EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Generalization Across Experimental Parameters in Neural Network Analysis of High-Resolution Transmission Electron Microscopy Datasets

Katherine Sytwu iD, Luis Rangel DaCosta iD, Mary C Scott

Microscopy and Microanalysis · 07.03.2024 · 30 (1) · 85-95

DOI: 10.1093/micmic/ozae001 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Microscopy and Microanalysis, online erschienen am 29.01.2024. Verfasst von Katherine Sytwu und 2 weiteren Autor:innen. Beteiligte Einrichtung: Molecular Foundry, Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA. Die Arbeit wurde bisher 18-mal zitiert (Crossref) und nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neural networks are promising tools for high-throughput and accurate transmission electron microscopy (TEM) analysis of nanomaterials, but are known to generalize poorly on data that is “out-of-distribution” from their training data. Given the limited set of image features typically seen in high-resolution TEM imaging, it is unclear which images are considered out-of-distribution from others. Here, we investigate how the choice of metadata features in the training dataset influences neural network performance, focusing on the example task of nanoparticle segmentation. We train and validate neural networks across curated, experimentally collected high-resolution TEM image datasets of nanoparticles under various imaging and material parameters, including magnification, dosage, nanoparticle diameter, and nanoparticle material. Overall, we find that our neural networks are not robust across microscope parameters, but do generalize across certain sample parameters. Additionally, data preprocessing can have unintended consequences on neural network generalization. Our results highlight the need to understand how dataset features affect deployment of data-driven algorithms.

Schlagwörter

Stichwörter der Autor:innen

generalizationmachine learningnanoparticlestransmission electron microscopy

Autor:innen und Institutionen

  • Molecular Foundry, Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA.
  • Materials Science and Engineering, University of California Berkeley, 2607 Hearst Ave, Berkeley, CA 94720, USA.

ORCID-Profile: Katherine Sytwu, Luis Rangel DaCosta

Zitieren

APA

Sytwu, K., Rangel DaCosta, L., & Scott, M. C. (2024). Generalization Across Experimental Parameters in Neural Network Analysis of High-Resolution Transmission Electron Microscopy Datasets. Microscopy and Microanalysis, 30(1), 85-95. https://doi.org/10.1093/micmic/ozae001

Vancouver

Sytwu K, Rangel DaCosta L, Scott MC. Generalization Across Experimental Parameters in Neural Network Analysis of High-Resolution Transmission Electron Microscopy Datasets. Microscopy and Microanalysis. 2024;30(1):85-95. doi:10.1093/micmic/ozae001

BibTeX
@article{Sytwu2024,
  author = {Sytwu, Katherine and Rangel DaCosta, Luis and Scott, Mary C},
  title = {Generalization Across Experimental Parameters in Neural Network Analysis of High-Resolution Transmission Electron Microscopy Datasets},
  journal = {Microscopy and Microanalysis},
  year = {2024},
  volume = {30},
  number = {1},
  pages = {85--95},
  doi = {10.1093/micmic/ozae001}
}
Kostenlos starten →