- Nano Express
- Open Access
Insight to Nanoparticle Size Analysis—Novel and Convenient Image Analysis Method Versus Conventional Techniques
© Vippola et al. 2016
- Received: 5 February 2016
- Accepted: 22 March 2016
- Published: 31 March 2016
The aim of this paper is to introduce a new image analysis program “Nanoannotator” particularly developed for analyzing individual nanoparticles in transmission electron microscopy images. This paper describes the usefulness and efficiency of the program when analyzing nanoparticles, and at the same time, we compare it to more conventional nanoparticle analysis techniques. The techniques which we are concentrating here are transmission electron microscopy (TEM) linked with different image analysis methods and X-ray diffraction techniques. The developed program appeared as a good supplement to the field of particle analysis techniques, since the traditional image analysis programs suffer from the inability to separate the individual particles from agglomerates in the TEM images. The program is more efficient, and it offers more detailed morphological information of the particles than the manual technique. However, particle shapes that are very different from spherical proved to be problematic also for the novel program. When compared to X-ray techniques, the main advantage of the small-angle X-ray scattering (SAXS) method is the average data it provides from a very large amount of particles. However, the SAXS method does not provide any data about the shape or appearance of the sample.
- Particle size analysis
- Image analysis
- Snake model
- Balloon force
Nowadays, nanoparticles are widely studied and used while their nanometer-scale size introduces such properties in them, which can differ significantly from those of the corresponding bulk material. Nanoparticles can be used in wide variety of applications where they can be considered for example as chemically inert additives like fillers in novel composite materials and high-refractive index and UV absorbing pigments in cosmetics and other consumer products or as chemically active particles in biomedical, in catalytic and in biotechnological use and like drug delivery agents for pharmaceutical industry [1–3]. Nanoparticles have unique properties which directly correlate to their size, shape, and size distribution, and therefore, to ensure the full exploitation of their properties, it is important to be able to measure these features efficiently and accurately. In recent years also huge efforts for estimating nanoparticles and related products health effects have been made [1, 3–5]. In order to provide reproducible nanoparticle characteristics for nanosafety studies and for compliance of nanoregulations, a precise determination of nanoparticle size and size distribution is crucial. Examples of methods used to characterize nanoparticles both for technical and academic use are transmission electron microscopy (TEM), scanning electron microscopy (SEM), scanning transmission electron microscopy (STEM), small-angle X-ray scattering (SAXS), wide-angle X-ray scattering (WAXS), atomic force microscopy (AFM), dynamic light scattering (DLS), differential mobility analysis (DMA), and time-of-flight secondary ion mass spectroscopy (TOF-SIMS) [2, 6–11]. Typically, different methods are seen as complementary techniques and they are recommended to be used together. Many of these methods mentioned above are different in terms of suitability to certain sized particles, easiness to use, time required, and other characteristics possible to obtain simultaneously.
In this paper, we are concentrating on transmission electron microscopy (TEM) methods into which different image analysis techniques are linked and on X-ray diffraction (SAXS and WAXS) techniques. The novelty of this paper is our new image analysis program “Nanoannotator”. This paper describes its usefulness and efficiency when analyzing nanoparticles, and at the same time, we compare it to conventional nanoparticle analysis techniques.
Description of the Developed Image Analysis Software Nanoannotator
TEM images provide a visualization of the studied nanoparticles and thus essential information not only about the size, but also on other characteristics of the particles. However, the size distribution measurement from TEM images is typically a challenging task due to the tendency of nanoparticles to accumulate together on TEM grid. Manual measurement is a widely used approach, but is limited by a large variance between subjects and limited throughput. On the other hand, the ability of image processing algorithms to separate particles that are in contact with each other is often limited, as well. Thus, the problem grows around the topic of identifying individual particles and their shape from agglomerates. Once this task has been solved, the characteristics, such as the size, orientation, or shape factor of individual particles are easier to resolve.
