AtomAI Brings Deep Learning to Microscopy Data Analysis Software

Package provides end-to-end analysis of microscopy images for accelerated materials research.

: Maxim Ziatdinov, Oak Ridge National Laboratory
An AI-generated image representing atoms and artificial neural networks.

The Science

Researchers have developed a new software package for analyzing images from electron and scanning probe microscopy. The package, AtomAI, uses deep learning. This is a type of machine learning that allows a program to train itself to accurately identify the contents of an image or block of text. Deep learning models automatically learn relevant features by using a network of layered “neurons.” These artificial neurons, inspired by biological neurons, serve as nodes through which data and computations flow. They are trained to detect various aspects of an image at different levels of complexity. This gives AtomAI greater precision than traditional machine learning and the ability to analyze a wider range of information.

The Impact

Electron and scanning probe microscopy allow scientists to engineer materials at the nanoscale. These techniques help scientists study a material’s structure and its functional properties. The AtomAI package applies deep learning to microscopy data at atomic resolutions. This provides quantifiable physical information such as the precise position and type of each atom in a sample. AtomAI also allows researchers to conduct real-time analysis of data. The package can import this information directly into theoretical simulations to gain more insight into a material’s structure.

Summary

AtomAI, which was developed partially at Oak Ridge National Laboratory (ORNL) Center for Nanophase Materials Science, is an end-to-end software package for image analysis. The package includes a unique model architecture to identify thin objects such as nanofibers or domain walls — the interfaces separating magnetic domains — in microscopy data. The software package is also built to reduce errors in image processing by accounting for unintended changes in the image data, such as incoming cosmic rays or images of non-target materials, and by incorporating certain unchanging physical characteristics into the model.

Electron and scanning probe microscopes have become critical tools for condensed matter physics, materials science, and chemistry research. However, researchers have lacked infrastructure to establish broad connections between microscopy observations and materials behavior. AtomAI is designed to help address this gap.

Contact

Maxim Ziatdinov
Oak Ridge National Laboratory
[email protected]

Funding

This effort was performed and partially supported by Oak Ridge National Laboratory’s Center for Nanophase Materials Sciences, a Department of Energy (DOE) Office of Science user facility, and by the DOE Office of Science, Office of Basic Energy Sciences Data, Artificial Intelligence and Machine Learning at DOE Scientific user facilities program under the Digital Twin Project and MLExchange Project.

Publications

Ziatdinov, M., Ghosh, A., Wong, C.Y., & Kalinin, S.V., AtomAI framework for deep learning analysis of image and spectroscopy data in electron and scanning probe microscopy. Nature Machine Intelligence 4, 1101–1112 (2022). [DOI: 10.1038/s42256-022-00555-8]

Related Links

ORNL News: Deep learning–based data analysis software by ORNL promises to accelerate materials research

Learn more about AtomAI at its Github page.

Highlight Categories

Program: BES , SUF

Performer: University , DOE Laboratory , SC User Facilities , BES User Facilities , CNMS