CoChat
Sign Up
Cited by 48

The DOI matches its Crossref record; 5 of the 5 references it cites were found in a registry or an index.

Single‐cell segmentation in bacterial biofilms with an optimized deep learning method enables tracking of cell lineages and measurements of growth rates

Jelli, Eric, Ohmura, Takuya, Netter, Niklas, Abt, Martin, Jiménez‐Siebert, Eva, Neuhaus, Konstantin, Rode, Daniel K. H., Nadell, Carey D., Drescher, Knut·2023·Molecular Microbiology
DOI 10.1111/mmi.15064
Open this paper in CoChat — free account

Abstract

Bacteria often grow into matrix‐encased three‐dimensional (3D) biofilm communities, which can be imaged at cellular resolution using confocal microscopy. From these 3D images, measurements of single‐cell properties with high spatiotemporal resolution are required to investigate cellular heterogeneity and dynamical processes inside biofilms. However, the required measurements rely on the automated segmentation of bacterial cells in 3D images, which is a technical challenge. To improve the accuracy of single‐cell segmentation in 3D biofilms, we first evaluated recent classical and deep learning segmentation algorithms. We then extended StarDist, a state‐of‐the‐art deep learning algorithm, by optimizing the post‐processing for bacteria, which resulted in the most accurate segmentation results for biofilms among all investigated algorithms. To generate the large 3D training dataset required for deep learning, we developed an iterative process of automated segmentation followed by semi‐manual correction, resulting in >18,000 annotated Vibrio cholerae cells in 3D images. We demonstrate that this large training dataset and the neural network with optimized post‐processing yield accurate segmentation results for biofilms of different species and on biofilm images from different microscopes. Finally, we used the accurate single‐cell segmentation results to track cell lineages in biofilms and to perform spatiotemporal measurements of single‐cell growth rates during biofilm development.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1111/mmi.15064

