Citation

If you use MicroStructFormer, the released model weights, or the associated EM ultrastructure segmentation workflow in your research, please cite:

Deep learning-based decoding of axonal ultrastructure in gene-edited mice using electron microscopy imaging
Nianqi Deng, Guillaume Miao, Anmar Khadra, Alan C Peterson, Hooman Bagheri
bioRxiv 2026.05.26.727755; doi: https://doi.org/10.64898/2026.05.26.727755

BibTeX Format

For LaTeX users or reference managers, please use the following BibTeX entry:

@article {Deng2026.05.26.727755,
    author = {Deng, Nianqi and Miao, Guillaume and Khadra, Anmar and Peterson, Alan C and Bagheri, Hooman},
    title = {Deep learning-based decoding of axonal ultrastructure in gene-edited mice using electron microscopy imaging},
    elocation-id = {2026.05.26.727755},
    year = {2026},
    doi = {10.64898/2026.05.26.727755},
    publisher = {Cold Spring Harbor Laboratory},
    abstract = {Myelin forms an insulating sheath around axons enabling both rapid and energy-efficient conduction of action potentials and myelin abnormalities or loss can lead to severe motor, sensory, and cognitive impairment. While electron microscopy can resolve multiple axonal components that are affected myelin, their large-scale quantitative analysis is both difficult and time consuming. To overcome such limitations, we developed a machine learning framework that automatically recognizes and quantifies multiple features of axons and myelin including axonal mitochondrial density and periaxonal area. Applying that framework to fibers in the spinal cord of variably hypomyelinated mice, we show here that reduction in the thickness and length of myelin sheaths results in correlating changes in mitochondrial density and periaxonal area. The machine learning framework introduced here should contribute to future insight into the axon, myelin, and mitochondrial relationships that change during neurological plasticity and myelin disease progression.Competing Interest StatementThe authors have declared no competing interest.Natural Sciences and Engineering Research Council of Canada (NSERC), RGPIN-2019-04520, ALLRP 588367-23China Scholarship CouncilUNIQUE},
    URL = {https://www.biorxiv.org/content/early/2026/05/27/2026.05.26.727755},
    eprint = {https://www.biorxiv.org/content/early/2026/05/27/2026.05.26.727755.full.pdf},
    journal = {bioRxiv}
}

External tool used in the reference pipeline

When analyses require explicit delineation of axon and myelin regions, an external axon–myelin segmentation tool is needed. In the reference pipeline described here, AxonDeepSeg is used to provide axon and myelin masks that complement the subcellular and periaxonal segmentation produced by MicroStructFormer.

For the use of AxonDeepSeg, please refer to their official documentation and cite the following publication:

Zaimi, A., Wabartha, M., Herman, V., Antonsanti, P.-L., Perone, C. S., & Cohen-Adad, J. (2018). AxonDeepSeg: automatic axon and myelin segmentation from microscopy data using convolutional neural networks. Scientific Reports, 8(1), 3816.
Link to the AxonDeepSeg paper.