An image analysis framework for studying the primary nanoparticle size distribution was developed. The particles are modeled using an active shape model : the method fits an active contour to the edge map of a TEM image, and it is able to identify individual nanoparticles and to define their size and shape parameters.
The image analysis pipeline consists of three steps: center point approximation, particle segmentation, and parameter computation. Center point approximation aims to extract the center point of each individual particle in the image, thus resulting in a set of estimated coordinates of all particles present in the image. The second step finds a local homogeneous area around the center point of each particle. As a result, we will have an exact map or the area occupied by each particle. After all particles and their shapes have been separated, the final step is to compute any area- or shape-based quantities from the mutual arrangement of the pixels in a segment.
The proposed method partitions first image roughly into particle region and background region using Otsu’s thresholding. The first step begins by a rough segmentation of the image into foreground and background based on local brightness. For this purpose, we use the well-known Otsu segmentation , which finds a gray level minimizing the sum of variances of the foreground and background partitions. The result of the initial segmentation step gives a preliminary separation of the particles (dark) from the background (bright). The centroids of the particles are then located using the distance transformation, where the distance to the background region is computed for each foreground (particle) pixel. Since distance transform is known to result in high values close to particle centers and low values near borders, we can define the centroids as the locations of the local maxima in the transform.
The center points are needed in the second step where enclosed contour v(s) = (x(s), y(s)) is moved within the image, and its position is updated iteratively into the direction of lowest energy. Here, the contour v(s) needs to be initialized before iterations can be run. In the case of round particles, one approach to form the initial contours is to draw a circle around each center point and to set their radii slightly larger than the distance from the center to the background region. Other initializations may be devised according to the shape characteristics of the analyzed particles.
Parameters computed for each particle segment and their descriptions
The number of pixels in a region
The center of pixel mass
The ratio of shared pixels between the segment and its convex hull
The length of the resulting active contour around the studied particle.
Major axis length
Major axis length of an ellipse with the same second central moment as the segmented object
Minor axis length
Major axis length of an ellipse with the same second central moment as the segmented object
Major axis orientation of an ellipse with the same second central moment as the segmented object
Diameter of a circle that has the same area as the particle
Ratio of shared pixels between the segment and its convex hull
The ratio of major and minor axis (see above)
In addition to the internal and external forces, the active contour model can be extended by including a balloon force λ that acts perpendicular to the contour. Depending on initialization of λ, the force inflates (λ > 0) or deflates (λ < 0) the contour on each iteration. It improves the original model by allowing the contour to be initialized far from the desired particle contour, which is beneficial since the framework relies on automatic contour initialization .
In this study, the size distribution of three nanomaterials was studied. The studied materials were silver nanoparticles, iron oxide whiskers, and graphite nanoparticles. These materials were selected since they represent different extremes in terms of factors affecting the particle size analysis, as described in the next chapter. Silver nanoparticles from NANOGAP s.a. (Spain) have a product name of NGAP NP Ag-2103 and they are a mixture of quasi-spherical and rod-like particles with the mean particle size of 40–55 nm . Iron oxide whiskers from Nanostructured & Amorphous Materials Inc., (USA) have a product name of α Fe2O3 fiber with the fiber diameter of 40–150 nm and fiber length of 250–600 nm . Graphite nanoparticles from SkySpring Nanomaterials Inc., (USA) have a product name of graphite nanoparticles #0520BX with spherical particle morphology and the average particle size of 3–4 nm .
The size distribution of the three nanomaterials was studied by image analysis based on transmission electron microscopy (TEM) images and by small- and wide-angle X-ray scattering (SAXS/WAXS). JEOL JEM 2010 transmission electron microscope was used to study the nanomaterials. The samples were prepared by slightly crushing the nanomaterial powder between laboratory glass slides, mixing the powder with ethanol and by dropping the dispersion on the copper TEM grid with a holey carbon film. Similar imaging conditions were used for all nanoparticles (acceleration voltage 200 kV, large objective aperture).
