Deep Learning-Based Segmentation of Cryo-Electron Tomograms

November 11, 2022

Jessica E. Heebner (1), Carson Purnell (1), Ryan K. Hylton (1), Mike Marsh (2), Michael A. Grillo (1), Matthew T. Swulius (1)
JoVE Journal, Biology. (November 11, 2022). DOI: 10.3791/64435


Cryo-electron tomography (cryo-ET) allows researchers to image cells in their native, hydrated state at the highest resolution currently possible. The technique has several limitations, however, that make analyzing the data it generates time-intensive and difficult. Hand segmenting a single tomogram can take from hours to days, but a microscope can easily generate 50 or more tomograms a day. Current deep learning segmentation programs for cryo-ET do exist, but are limited to segmenting one structure at a time. Here, multi-slice U-Net convolutional neural networks are trained and applied to automatically segment multiple structures simultaneously within cryo-tomograms. With proper preprocessing, these networks can be robustly inferred to many tomograms without the need for training individual networks for each tomogram. This workflow dramatically improves the speed with which cryo-electron tomograms can be analyzed by cutting segmentation time down to under 30 min in most cases. Further, segmentations can be used to improve the accuracy of filament tracing within a cellular context and to rapidly extract coordinates for subtomogram averaging.

How Our Software Was Used

Dragonfly was used for preprocessing and annotating cryo-electron tomograms to train artificial neural networks for multi-class segmentation.

Author Affiliation

(1) Pennsylvania State University-College of Medicine
(2) Object Research Systems