Segmentation and Morphology Computation of a Spiky Nanoparticle Using the Hourglass Neural Network

Area of Science:

• Materials Science
• Nanotechnology
• Artificial Intelligence

Background:

• Manual measurement of nanoparticle morphology from electron microscopy images is time-consuming and prone to errors.
• Advancements in artificial intelligence (AI) and deep learning offer potential for automated image analysis.

Purpose of the Study:

• To develop and validate a deep neural network (DNN) for automated segmentation and morphological analysis of gold spiky nanoparticles (Au SNPs) in electron microscopy images.
• To assess the accuracy of DNN-based measurements against manual segmentation for nanoparticle growth analysis.

Main Methods:

• A deep neural network (DNN) was designed for segmenting Au SNPs in electron microscopy images.
• A spike-focused loss function was employed during training to enhance the detection of nanoparticle spikes.
• The DNN was evaluated for its accuracy in measuring particle growth compared to manual segmentation.

Main Results:

• The proposed DNN achieved accurate segmentation of Au SNPs, comparable to manual methods.
• Measurements of Au SNP growth using the DNN showed high agreement with those obtained from manually segmented images.
• The DNN demonstrated potential for real-time morphological analysis when tested on an embedded system.

Conclusions:

• The developed DNN provides an accurate and automated method for morphological analysis of Au SNPs.
• The spike-focused loss function effectively improves segmentation accuracy, particularly at the nanoparticle borders.
• Integration with microscope hardware enables real-time nanoparticle characterization, advancing materials science research.