variable abilities of the human eye. An automated system based on image analysis enables
the fixing of size and color thresholds and can therefore improve the precision and reliabil-
ity of visible particle counts well beyond human capability.

4
2.1. Manual particle counting
There are several ways to manually count particles in excipients. A general way to count them
is by spreading a layer of powder in a 30x30 cm square. Magnifying glass and lighting are used
to facilitate the manual counting of all contrasting particles. To increase the chance of detecting
particles, sieving can be used to concentrate large visible particles by removing a large part of the
fine fraction, which is thought to not contain visible particles. Counting is based on the subjective
evaluation of the analyst who defines if a particle is different from the bulk or not. Making this
decision can be challenging, particularly for faint and small particles.

2.2. Automated particle counting; definition of particle color

In automated particle counting, the particle color and size must be defined upfront, starting from
the color. The most convenient method is to use a three-dimensional color space, as shown in
Figure 2. The first dimension is the color hue, which is a circular value, where an angle of 0° (360°)
corresponds to red. The second dimension is color saturation, which describes the strength of
the color. High saturation is a strong color and low saturation is a weak, faint, light-greyish color.
The third dimension is the color intensity, where an exceedingly high intensity indicates a very
bright color, and a low intensity indicates no color or darkness. For example, the combination of
low intensity and low saturation is black. Each pixel of the digital camera provides values for all
three dimensions. Software can then analyze the matrix of pixels on their location in a three-di-
mensional color space and determine if certain pixels are different from the background, that is,
a detectable particle in a bed of standardly colored particles. Please note the difference in defini-
tion: a detectable particle found by pre-programmed software is different from a visible particle
found by an operator (prone to bias).
Hue (0-360°)
Saturation
Saturation (0-100%)

Intensity (0-100%)

Hue
Desatur
ation
Value

Figure 2. Three-dimensional color space projected in a two-dimensional color space and one-dimensional intensity (left im-
age) or in a circular manner (right image), with hue in the form of a cylinder, saturation as the radius of the cylinder, and inten-
sity (here designated as ‘value’) as the axis. Both representations provide the same three-dimensional color space.

It is necessary to define thresholds for color dimensions that are relevant for detection, that is,
the color of a standard particle. First, a background image of the material must be created to
determine its specific color. The whitest materials have low saturation levels and high intensities;
however, these strongly depend on the material. A detectable particle is a combination of pix-
els with distinct locations in the color space compared to the standard color dimensions of the
background. Detection limits should not be set too close or too far from the standard color di-
mensions, thus risking the over- or underestimation of the number of detectable particles. Proper
method development is required to set boundaries for the color dimensions that determine what
an atypical color is, considering that natural variability in powder colors also exists.
5
2.3. Automated counting; definition of particle size
The camera images an area by projecting it onto a chip with a certain number of pixels. The magnifi-
cation is determined by hardware settings, such as the distance, lens system, and detector definition.
In theory, a particle projected onto a single pixel can be distinguished; however, in practice, a minimum
number of approximately 5×5 pixels should be used to distinguish a particle from its surroundings (Fig-
ure 3). This implies that the camera system should be capable of projecting the smallest detectable
particles onto at least 5×5=25 pixels. The software can identify this group of pixels as particles with
a certain (average) color, size, and shape. The more pixels per particle, the better the color, size, and
shape that can be determined.

The USP defined a gray area in the diameters of visible particles ranging from 50 to 150 μm. There-
fore, a representative automated system should be capable of detecting particles at least 50 μm in
diameter. The magnification of the system should be such that 1 pixel represents a maximum of 10 μm.

Figure 3. This image illustrates how a particle is projected onto a 5×5-pixel grid and how the software translates
it into color dimensions and size. Left: a hypothetical particle projected onto a grid; right: how the grid ‘sees’
the particle, i.e., as an assembly of pixels with varying intensity, depending on the overlap of particles with each
pixel. The greater the number of pixels per particle, the better the identification and determination of the color
dimensions. In this example, the particle fills 15 pixels, corresponding to a diameter of ~4 pixels, and an average
color intensity of 5x100%+1x75%+4x50%+5x25%=60%.

