Light and electron microscopy continuum-resolution imaging of 3D cell cultures and image analysis using Dragonfly

Researchers have published a study on 3D cancer organoids using multiple imaging methods to capture disease development across scales. Non-invasive light microscopy that is essential for probing organoid development was correlated with high-resolution electron microscopy to image subcellular structures. After performing FIB-SEM volume imaging to capture 3D diffraction-limited subcellular structures, Dragonfly was used to train neural networks for automatic segmentation of FIB-SEM data. CNN models were prepared with the 2.5D-UNet architecture. Using the segmentation wizard module, ground truth data was manually annotated and CNNs models were trained to segment subcellular structures such as cell nuclei, mitochondria, microvillar actin bundles and cell-junctions. Because each acquisition includes different features and noise levels, it was necessary to re-train the CNNs on each acquisition. However, the procedure was much quicker through using the previously trained CNNs models. Dragonfly was also used to perform instance segmentation using nearest-neighbour analysis and to compute local density maps of segmentations using the "Bone analysis" wizard.

According to the first author Edoardo D’Imprima, "Dragonfly was essential to data-mine FIBSEM data of patient-derived cancer organoids. The user-friendly interface and the large amount of freely available teaching material allowed me to quickly master the software down to advanced operations. I believe Dragonfly will make Deep Learning approaches for bio-image analysis amenable to non-experts."

 

Video Presentation

External Content

This is a Youtube video.
To be able to watch, you need to accept Marketing Cookies.

Manage Consent

I agree with external content being displayed to me. Personal data may be transmitted to third-party platforms. More Information - Data protection

Publication:

D’Imprima, E., Montero, M.G., Gawrzak, S., Ronchi, P., Zagoriy, I., Schwab, Y., Jechlinger, M. and Mahamid, J., 2023. Light and electron microscopy continuum-resolution imaging of 3D cell cultures. Developmental Cell, 58(7), pp.616-632 (https://doi.org/10.1016/j.devcel.2023.03.001).

Keywords: FIB-SEM, Organoids, deep-learning image segmentation, cryo.confocal light microscopy, correlative light and electron microscopy

FIB-SEM volumes of patient-derived colorectal cancer organoids visualize fine ultrastructural details.

(A–K) Visualization and segmentation of subcellular structures using Dragonfly. In (A), (D), (G), and (H), FIB-SEM frames showing mitochondria (A–C), actin bundles in microvilli (D–F), and cell junctions (G and H, white arrows) and desmosomes (H, red arrows) overlaid with the corresponding automatic segmentation (B, E, I, and J, in color).

In (C), (F), and (K), 3D rendering of 50 consecutive frames with embossed segmentations. Frame indicates the areas shown in left panels. (L) 3D rendering of segmentations of nuclei (blue), actin bundles (magenta), and cell junctions (cyan). (M) 3D rendering of segmentation with nuclei (blue) and mitochondria colored according to their volumes (purple to yellow gradient); note the two-order of magnitude larger mitochondria around the nuclei of two cells (green-yellow).