CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy

Understanding the complex three-dimensional structure of cells is crucial across many disciplines in biology and especially in neuroscience. Here, we introduce a set of models including a 3D transformer (SwinUNetR) and a novel 3D self-supervised learning method (WNet3D) designed to address the inher...

Full description

Saved in:
Bibliographic Details
Main Authors: Cyril Achard, Timokleia Kousi, Markus Frey, Maxime Vidal, Yves Paychere, Colin Hofmann, Asim Iqbal, Sebastien B Hausmann, Stéphane Pagès, Mackenzie Weygandt Mathis
Format: Article
Language:English
Published: eLife Sciences Publications Ltd 2025-06-01
Series:eLife
Subjects:
Online Access:https://elifesciences.org/articles/99848
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849685012492845056
author Cyril Achard
Timokleia Kousi
Markus Frey
Maxime Vidal
Yves Paychere
Colin Hofmann
Asim Iqbal
Sebastien B Hausmann
Stéphane Pagès
Mackenzie Weygandt Mathis
author_facet Cyril Achard
Timokleia Kousi
Markus Frey
Maxime Vidal
Yves Paychere
Colin Hofmann
Asim Iqbal
Sebastien B Hausmann
Stéphane Pagès
Mackenzie Weygandt Mathis
author_sort Cyril Achard
collection DOAJ
description Understanding the complex three-dimensional structure of cells is crucial across many disciplines in biology and especially in neuroscience. Here, we introduce a set of models including a 3D transformer (SwinUNetR) and a novel 3D self-supervised learning method (WNet3D) designed to address the inherent complexity of generating 3D ground truth data and quantifying nuclei in 3D volumes. We developed a Python package called CellSeg3D that provides access to these models in Jupyter Notebooks and in a napari GUI plugin. Recognizing the scarcity of high-quality 3D ground truth data, we created a fully human-annotated mesoSPIM dataset to advance evaluation and benchmarking in the field. To assess model performance, we benchmarked our approach across four diverse datasets: the newly developed mesoSPIM dataset, a 3D platynereis-ISH-Nuclei confocal dataset, a separate 3D Platynereis-Nuclei light-sheet dataset, and a challenging and densely packed Mouse-Skull-Nuclei confocal dataset. We demonstrate that our self-supervised model, WNet3D – trained without any ground truth labels – achieves performance on par with state-of-the-art supervised methods, paving the way for broader applications in label-scarce biological contexts.
format Article
id doaj-art-ec9093310ef741babed0a29b14768417
institution DOAJ
issn 2050-084X
language English
publishDate 2025-06-01
publisher eLife Sciences Publications Ltd
record_format Article
series eLife
spelling doaj-art-ec9093310ef741babed0a29b147684172025-08-20T03:23:18ZengeLife Sciences Publications LtdeLife2050-084X2025-06-011310.7554/eLife.99848CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopyCyril Achard0https://orcid.org/0009-0003-6992-6928Timokleia Kousi1Markus Frey2https://orcid.org/0000-0003-0291-3391Maxime Vidal3Yves Paychere4Colin Hofmann5Asim Iqbal6https://orcid.org/0000-0003-2174-4554Sebastien B Hausmann7Stéphane Pagès8https://orcid.org/0000-0003-0618-777XMackenzie Weygandt Mathis9https://orcid.org/0000-0001-7368-4456Brain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandWyss Center for Bio and Neuroengineering, Geneva, SwitzerlandBrain Mind Institute and Neuro X, École Polytechnique Fédérale de Lausanne (EPFL), Geneva, SwitzerlandUnderstanding the complex three-dimensional structure of cells is crucial across many disciplines in biology and especially in neuroscience. Here, we introduce a set of models including a 3D transformer (SwinUNetR) and a novel 3D self-supervised learning method (WNet3D) designed to address the inherent complexity of generating 3D ground truth data and quantifying nuclei in 3D volumes. We developed a Python package called CellSeg3D that provides access to these models in Jupyter Notebooks and in a napari GUI plugin. Recognizing the scarcity of high-quality 3D ground truth data, we created a fully human-annotated mesoSPIM dataset to advance evaluation and benchmarking in the field. To assess model performance, we benchmarked our approach across four diverse datasets: the newly developed mesoSPIM dataset, a 3D platynereis-ISH-Nuclei confocal dataset, a separate 3D Platynereis-Nuclei light-sheet dataset, and a challenging and densely packed Mouse-Skull-Nuclei confocal dataset. We demonstrate that our self-supervised model, WNet3D – trained without any ground truth labels – achieves performance on par with state-of-the-art supervised methods, paving the way for broader applications in label-scarce biological contexts.https://elifesciences.org/articles/99848self-supervised learningartificial intelligenceneurosciencemesoSPIMconfocal microscopyplatynereis
spellingShingle Cyril Achard
Timokleia Kousi
Markus Frey
Maxime Vidal
Yves Paychere
Colin Hofmann
Asim Iqbal
Sebastien B Hausmann
Stéphane Pagès
Mackenzie Weygandt Mathis
CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy
eLife
self-supervised learning
artificial intelligence
neuroscience
mesoSPIM
confocal microscopy
platynereis
title CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy
title_full CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy
title_fullStr CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy
title_full_unstemmed CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy
title_short CellSeg3D, Self-supervised 3D cell segmentation for fluorescence microscopy
title_sort cellseg3d self supervised 3d cell segmentation for fluorescence microscopy
topic self-supervised learning
artificial intelligence
neuroscience
mesoSPIM
confocal microscopy
platynereis
url https://elifesciences.org/articles/99848
work_keys_str_mv AT cyrilachard cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT timokleiakousi cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT markusfrey cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT maximevidal cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT yvespaychere cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT colinhofmann cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT asimiqbal cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT sebastienbhausmann cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT stephanepages cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy
AT mackenzieweygandtmathis cellseg3dselfsupervised3dcellsegmentationforfluorescencemicroscopy