2.4. Automated counting systems

Several systems exist that enable automated detection of particles. The easiest method is to use a
microscope to image a specific surface and detect the particles in that image using a software such
as ImageJ [7]. However, the surface that can be imaged in this manner is limited to a few square centi-
meters, which is sufficient only when a high concentration of detectable particles is present. In typical
pharmaceutical excipient powders, the number of visible particles is extremely low (i.e., few particles
per m2). To properly count detectable particles and provide statistically relevant and reproducible data,
several square meters of the powder must be analyzed. OCS-PT2C (OCS Optical Control Systems
GmbH, Witten, Germany) is a commercially available powder analyzer system, which has a vibratory
transporter that carries powder from a hopper to a collection point along a camera system. During
transport, the powder is continuously photographed, and the images are analyzed using a software.
By “teaching” the system what a particle is, based on color and size as discussed above, the particles
can be counted and classified based on certain parameters such as color, size, and shape.

6
3. Microcrystalline Cellulose (MCC)
MCC is prepared from cellulose in the form of purified wood pulp. During the production process, long
cellulose fibers are hydrolyzed by acid under mechanical shear, yielding a product composed of much
shorter particles with a reduced level of polymerization. The final product is obtained by spray drying.
Colored particles in MCC are either already present in the natural raw material or are formed during
the manufacturing process in the hydrolysis reaction and/or the drying stage. Despite the optimized
production processes used to yield clean products, traces of colored particles cannot be excluded or
avoided completely. Therefore, the visible particles in MCC, according to the IPEC definition, are tech-
nically unavoidable particles (TUPs) generated from the raw material or process. Excessive heating of
carbohydrates results in particles characterized by a yellow/brown to dark color with medium to low
saturation and intensity. Visible particles that are not technically unavoidable according to the IPEC
definition are designated as ‘atypical visible particles,’ (AVPs) such as unusually colored (green, blue) or
exceptionally large particles.

Therefore, it is important to have reliable testing techniques that can provide information on the typ-
ical number of detectable particles and unexpected deviations due to raw material or process incon-
sistencies.

4. Results and discussion

The MCC powder was analyzed using an OCS-PT2C system. An example of the results for the de-
tected particles in 50 g of Pharmacel® MCC 101 is shown in Table 1. The lower particle size limit was set
at 50 μm, which was in line with the lower limits of potentially visible particles.
Table 1. Detectable particles were found in 50 g of Pharmacel® MCC 101. An empty cell indicates that we did not
detect any particles in the class. The detectable particle count is the sum of all particles, and the visible particle
count is obtained by subtracting the reported particles in the gray fields from the total sum.

Size class Black Brown Yellow Other Sum

(μm)
50-100 1 1 2
100-150 2 2
150-200 1 1 2
200-250 3 3
250-300 1 1
300-350 0
350-400 1 1 2
400-450 0
450-500 1 1
>500 0
Sum 5 6 2 0 13

A total of 13 detectable particles were counted. This was counted in 50 g of powder spread over an
area of approximately 1 × 106 mm2 (assumptions: bulk density =0.3 g/ml, layer thickness = 0.2 mm) us-
ing a vibratory feeder. All particles were classified based on their size and color dimensions. The system
also provided images of these particles (Figure 4).
7
Figure 4. Images of 13 particles were captured in Pharmacel® MCC 101.

These detected particles can be regarded as TUPs, as they are typically yellow, brown, and
black particles expected in a process in which the raw material is natural, and drying is re-
quired. No atypical particles, based on color or excessive size, were observed.

This standardized count is likely to overestimate the number of particles compared with the
number that would be visible to the human eye. First, two of the 13 detected particles were <
100 μm in size and could not be included in the visible particle count. In addition, small yellow
and brown particles may not be detected by the human eye because of their limited color
contrast with the white bulk powder background. We can speculate that of the 13 particles
detected by the automated method, 4–11 particles may have been detected by an operator
based on their individual eye sensitivity. The use of the automated method for particle de-
tection enabled a fully objective particle count and eliminated operator bias.

