Automatic segmentation of inorganic nanoparticles in BF TEM micrographs

Area of Science:

• Materials Science
• Nanotechnology
• Microscopy

Background:

• Transmission Electron Microscopy (TEM) is crucial for nanoparticle characterization.
• Accurate segmentation of nanoparticles in TEM images is challenging, leading to poor statistical analysis.
• Existing methods struggle with overlapping particles and high error rates.

Purpose of the Study:

• To develop an automated method for accurate nanoparticle segmentation in TEM micrographs.
• To overcome limitations of existing segmentation techniques, particularly in handling particle overlap.
• To improve the statistical reliability of nanoparticle analysis from TEM data.

Main Methods:

• Development of an automatic particle picking device utilizing variance hybridized mean local thresholding.
• Implementation of the algorithm in Matlab for processing TEM images.
• Validation using a dataset of 150 bright field TEM micrographs with ~2,000 nanoparticles.

Main Results:

• The novel method significantly reduces false positives and false negatives compared to global thresholding, local thresholding, and manual segmentation.
• The device effectively increases the number of individual particles identified in regions of overlap.
• Improved accuracy in nanoparticle size and shape analysis.

Conclusions:

• The proposed automatic particle picking device offers a robust solution for nanoparticle segmentation in TEM.
• This method enhances the quality and reliability of quantitative nanoparticle analysis.
• The technique has the potential to advance nanoparticle research by enabling better statistical insights.