Three different image analysis methods were compared: traditional manual image analysis, an open source image processing program ImageJ (http://imagej.net) and the MATLAB-based image analysis program Nanoannotator described in the previous chapter. The details of the image analysis practices are described together with the results. Particle size distribution by number (D n ) was determined by the image analysis methods. The results were compared to the volume-weighted particle-size distributions (D v ) defined by small-angle X-ray scattering (SAXS) method. Although in general, size distributions by number and by volume are not comparable, these two distributions result in similar average and most frequent particle sizes if the size distribution is relatively narrow . In addition, to compare the particle size and crystallite size of three nanomaterials, also wide-angle X-ray scattering (WAXS) measurements were performed.
The SAXS and WAXS measurements were done by Panalytical Empyrean Multipurpose Diffractometer with Cu Kα X-ray source (λ = 0.15418 nm) and a solid-state detector (PIXcel3D). For SAXS measurements, a focusing mirror for Cu radiation was used. Mylar foil was applied on both sides of the sample, and a double Mylar film was used as the background sample. The studied angle range 2θ was −0.1…5° with a step size of 0.01° and step duration of 3 s. The EasySAXS software (version 2.0a) was used to derive the volume-weighted particle-size distributions (D v ) by indirect Fourier-transformation. An average crystallite size for nanomaterials was determined from WAXS patterns with the aid of the HighScore Plus software (version 3.0.5) and based on the well-known Scherrer equation. Before that, phases were identified by using the database (PDF-4+ 2014) from International Centre for Diffraction Data (ICDD).
Manual Image Analysis
The particle size results obtained by different methods. The relative standard deviations values shall be compared only within one sample type, since it is dependent on the particle size
The size reported by manufacturer
Manual image analysis (D n )
Nanoannotator program (D n )
SAXS (D v )
Silver nanoparticles (diameter)
40–55 nm 
Most frequent value
Relative standard deviation
Iron oxide whiskers (length and width)
40–150 nm 
250–600 nm 
Most frequent value
Relative standard deviation
3–4 nm 
Most frequent value
Relative standard deviation
“ImageJ” Open Source Image Analysis Program
Nanoannotator Image Analysis Program
SAXS and WAXS Measurements
The SAXS measurement is a relatively straightforward method to analyze nanoparticles. The sample preparation with Mylar film takes only few minutes per sample, and the actual measurement duration is minutes (in this study 25 minutes, but also shorter runs could be used). The particle size data obtained by different methods are summarized in Table 1.
The Efficiency of the Different Methods
The efficiency of the different methods is estimated by comparing the time required in each case to achieve the results
TEM imaging and manual image analysis
Number of analyzed particles
TEM imaging and analysis with Nanoannotator
Number of analyzed particles
SAXS data acquisition and analysisa
In the case of silver nanoparticles, the accuracy of the automatic particle recognition of the Nanoannotator program was so good that the efficiency was three times better than in the manual image analysis. However, the challenging shape of the iron oxide whiskers decreased the efficiency of the Nanoannotator program, and it did not offer any advantage when compared with manual analysis. It is assumed that the manual analysis efficiency would be the same for different particle shapes. It should be noted that the described measurement durations do not represent any absolute values for these methods since the time is dependent on the operator and on the specimen, among other things.
