Skip to main content

ACORN Methods Paper Reproducibility Package: Turning Large Electron Microscopy Datasets into Quantitative Measurements with Model-Assisted Analysis

    Alexis N Williams | Oak Ridge National Laboratory
    Chanda R. Harris | Oak Ridge National Laboratory
    Juhyun Oh | Oak Ridge National Laboratory
    Michael J Zachman | Oak Ridge National Laboratory
    Albina Y Borisevich | Oak Ridge National Laboratory
Download Dataset on Globus

Description

This dataset contains the experimental micrographs and synthetic training data used to evaluate ACORN (Annotate, Curate, Observe, Review, and Navigate), an open-source application for converting electron microscopy (EM) images into calibrated, quantitative measurements. The experimental data include 30 low-dose cryogenic transmission electron microscopy (cryo-TEM) micrographs of poly(lactic-co-glycolic acid) (PLGA) nanoparticles in vitreous ice and 26 scanning electron microscopy (SEM) micrographs of uncoated bacterial spores. The cryo-TEM micrographs were acquired on a Titan Krios G4 at 300 kV with a Gatan K3 detector (2.229 Å/pixel). The SEM micrographs were acquired on a Zeiss Merlin at 1 kV (3.6 to 19.3 nm/pixel). All micrographs are provided in their native MRC and TIFF formats with instrument calibration retained in the file headers. The synthetic data comprise two datasets generated with the ACORN TEM and SEM simulators: 400 cryo-TEM images of nanoparticles and 400 SEM images of spores. Each dataset is split into 280 training, 60 validation, and 60 test images. Each image has exact instance segmentation labels in YOLO polygon format and a manifest of its simulation parameters, including pixel size, dose, defocus, ice thickness, coating, and accelerating voltage. The synthetic datasets contain 18,354 and 24,242 labeled objects, respectively. These data support training and benchmarking segmentation models for EM and testing measurement accuracy against known object geometries. ACORN software can be found at https://github.com/vitrifyai/acorn_software

Funding Information

DOE Contract Number

AC05-00OR22725

Originating Research Organization

Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)

Sponsoring Organization

Office of Science (SC)

Related Works

Details

DOI

https://doi.org/10.13139/ORNLNCCS/3920044

Release Date

November 2, 2026

Subject

MATERIALS SCIENCE

Keywords

electron microscopy, scanning electron microscopy, instance segmentation, synthetic training data, nanoparticles, bacterial spores, quantitative image analysis

Dataset

Dataset Type

IP Still Images or Photos

Software

ACORN

Cite This Dataset:

Williams, A., Harris, C., Oh, J., Zachman, M., Borisevich, A., Hoglund, E., Haque, A., Harris, S., Lasseter, J., Plechaty, A., Hochanadel, L., Vasileva, D., Morrell-Falvey, J., Michener, J., Davern, S., Sanders, B., Retterer, S. (2026). ACORN Methods Paper Reproducibility Package: Turning Large Electron Microscopy Datasets into Quantitative Measurements with Model-Assisted Analysis. Oak Ridge National Laboratory. https://doi.org/10.13139/ORNLNCCS/3920044.

Acknowledgements

This work was carried out [in part] at Oak Ridge National Laboratory, managed by UT-Battelle, LLC for the U.S. Department of Energy under contract DE-AC05-00OR22725.