Deep learning for automated size and shape analysis of nanoparticles in scanning electron microscopy

Bals, Jonas; Epple, Matthias

doi:10.1039/d2ra07812k

Cited by 18 publications

(11 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For this, a previously described routine was used (see ref. [35] for details). The performance of a network was defined by the Intersection over Union (IoU), defined as

IoU = \frac{100 \cdot TP}{TP + FP + FN} false[% false]

with TP: true positive, FP: false positive, TN: true negative, and FN: false negative, ranging from 0 to 100.…”

Section: Resultsmentioning

confidence: 99%

“…For this, a previously described routine was used (see ref. [35] for details). The performance of a network was defined by the Intersection over Union (IoU), defined as…”

Section: Segmentation and Analysis Of Sem Images By The Cnns Trained ...mentioning

confidence: 99%

“…[ 4,9,10 ] We demonstrate here that the performance of UNet++ trained on artificial data is comparable to the same model trained on real SEM images. [ 35 ] We also demonstrate the limitations of this approach.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Artificial Scanning Electron Microscopy Images Created by Generative Adversarial Networks from Simulated Particle Assemblies

Bals

Epple

2023

Advanced Intelligent Systems

Self Cite

View full text Add to dashboard Cite

Particle assemblies created by software package Blender are converted into artificial scanning electron micrographs (SEM) with a generative adversarial network (GAN). The introduction of height maps (i.e., surface topography or relief structure) considerably enhances the quality of the artificial SEM images by providing 3D information on the input data. These artificial images serve as input data to train a convolutional neural network (CNN) to identify and classify nanoparticles. Although the performance of the CNN trained with artificial SEM images is slightly inferior to the same CNN trained with real SEM images, this offers a pathway to create training data for segmentation and classification networks for SEM image analysis by deep learning algorithms.

show abstract

“…For this, a previously described routine was used (see ref. [35] for details). The performance of a network was defined by the Intersection over Union (IoU), defined as

IoU = \frac{100 \cdot TP}{TP + FP + FN} false[% false]

with TP: true positive, FP: false positive, TN: true negative, and FN: false negative, ranging from 0 to 100.…”

Section: Resultsmentioning

confidence: 99%

“…For this, a previously described routine was used (see ref. [35] for details). The performance of a network was defined by the Intersection over Union (IoU), defined as…”

Section: Segmentation and Analysis Of Sem Images By The Cnns Trained ...mentioning

confidence: 99%

See 1 more Smart Citation

Artificial Scanning Electron Microscopy Images Created by Generative Adversarial Networks from Simulated Particle Assemblies

Bals

Epple

2023

Advanced Intelligent Systems

Self Cite

View full text Add to dashboard Cite

show abstract

“…Additionally, reference data for selected characterization methods will be made accessible to ensure the accuracy, comparability, standardization, and compatibility of characterization methods for complex nanomaterials, which can also be utilized for the feeding of deep learning approaches. 33 ■ ASSOCIATED CONTENT * sı Supporting Information…”

Section: ■ Conclusion and Outlookmentioning

confidence: 99%

Iron Oxide Nanocubes as a New Certified Reference Material for Nanoparticle Size Measurements

et al. 2023

View full text Add to dashboard Cite

The rational design and increasing industrial use of nanomaterials require a reliable characterization of their physicochemical key properties like size, size distribution, shape, and surface chemistry. This calls for nanoscale reference materials (nanoRMs) for the validation and standardization of commonly used characterization methods closely matching real-world nonspherical nano-objects. This encouraged us to develop a nonspherical nanoRM of very small size consisting of 8 nm iron oxide nanocubes (BAM-N012) to complement spherical gold, silica, and polymer nanoRMs. In the following, the development and production of this nanoRM are highlighted including the characterization by transmission electron microscopy (TEM) and small-angle X-ray scattering (SAXS) as complementary methods for size and shape parameters, homogeneity and stability studies, and calculation of a complete uncertainty budget of the size features. The determination of the nanocubes’ edge length by TEM and SAXS allows a method comparison. In addition, SAXS measurements can also provide the mean particle number density and the mass concentration. The certified size parameters, area equivalent circular diameter and square edge length, determined by TEM with a relative expanded uncertainty below 9%, are metrologically traceable to a natural constant for length, the very precisely known (111) lattice spacing of silicon. Cubic BAM-N012 qualifies as a certified nanoRM for estimating the precision and trueness, validation, and quality assurance of particle size and shape measurements with electron microscopy and SAXS as well as other sizing methods suitable for nanomaterials. The production of this new iron oxide nanocube RM presents an important achievement for the nanomaterial community, nanomaterial manufacturers, and regulators.

show abstract

“…The rise of articial intelligence/machine learning/deep learning has considerably enhanced our ability to train computers to recognize and autonomously analyse particles. Machine learning techniques have already been applied to electron microscopic images where they usually outperform classical image analysis approaches, especially when noisy images or overlapping particles are involved 15,[17][18][19][20][21][22][23][24][25] (see ref. [26][27][28] for recent reviews).…”

Section: Introductionmentioning

confidence: 99%

Automated analysis of transmission electron micrographs of metallic nanoparticles by machine learning

et al. 2023

Self Cite

View full text Add to dashboard Cite

show abstract

Deep learning for automated size and shape analysis of nanoparticles in scanning electron microscopy

Abstract: Particles depicted in scanning electron micrographs are automatically identified and classified according to size and shape with a deep-learning algorithm. The procedure works for both SE images and STEM images.

Cited by 18 publications

References 35 publications

Artificial Scanning Electron Microscopy Images Created by Generative Adversarial Networks from Simulated Particle Assemblies

Artificial Scanning Electron Microscopy Images Created by Generative Adversarial Networks from Simulated Particle Assemblies

Iron Oxide Nanocubes as a New Certified Reference Material for Nanoparticle Size Measurements

Automated analysis of transmission electron micrographs of metallic nanoparticles by machine learning

Contact Info

Product

Resources

About