Cites

5 of 5 references verified0 citation errors · 0 not checkable

Statistics

Figures

Image 1
CC-BY-NC · From the PDF
Image 2
CC-BY-NC · From the PDF
FIGURE 3 Algorithm for optimizing post-­ processing of the StarDist segmentation output avoids over-­ segmentation. (a) When the StarDist convolutional neural network (CNN) model is used to generate a single-­ cell segmentation of a 3D raw image, the output of StarDist is a 3D map of the 'label probability', which quantifies the probability of each pixel to be inside a cell. In addition, StarDist generates multiple 3D distance maps, which describe the length of the StarDist rays spanning the star-­ convex polyhedra that approximate the cell shapes. Depicted are three distance maps on top of each other, using RGB values of the three distance map values. (b) Schematic illustration of the strategy underlying StarDist OPP: In Step 1, the pixel with the highest label probability is chosen and the corresponding polyhedron P 1 is constructed based on the distance map entries at that position, and P 1 is assumed to be the current cell shape. In Step 2, the polyhedron P 2 corresponding to the pixel within the current cell shape with the next highest probability is generated. In Step 3, cell shapes P 1 and P 2 are fused together to yield a new cell shape P 1 U P 2 . Steps 2 and 3 are iteratively repeated until the volume V of the cell shape reaches V max and no longer increases (see schematic graph), which yields the final fused cell segmentation label (Step 4). (c) Comparison of predictions by StarDist (left) and StarDist OPP (right) for the 3D region depicted in panel A. Cell labels in StarDist involve the overlap of regions with high label probability, which indicate segmentation problems. Colors indicate different cells, overlaps between different cell predictions are white, and the background is black.
FIGURE 3 Algorithm for optimizing post-­ processing of the StarDist segmentation output avoids over-­ segmentation. (a) When the StarDist convolutional neural network (CNN) model is used to generate a single-­ cell segmentation of a 3D raw image, the output of StarDist is a 3D map of the 'label probability', which quantifies the probability of each pixel to be inside a cell. In addition, StarDist generates multiple 3D distance maps, which describe the length of the StarDist rays spanning the star-­ convex polyhedra that approximate the cell shapes. Depicted are three distance maps on top of each other, using RGB values of the three distance map values. (b) Schematic illustration of the strategy underlying StarDist OPP: In Step 1, the pixel with the highest label probability is chosen and the corresponding polyhedron P 1 is constructed based on the distance map entries at that position, and P 1 is assumed to be the current cell shape. In Step 2, the polyhedron P 2 corresponding to the pixel within the current cell shape with the next highest probability is generated. In Step 3, cell shapes P 1 and P 2 are fused together to yield a new cell shape P 1 U P 2 . Steps 2 and 3 are iteratively repeated until the volume V of the cell shape reaches V max and no longer increases (see schematic graph), which yields the final fused cell segmentation label (Step 4). (c) Comparison of predictions by StarDist (left) and StarDist OPP (right) for the 3D region depicted in panel A. Cell labels in StarDist involve the overlap of regions with high label probability, which indicate segmentation problems. Colors indicate different cells, overlaps between different cell predictions are white, and the background is black.CC-BY-NC · From the PDF
Image 4
CC-BY-NC · From the PDF
Image 5
CC-BY-NC · From the PDF
FIGURE 6 Based on accurate single-­ cell segmentation in biofilms, all individual cells can be tracked and cellular growth rates can be measured. (a, b) Tracked trajectories of cells during Vibrio cholerae biofilm formation are shown in light gray. A few example trajectories are indicated by dark gray and black lines. Panel (a) shows the trajectories in xyz -­ space. Panel (b) shows the trajectories in the xy -­ plane and time. (c) Spatiotemporal cell division rate during biofilm formation. The number of cells in the biofilm ( N cell ) indicates the time of development. The y -­ axis shows the distance between the cell center and the center of the biofilm at the bottom of the biofilm (in the z = 0 plane). Inset: Schematic figure illustrates the coordinates used in panels (c) and (d). (d) Spatial distribution of cell division events in a Vibrio cholerae biofilm that grew from 2074 cells to 5190 cells during the measurement interval. The horizontal axis shows the distance between the xy -­ projected cell centroid and the bottom center of the biofilm. The vertical axis shows the z -­ position of the cell centroid.
FIGURE 6 Based on accurate single-­ cell segmentation in biofilms, all individual cells can be tracked and cellular growth rates can be measured. (a, b) Tracked trajectories of cells during Vibrio cholerae biofilm formation are shown in light gray. A few example trajectories are indicated by dark gray and black lines. Panel (a) shows the trajectories in xyz -­ space. Panel (b) shows the trajectories in the xy -­ plane and time. (c) Spatiotemporal cell division rate during biofilm formation. The number of cells in the biofilm ( N cell ) indicates the time of development. The y -­ axis shows the distance between the cell center and the center of the biofilm at the bottom of the biofilm (in the z = 0 plane). Inset: Schematic figure illustrates the coordinates used in panels (c) and (d). (d) Spatial distribution of cell division events in a Vibrio cholerae biofilm that grew from 2074 cells to 5190 cells during the measurement interval. The horizontal axis shows the distance between the xy -­ projected cell centroid and the bottom center of the biofilm. The vertical axis shows the z -­ position of the cell centroid.CC-BY-NC · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 5 references

