We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Deep Neural Networks for Image Processing and Biomolecular Single-Particle 3D Reconstruction from Electron Micrographs
Summary
Researchers developed novel deep learning models for processing electron micrograph images and performing biomolecular single-particle 3D reconstruction, addressing the inverse problem of extracting accurate sample information from noisy electron micrographs affected by aberrations, astigmatism, and detection noise in imperfect imaging systems.
Electron microscopy (EM) has become a powerful tool for materials characterization. However, the existence of limiting factors in imperfect imaging systems, such as aberrations, astigmatism, detection noise, etc., makes extracting sample information from noisy electron micrographs an inverse problem. Differences in image formation models and features compared to optical cameras hinder the direct application of advanced computer vision techniques developed for visual objects to electron microscopy. This Thesis focuses on developing novel deep learning models for image processing of electron micrographs to enable faster and more accurate characterization of diverse samples. Specifically, (1) U-Net and its variants were employed for high-performance semantic/instance segmentation and shape classification, facilitating automatic quantification and classification of microplastics. (2) Noise2Noise, zero-shot Noise2Noise and Noise2SR were adapted to eliminate common corruptions in experimental SEM images of minerals, carbon nanotubes and mesoporous carbon. (3) Three deep learning models (VampPrior-SPR, ExemplarPrior-SPR and LSGM-SPR) were proposed for biomolecular heterogeneous 3D reconstruction from low-SNR 2D cryo-EM images. These models successfully resolved discrete compositional and continuous conformational heterogeneity across various simulated and experimental datasets. These advancements contribute to enhancing the capabilities of electron microscopy for broader applications with higher inference accuracy and speed.
Sign in to start a discussion.
More Papers Like This
Automatic quantification and classification of microplastics in scanning electron micrographs via deep learning
Researchers developed a deep learning system that can automatically detect and classify microplastics in scanning electron microscope images, replacing the time-consuming process of manual analysis. The system achieved high accuracy in identifying different types and shapes of microplastic particles, even very small ones that are difficult to spot by eye. This automated approach could significantly speed up microplastic monitoring and pollution assessment efforts.
No ground truth needed: unsupervised sinogram inpainting for nanoparticle electron tomography (UsiNet) to correct missing wedges
Researchers developed an unsupervised deep learning method called UsiNet for correcting missing wedge artifacts in electron tomography without needing ground truth data. The technique improves three-dimensional imaging of complex materials at the nanoscale, which is relevant for characterizing synthetic nanoparticles and microstructural domains.
Precise Sizing and Collision Detection of Functional Nanoparticles by Deep Learning Empowered Plasmonic Microscopy
Deep learning-empowered plasmonic microscopy (Deep-SM) enabled precise sizing and collision detection of individual nanoparticles on sensor surfaces by enhancing signal detection and suppressing noise from sequential image data, advancing single-nanoparticle analysis for biological and materials applications.
On the use of deep learning for phase recovery
Researchers reviewed how deep learning — a type of artificial intelligence — can recover phase information from light, which is typically lost when cameras capture images, enabling sharper microscopy and better materials analysis. These advances improve the tools scientists use to study tiny particles, including microplastics, at very fine scales.
A Comparative Study of State-of-the-Art Deep Learning Models for Semantic Segmentation of Pores in Scanning Electron Microscope Images of Activated Carbon
Researchers developed deep learning models to automatically identify and measure pores on activated carbon surfaces from electron microscope images. The study introduced a new dataset and tested several state-of-the-art models for this task, showing promising results compared to slow and costly manual analysis. The findings suggest that AI-based approaches could make quality assessment of activated carbon more efficient for applications in water purification and air filtration.