Given that the count of detectable particles might be an overestimation compared to visi-
ble particles, some assumptions can be made to derive an “average” visible particles count.
It can be expected that the detected particles with the highest contrast (i.e., black parti-
cles) will be visible at the lowest threshold of 100 μm; medium-contrast particles, which are
brown, at a medium threshold of 150 μm; and lowest-contrast particles, that is, yellow, at
the highest threshold of 200 μm. These thresholds are also depicted in Table 1 in gray color.
Based on these settings, the visible particle count was 11. Two out of the 13 detected parti-
cles were removed, as they are below the set thresholds for visible particles. In the following
sections, the detectable and visible particle counts of each tested sample are presented.
Particle count comparisons within a batch, batch-to-batch, grade-to-grade, and brand-to-
brand ratio can now be performed with an unprecedented level of reliability and accuracy
using an automated detection method.

In Table 2, a selection of detected particles in all size and color classes is presented. These
images are collected from several different batches and do not represent a typical batch.
However, overall these images give an overview of the typical particles that can be expected
in a batch of MCC. The smallest and faintest detected particles are regarded as invisible,
however, this is highly subjective and it is hard to substantiate the removal of these particles
from counting.

8
Table 2. Selection of typical particles found in all size and color classes.

Size class (μm) Black Brown Yellow

50-100

100-150

150-200

200-250

250-300

300-350

350-400

400-450

450-500

>500

9
Next, we measured the number of detectable particles for Pharmacel® MCC 101, 102, and 112
from DFE Pharma and compared it with the number of detectable particles of pharma-grade
MCC from two other global manufacturers, as shown in Figure 5. The reported results are an
average of 10 samples of 50 g, so a total of 500 g per sample is analyzed and reported as an
average count in 50 g of MCC. It is worth noting that over 40 samples of Pharmacel® MCC 101,
102, and 112 were measured, whereas only three samples of MCC 101, 102, and 112 from other
manufacturers were tested.

250

200
Counts/50 g MCC

150

100

50

0
1

2
10

10

10
10

11

10

11

10

11
el

B
el

B
el

B
ac
ac

ac

m
m

ar
ar

ar

Ph
Ph

Ph

Detectable Count Visible Count

Figure 5. Detectable particles of samples of various MCC products from DFE Pharma and two other manufacturers,
indicated as A and B. All colored particles >50 μm in size were included in the count of detectable particles. Visible
particles were considered black particles >100 μm, brown particles > 150 μm, and yellow particles >200 μm.

No significant differences were noted between the grades within the same MCC manufac-
turer. However, obvious differences were observed in the visible particle counts between
manufacturers. For all grades, Pharmacel® and MCC from manufacturer A contained fewer
particles than MCC from manufacturer B. Converting the data to a visible count resulted in
a decrease in the number of counted particles. For MCC from manufacturer B, a substantial
portion of detectable, but not visible, particles can be observed. An analysis of the tables,
which report the particle counts of these batches, revealed the presence of more than 100
small black particles. This deviation may have been caused by an excessive number of black
fines in the raw material or could have been generated during the production process at
manufacturer B. Although we classified these particles as non-visible based on thresholds,
the visibility of these particles is very subjective. It could be argued that these 50-100 μm
black particles lying on a white powder bed background could be distinguished by some
people (visible gray zone in Figure 1).

The brand-to-brand comparison highlighted here demonstrates the strength of this meth-
od. It was not only possible to count particles based on two different thresholds (i.e., detect-
able, and visible), but also to qualitatively evaluate the origin of the deviation of detectable
particles observed for manufacturer B (i.e., many small black visible particles). Therefore,
the automated method presented here provides an unprecedented level of reliability, accu-
racy, and details for the investigation of visible particles in powders.

10
5. Conclusions
The manual counting of particles in powdered products, such as MCC, is intrinsically biased
and may lead to imprecise counts and unreliable findings. Automatic counting using image
analysis eliminated the subjectivity of counting. Nevertheless, two prerequisites must be
satisfied to ensure reliable visible particle count. First, given the extremely low frequency
of visible particles in pharmaceutical excipients, the analytical method should ensure that
a sufficiently large amount of powder is analyzed. Second, the particle color, contrast, and
size thresholds must be set correctly and fixed. Based on this, DFE Pharma has developed
a visible particle counting method that enables the collection of highly objective data on
visible particle detection in excipient powders.