The studied three different nanomaterials exemplified well how significant the effect of the material itself is on the practicality and efficiency of the used characterization method. The composition of the material, the particle shape and size, and the crystal size versus the particle size define the mass-thickness contrast of the TEM image. If high magnifications are used for small particles, also the inner structure of the particles can become visible, as for the graphite in our case. In this study, the focus was rather on image analysis than on optimizing the imaging conditions for each nanomaterial separately. In optimal case for particle size analysis, mass-thickness contrast solely provides best images. However, especially light and small particles, like graphite, require imaging conditions in which other contrast mechanisms are also visible. For example, increasing the objective aperture size decreases diffraction contrast in the image, but simultaneously, the effect of phase contrast is increased. Optimizing imaging conditions for a certain material can offer advantages for the particle size analysis, but in the end, the different contrast mechanisms always have some contribution to the formation of TEM images. In addition, the tendency of the material to agglomerate affects to the easiness to interpret the TEM images. The typical performance of an average observer can be estimated by the Rose criterion  which states that if the change in signal of an image exceeds the noise by a factor of five (ΔS > 5 N), it is visible to the human eye. So, if the outlines of the overlapping particles or the inner structure or topography of the particles itself cause high noise level, it can be very difficult to distinguish the particles. In these cases, the determination of the size distribution by image analysis techniques is not possible. Conventional TEM image analysis techniques require good dispersion of nanoparticles which needs a lot of resources and time for sample preparation [8, 9]. However, the ability of Nanoannotator to resolve individual particles when they are touching each other or even overlapping to some extent enables faster and straightforward sample preparation.
If the particle outlines are visible to the human eye from the image, different image analysis methods are applicable or at least the accuracy of the results can be evaluated. The general advantage of image analysis techniques is that in addition to the size distribution data, it provides other particle characteristics in the form of visual image of the particle. Also, the acquisition of elemental data by energy-dispersive X-ray spectroscopy (EDS) or crystallinity information by diffraction patterns is possible simultaneously to the imaging. The developed Nanoannotator program is a good supplement to the field of particle analysis techniques, since the traditional image analysis programs suffer from the inability to separate the individual particles from agglomerates and are thus useless. It is three times more efficient, and it offers a lot more detailed information of the particles than the manual technique. However, particle shapes very different from spherical proved to be problematic also for the Nanoannotator program and no benefit was achieved by its use. However, the program could be modified for different shaped particles and a similar efficiency as seen for the silver nanoparticles could be achieved. If the task is to measure the size distribution of agglomerates, which is a realistic case, for example, when evaluating the safety of some nanomaterials in air, also the traditional programs are practical. The only limitation of the manual image analysis is the resources it takes and the limitedness of the obtained data. It gives accurate results on particle dimensions and typically even the cases which are problematic to the image analysis methods are possible to analyze visually.
The advantage of the SAXS method is the average data it provides from a very large amount of particles. Also, the acquisition of the WAXS data from the same sample does not significantly increase the measurement time, but it provides additional data of the composition and crystal structure and size of the sample. However, the SAXS method does not provide any data about the shape or appearance of the sample. Since the shape data is required or at least helpful when analyzing the SAXS data, this method typically requires a complementary method. For small spherical particles, as for graphite, the SAXS method is optimal whereas the image analysis methods are problematic. However, for non-spherical particles, the SAXS data can be only compared to the simulated curves and the accuracy of the measurement decreases.
For non-uniform samples in which particle size or shape has a lot of variations, none of the aforementioned methods is suitable. However, the recent development of the computer technology offers new possibilities also for accurate and efficient image analysis techniques: the combination of excellent ability of the human eye to distinguish the particle outlines, touch screen, and stylus pen could offer the advantages of manual image analysis (accuracy) and image analysis programs (detailed size and shape information). In general, it can be stated that the nanoparticle analysis method has to be chosen according to the requirements of the case in question and the comparative studies between different nanomaterials is difficult.