Research works printed but not linked: 66

  1. 10.1038/nrmicro3491 · DOI
  2. Ahrens J., 2005, ParaView: an end‐user tool for large data visualization
  3. 10.1038/s41564‐021‐00900‐4 · DOI
  4. 10.1002/cyto.a.23863 · DOI
  5. 10.1101/491035 · DOI · Paper page
  6. 10.48550/arXiv.1512.01274 · DOI · Paper page
  7. 10.1007/978-3-319-46723-8_49 · DOI · Paper page
  8. 10.1146/annurev.mi.41.100187.002251 · DOI · Paper page
  9. 10.1038/scientificamerican0178‐86 · DOI
  10. 10.1371/journal.pone.0268199 · DOI
  11. 10.1038/s41592‐022‐01639‐4 · DOI
  12. 10.7554/eLife.70794 · DOI
  13. 10.1073/pnas.1601702113 · DOI
  14. 10.1038/nmicrobiol.2016.77 · DOI
  15. 10.14440/jbm.2014.36 · DOI
  16. 10.1038/s41592‐022‐01507‐1 · DOI
  17. 10.1016/j.tim.2020.03.008 · DOI
  18. 10.1038/s41592‐018‐0261‐2 · DOI
  19. 10.1038/nrmicro2415 · DOI · Paper page
  20. 10.1038/s41579‐019‐0158‐9 · DOI · Paper page
  21. 10.1128/JB.01826‐08 · DOI
  22. Griffa A., 2010, Comparison of deconvolution software in 3D microscopy: a user point of view—part 1, G.I.T. Imaging & Microscopy, 12, 43
  23. 10.1186/s12915‐022‐01453‐6 · DOI
  24. 10.1038/s41586‐020‐2649‐2 · DOI · Paper page
  25. 10.1038/s41564‐020‐00817‐4 · DOI
  26. 10.1038/s41567‐018‐0356‐9 · DOI
  27. 10.1111/mmi.14501 · DOI
  28. 10.1371/journal.pbio.3001846 · DOI
  29. 10.1093/femsre/fuaa062 · DOI
  30. 10.1038/s41579‐022‐00692‐2 · DOI
  31. 10.1111/j.1365‐2818.2012.03675.x · DOI
  32. 10.1038/s41592‐021‐01284‐3 · DOI
  33. 10.1371/journal.pcbi.1007673 · DOI
  34. 10.25080/Majora‐92bf1922‐00a · DOI
  35. 10.1038/s41592‐019‐0403‐1 · DOI · Paper page
  36. 10.1038/s41522‐021‐00214‐7 · DOI · Paper page
  37. 10.1371/journal.pcbi.1009797 · DOI
  38. 10.1111/mmi.13264 · DOI
  39. 10.1109/ICCVW54120.2021.00110 · DOI
  40. 10.1126/science.abb8501 · DOI
  41. 10.1128/JB.180.3.722‐731.1998 · DOI
  42. 10.1046/j.1365‐2958.1998.00791.x · DOI
  43. 10.1038/nmeth.2019 · DOI · Paper page
  44. 10.1007/978-3-030-00934-2_30 · DOI · Paper page
  45. 10.48550/arXiv.1802.05799 · DOI
  46. 10.1128/mBio.00103‐13 · DOI
  47. Sofroniew N., 2022, Napari: a multi‐dimensional image viewer for python
  48. 10.1021/la401322k · DOI
  49. 10.1038/nrmicro1838 · DOI · Paper page
  50. 10.1186/s12859‐021‐04344‐9 · DOI
  51. 10.1038/s41592‐020‐01018‐x · DOI · Paper page
  52. 10.1101/2021.06.11.448083 · DOI
  53. 10.1186/s12880‐015‐0068‐x · DOI
  54. The Pandas Development Team, 2023, Pandas‐Dev/Pandas: Pandas
  55. 10.1016/j.ymeth.2016.09.016 · DOI
  56. 10.1109/ISBI52829.2022.9761655 · DOI
  57. 10.1042/BST20180391 · DOI
  58. 10.1038/s41467‐021‐22518‐0 · DOI
  59. 10.1016/j.cell.2021.11.024 · DOI
  60. 10.1038/s41592‐018‐0216‐7 · DOI
  61. 10.1109/WACV45572.2020.9093435 · DOI
  62. 10.7554/eLife.57613 · DOI
  63. 10.1073/pnas.1611494113 · DOI
  64. 10.1364/BOE.420788 · DOI
  65. 10.1038/s41522‐022‐00362‐4 · DOI · Paper page
  66. 10.1038/s41467‐020‐19866‐8 · DOI

Cited by

Sorted by citations

See all 48

Metadata

Authors
Jelli, E., Ohmura, T., Netter, N., Abt, M., Jiménez‐Siebert, E., Neuhaus, K., Rode, D. K. H., Nadell, C. D., & Drescher, K.
Venue
Molecular Microbiology
Publisher
Wiley
Published
2023-04-17
Volume
119
Issue
6
Pages
659-676
ISSN
  • 0950-382X
  • 1365-2958
DOI
10.1111/mmi.15064
Cited by
48
References
5
Language
en
Licence
creativecommons.org/licenses/by-nc/4.0/ (vor)
Funders
  • Bundesministerium für Bildung und Forschung (TARGET‐Biofilms)
  • Deutsche Forschungsgemeinschaft (DR 982/5‐1, DR 982/6‐1)
  • European Commission (716734, 955910)
  • Human Frontier Science Program (LT000013/2019‐C)
  • Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (51NF40_180541, TMCG‐3_213801)

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Single‐cell segmentation in bacterial biofilms with an optimized deep learning method enables tracking of cell lineages and measurements of growth rates. https://app.cochat.ai/papers/10.1111/mmi.15064
The paper
Jelli, E., Ohmura, T., Netter, N., Abt, M., Jiménez‐Siebert, E., Neuhaus, K., Rode, D. K. H., Nadell, C. D., & Drescher, K. (2023). Single‐cell segmentation in bacterial biofilms with an optimized deep learning method enables tracking of cell lineages and measurements of growth rates. Molecular Microbiology. https://doi.org/10.1111/mmi.15064