Introduction

Nanoparticles are defined as material between 1 and 100 nm in size with a mass of 105 atoms (Issa et al. 2013). Nanoparticles are known to have a wide range of diagnostic and therapeutic applications. As examples, vascular and tumor imaging, drug delivery and iron supplement may be mentioned (Arami et al. 2015). The strong interest in nanoparticles results from the unique physical properties such as the area of their surface and the surface tension (Paranjpe and Müller-Goymann 2014. The therapeutic effect of nanoparticles, however, is strongly dependent on their size (de Jong et al. 2008; Dwivedi et al. 2014). The importance of nanoparticles including quality control was mentioned in the recent literature (Kite et al. 2019; Patil and Bhosale 2019; Patil et al. 2018, 2020). This frequent control of the synthesized nanoparticles is therefore a condition to maximize the therapeutic effect (Nischwitz and Goenaga-Infante 2012; Vippola et al. 2016). The gold standard for quality control is the analysis of images created by transmission electron microscopy, scanning electron microscopy, scanning transmission electron microscopy, small-angle X-ray scattering, wide-angle X-ray scattering, atomic force microscopy, or dynamic light scattering (Vippola et al. 2016). The nature of the image analysis may be manual or automated. A manual analysis is tedious, time-consuming and, most important, not reproducible. An automated analysis on the other hand is reproducible, fast and manages bigger sample sizes (Buzin et al. 2015, 2017; Macedo et al. 2017). Last, not least, the automated analysis does not require a highly educated specialist, resulting in a more inexpensive analysis. A recent study stated the inability of common image analysis software to identify agglomerated nanoparticles. To this end, the authors developed a software capable of identifying single nanoparticles also when agglomerated. The authors compared their results with ImageJ and a manual analysis (Vippola et al. 2016). The results of the three methods seemed comparable. Gold nanoparticles (AuNPs) have been researched for applications in medical images, according to recent research; they are considered image contrast agents that benefit clinical practice (Han et al. 2019).

The authors of the present study hypothesized that agglomerated events might be excluded from an automated analysis, given, that a sufficient quantity of nanoparticles per sample is analyzed.

The aim of the present study is to analyze shape and size of nanoparticles with a free software for image analysis and subsequently compare the results with those of a manual analysis.

Sample preparation

In Synthesis and characterization of gold nanoparticles.

Gold nanoparticles (AuNPs) were synthesized based on (Turkevich et al. 1951) method with modifications. Briefly, 15 mL of chloroauric acid (HAuCl4, Merck Co. Ltd., Brazil) at 2.5 × 10−4 M was maintained under agitation (200 rpm). The gold precursor was reacted with 1 mL of trisodium citrate dihydrate at 1% m/v (Na3C6H5O7·2H2O, Merck Co. Ltd., Brazil) for 15 min at 100ºC. The solution was cooled in ice bath for 30 min. AuNPs were collected, and their optical properties were evaluated by UV–vis spectroscopy (FEMTO 800 XI). Size and morphology were examined by transmission electron microscope (JEOL, JEM-1400) operated at 120 kVwith Lab6 filament. AuNPs were characterized using ImageJ and Cellprofiler.

Materials

ImageJ

The monodispersity of the AuNPs was determined based on the aspect ratio manually (major axis length and minor axis length) using the ImageJ software as shown in Fig. 1.

Fig. 1
figure 1

TEM image used to count

Full size image

CellProfiler

The automated analysis was conducted using the free software CellProfiler (version 2.1.1) and analyzed 23 images (N = 423 nanoparticles), and each image was processed individually.

The identified objects that did not reach a certain minimum in terms of a user-defined, image-dependent form factor were not included in the subsequent analysis, as described in Table 1 (5-Filter Objects):

Table 1 Algorithm used for the image analysis with CellProfiler
Full size table

The original image, the initial identification and the result of the filtering are depicted in Fig. 2.

Fig. 2
figure 2

Original image, referring to module 1 of Table 1 (a); initial identification of the objects of interest, referring to module 3 of Table 1 (b); Remaining objects after filtering the objects, referring to module 5 of Table 1 (c)

Full size image

Results and discussion

A quality control, i.e., a morphological analysis of nanoparticles was conducted manually and automated. A mean of the values is described in Table 2. Subsequently, the results were compared. The results of the comparison are depicted (Fig. 3).

Table 2 Mean and standard deviation of manual and automated count
Full size table
Fig. 3
figure 3

A Bland–Altman graph for the comparison of manual and automated analysis. The compared parameters were minor axis length divided by the major axis length. The x-axis represents the media of the two analyses; the y-axis represents the difference between the two analyses. The middle line of the y-axis represents the mean of the differences ( − 0.01). The upper and lower lines are mean ± 1.96 *standard deviation of the differences

Full size image

The Bland–Altman assay shows the mean of the differences is close to zero. There is no systematic bias, i.e., the differences are mean, above and below the mean. The higher the mean, the smaller the differences between the two methods. Almost all the differences are within mean + 1.96 *standard deviation, i.e., the differences are normally distributed (Fig. 4).

Fig. 4
figure 4

The boxplot represents the variability of the manual and automatic data in the study. The central line represents the median, and the top and bottom lines are the data range (maximum–minimum). The spaces between the different parts of the box indicate the degree of dispersion

Full size image

The median of the manual method was 0.94, and in the automatic method, it was 0.97, as shown in the graph by the central line. The automatic method showed a more symmetrical distribution than the manual method. There is also a low variability and low standard deviation compared to the manual method.

The effect of nanoparticles is strongly dependent on their size (De Jong et al. 2008; Dwivedi et al. 2014). A frequent quality control of the synthesized nanoparticles is therefore a very important measure to maximize the effect of the objects (Nischwitz and Goenaga-Infante 2012; Vippola et al. 2016).

For the present study, a quality control, i.e., a morphological analysis of nanoparticles was conducted manually and automated. Subsequently, the results were compared. The data of the present study indicate that the results of the manual and automated analysis are comparable. In a recent study, a software was developed, capable of conducting a quality control of nanoparticles including the identification and analysis of single objects even when conglomerated (Vippola et al. 2016. The authors had compared the results obtained with their new software with those of the software imageJ and a manual analysis. The results of the three different methods seemed comparable; however, no statistical analysis of the comparability of the results was conducted. As the authors of the present study hypothesized that conglomerated objects can be excluded, the results of the present study indicate that the results of a manual analysis, including conglomerated objects, are comparable to those of an automated analysis that excluded conglomerated objects.

Conclusions

The results of the present study indicate that an automated analysis excluding conglomerated objects may deliver the same results as automated analysis including, given that the first method analyses a representative count of objects. Another conclusion of the present study is that an automated analysis is more accurate and faster as compared to a manual analysis. It may be concluded that there is no necessity to spend labor, money and time to develop a software.

Further comparisons may be necessary to verify the truth of the author’s hypothesis of the present study.