In addition to a standardized and objective methodology, the automated system enabled

the storage and categorization of particle images into a database. This database of images
can then be used to build a technically unavoidable particle profile (TUPP) for an excipient,
which can be shared with drug product manufacturers and regulatory bodies. Finally, the
automated particle counting method can provide early warnings of unexpected increases in
visible particles during production and can be used to optimize the production process. This
study also showed that pharma-grade MCC products differed in terms of the content of vis-
ible particles. MCC from one manufacturer contained three- to eight-fold more detectable
particles than MCC from other manufacturers, as determined using the automated system.

11
6. References
[1] J. Gross-Rother, M. Blech, E. Preis, U. Bakowsky, P. Garidel, Particle detection and characterization
for biopharmaceutical applications: Current principles of established and alternative techniques, Pharma-
ceutics. 12 (2020) 1–49. https://doi.org/10.3390/pharmaceutics12111112.

[2] C.C. DeMerlis, A. Smith, D.R. Schoneker, Regulatory Information for Excipients, in: Phar-
maceutical Excipients, John Wiley & Sons, Inc., Hoboken, NJ, USA, 2016: pp. 241–267. https://doi.
org/10.1002/9781118992432.ch6.

[3] S. Ardizzone, F. Dioguardi, T. Mussini, Microcrystalline cellulose powders: structure, surface fea-
tures and water sorption capability, Cellulose. 6 (1999) 57–69. https://doi.org/10.1023/A:1009204309120.

[4] G.A. Hebbink, B.H.J. Dickhoff, Application of lactose in the pharmaceutical industry, in: Lactose,
Elsevier, 2019: pp. 175–229. https://doi.org/10.1016/B978-0-12-811720-0.00005-2.

[5] N. Gordon, Colour blindness, Public Health. 112 (1998) 81–84. https://doi.org/10.1038/
SJ.PH.1900446.

[6] J. Rech, A. Fradkin, A. Krueger, C. Kraft, D. Paskiet, Evaluation of Particle Techniques for the Char-
acterization of Subvisible Particles From Elastomeric Closure Components, Journal of Pharmaceutical Sci-
ences. 109 (2020) 1725–1735. https://doi.org/10.1016/J.XPHS.2020.01.026.

[7] T.J. Collins, ImageJ for microscopy, Biotechniques. 43 (2007) S25–S30. https://doi.
org/10.2144/000112517.

12
DFE Pharma sales offices

Head Office India Singapore

Klever Strasse 187 Plot No. 148 3 Temasek Avenue

P.O. Box 20 21 20 Prestige Featherlite Tech Park #11-01 Centennial Tower
47568 Goch 2nd Phase, EPIP Zone Singapore 039190
Germany Whitefield, Bangalore 560066 T. +65 6580 8100
T. +49 2823 9288 770 India F. +65 6580 8191
F. +49 2823 9288 7799 T. +91 80 4925 6100
[email protected] F. +91 80 4925 6150

Japan North America China

Taiyo Seimei Bldg. 14F 61 South Paramus Road Room 302

2-16-2, Konan, Minato-ku, Tokyo Suite 535 No. 68 Bohang Road
108-0075 Japan Paramus, NJ 07652 Pudong Shanghai
T. +81 3 6260 0740 USA China
F. +81 3 6260 0754 T. +1 551 497 7355 T. +86 21 6134 5821
F. +1 551 497 7358 F. +86 21 6134 5800
Brazil

Rua Fradique Coutinho, 30 - sala 61

05416-000 São Paulo SP Brazil
T. +55 11 2395 1700

dfepharma.com
To learn more about the DFE Pharma approach for the detection of visible particles in excipients, please con-
tact: Dr. Gerald Hebbink- [email protected]

DFE Pharma GmbH & Co. KG

The details given here are merely intended for information purposes and are in
no way legally binding. Consequently we accept no responsibility in the broad-
est sense of the word for damage that may result from applications based upon
this information. Furthermore, this information does not constitute
permission to infringe patent and licence rights.