Filtern
Dokumenttyp
- Posterpräsentation (2) (entfernen)
Sprache
- Englisch (2)
Referierte Publikation
- nein (2)
Schlagworte
- Automatisation (1)
- Convolutional neural networks (1)
- Core-shell particles (1)
- Electron microscopy (1)
- Image analysis (1)
- Image segmentation (1)
- Nanoparticles (1)
- Roughness (1)
- Scanning electron microscopy (1)
Organisationseinheit der BAM
- 1 Analytische Chemie; Referenzmaterialien (2) (entfernen)
Core-shell (CS) particles have been increasingly used for a wide range of applications due to their unique properties by merging individual characteristics of the core and the shell materials. The functionality of the designed particles is strongly influenced by their surface roughness. Quantitative evaluation of the roughness of CS microparticles is, however, a challenging task for Scanning Electron Microscopy (SEM).
The SEM images contain two-dimensional (2D) information providing contour roughness data only from the projection of the particle in the horizontal plane. This study presents a practical procedure to achieve more information by tilting the sample holder, hence allowing images of different areas of a single particle to be recorded at different orientations under the same view angle. From the analysis of these images, quasi three-dimensional (3D) information is obtained.
Three types of home-made particles were investigated: i) bare polystyrene (PS) particles, ii) PS particles decorated with a first magnetic iron oxide (Fe3O4) nanoparticle shell forming CS microbeads, iii) PS/Fe3O4 particles closed with a second silica (SiO2) shell forming core-shell-shell (CSS) microbeads. A series images of a single particle were taken with stepwise tilted sample holder up to 10° by an SEM with high-resolution and surface sensitive SE-InLens® mode.
A reliable analysis tool has been developed by a script in Python to analyze SEM images automatically and to evaluate profile roughness quantitatively, for individual core-shell microparticles. Image analysis consists of segmentation of the images, identifying contour and the centre of the particle, and extracting the root mean squared roughness value (RMS-RQ) of the contour profile from the particle projection within a few seconds.
The variation in roughness from batch-to-batch was determined with the purpose to set the method as a routine quality check procedure. Flow cytometry measurements provided complementary data. Measurement uncertainties associated to various particle orientations were also estimated.
In this contribution different ways are explored with the aim to generate suitable training data for ‘non-ideal’ samples using various approaches, e.g., computer-generated images or unsupervised learning algorithms such as generative adversarial networks (GANs). We used these data to train simple CNNs to produce segmentation masks of SEM images and tested the trained networks on real SEM images of complex nanoparticle samples. The novel use of CNN for the automated analysis of the size of nanoparticles of complex shape and with a high degree of agglomeration has proved to be a promising tool for the evaluation of particle size distribution on a large number of constituent particles. Further development and validation of the preliminary model, respectively larger training and validation data sets are necessary.