A deep learning-based framework for automatic analysis of the nanoparticle morphology in SEM/TEM images

Area of Science:

• Materials Science
• Nanotechnology
• Computer Vision

Background:

• Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) are essential for nanomaterial characterization.
• Automated analysis of nanomaterial morphology in SEM/TEM images is critical for accelerating research.
• Current methods face challenges in high-throughput, automated online statistical analysis of complex images.

Purpose of the Study:

• To develop a universal deep learning framework for fast and accurate online statistical analysis of nanoparticle morphology in complex SEM/TEM images.
• To enable high-throughput characterization of nanomaterials.

Main Methods:

• A three-stage framework: nanoparticle segmentation using a lightweight deep learning network (NSNet), nanoparticle shape extraction, and statistical analysis.
• NSNet was developed for efficient and accurate segmentation of nanoparticles in challenging imaging conditions.

Main Results:

• The proposed framework achieved 86.2% accuracy and processed 11 SEM/TEM images per second on an embedded processor.
• NSNet demonstrated superior performance in segmenting small, dense nanoparticles with high background interference compared to other models.
• Statistical analysis results (equivalent diameter and Blaschke shape coefficient) closely matched manual analysis.

Conclusions:

• The developed deep learning framework offers a promising solution for automatic and intelligent analysis of nanomaterial morphology.
• This technology has the potential to significantly advance high-throughput nanomaterials science research.