A novel image analysis program Nanoannotator used for nanoparticle analysis from transmission electron microscopy images was introduced, and its usefulness and efficiency was evaluated with promising results. This new image analysis program proved out to be a good supplement to the field of particle analysis techniques while having several advantages over the conventional analysis methods. The developed Nanoannotator program can separate the individual particles from agglomerates, it is more efficient in time and it offers detailed information from the particles. Nevertheless, Nanoannotator program does not solve all the challenges there are within the comprehensive nanoparticle analysis. Like when the particle shapes are very different from spherical ones, there are still issues to be solved for having them analyzed efficiently. In general, it can be stated that the nanoparticle analysis method has to be chosen according to the requirements for what the analysis is used, and an approach of using more than one nanoparticle analysis method should be favored in order to gain comprehensive data on size, size distribution, and shape of nanoparticles.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
- Borm P, Robbins D, Haubold S, Kuhlbush T, Fissan H, Donaldson K et al (2006) The potential risks of nanomaterials: a review carried out for ECETOC. Part Fibre Toxicol 3:11View ArticleGoogle Scholar
- Boyd RD, Cuenat A (2011) New analysis procedure for fast and reliable size measurement of nanoparticles from atomic force microscopy images. J Nanoparticle Res 13:105–113View ArticleGoogle Scholar
- Stark WJ, Stoessel PR, Wohlleben W, Hafner A (2015) Industrial applications of nanoparticles. Chem Soc Rev 44:5793–5805View ArticleGoogle Scholar
- Maynard AD, Aitken RJ (2007) Assessing exposure to airborne nanomaterials: current abilities and future requirements. Nanotoxicology 1:26–41View ArticleGoogle Scholar
- Savolainen K, Pylkkänen L, Norppa H, Falck G, Tuomi T et al (2010) Nanotechnologies, engineered nanomaterials and occupational health and safety—a review. Saf Sci 48:957–963View ArticleGoogle Scholar
- Karlsson L, Deppert K, Malm J-O (2006) Size determination of Au aerosol nanoparticles by off-line TEM/STEM observations. J Nanoparticle Res 8:971–980View ArticleGoogle Scholar
- Temmerman P, Verleysen E, Lammertyn J, Mast J (2014) Size measurement uncertainties of near-monodisperse, near-spherical nanoparticles using transmission electron microscopy and particle-tracking analysis. J Nanoparticle Res 16:2628View ArticleGoogle Scholar
- Agbabiaka A, Wiltfong M, Park C (2013) Small angle X-ray scattering technique for the particle size distribution of nonporous nanoparticles. J Nanopar 2013:640436Google Scholar
- Pabisch S, Feichtenschlager B, Kickelbick G, Peterlik H (2012) Effect of interparticle interactions on size determination of zirconia and silica based systems—a comparison of SAXS, DLS, BET, XRD and TEM. Chem Phys Lett 521:91–97View ArticleGoogle Scholar
- Sasaki A (2005) Size distribution analysis of nanoparticles using small angle X-ray scattering technique. The Rikagu J 22:31–38Google Scholar
- Ohyama J, Esaki A, Koketsu T, Yamamoto Y, Arai S, Satsuma A (2016) Atomic-scale insight into the structural effect of a supported Au catalyst based on a size-distribution analysis using Cs-STEM and morphological image-processing. J Catal 335:24–35View ArticleGoogle Scholar
- Cootes TF, Taylor CJ, Cooper DH, Graham J (1995) Active shape models—their training and application. Comput Vis Image Underst 61:38–59View ArticleGoogle Scholar
- Otsu N (1979) A threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern 9:62–66View ArticleGoogle Scholar
- Kass M, Witkin A, Terzopoulos D (1988) Snakes: active contour models. Int J Comput Vis 1:321–331View ArticleGoogle Scholar
- L. Cohen (1991) On active contour models and balloons, CVGIP: Image Understanding 53/2: 211–218Google Scholar
- Data sheet of NGAP NP Ag-2103 particles, www.nanogap.es/#!ngap-np-ag-2103/cv7f (accessed 15.1.2016)
- Data sheet of alpha Fe2O3 fibers, www.nanoamor.com (accessed 2013)
- Data sheet of graphite nanoparticles #0520BX, www.ssnano.com/inc/sdetail/graphite_nanoparticles/303 (accessed 15.1.2016)
- T. Ferreira, W. Rasband: ImageJ User Guide IJ 1.46r. Available online http://imagej.nih.gov/ij/docs/guide/user-guide.pdf (accessed 15.1.2016)
- J. Goldstein, D. E. Newbury, P. Echlin, D. C. Joy, et.al., (2003) Scanning electron microscopy and X-ray microanalysis. 3rd ed. Springer Science, New York, USA.Google Scholar