SELMA3D 2025: Self-supervised learning for 3D light-sheet microscopy image segmentation

In modern biological research, the ability to visualize and analyze complex structures within tissues and organisms is crucial. Traditional imaging techniques often struggle to provide a cellular-resolution, 3D view of bio-samples while preserving their structural integrity. The combination of tissue clearing and light-sheet microscopy (LSM) overcomes these limitations, serving as a powerful method for high-contrast, ultra-high-resolution imaging. This approach enabled detailed visualization of a wide range of biological structures, including cellular and subcellular structures, organelles and processes, across diverse samples [1]. Tissue clearing techniques render inherently opaque biological samples transparent, allowing light to penetrate deeply into the tissue [2] and imaging reagents (e.g., fluorophores or antibodies), while preserving their structural integrity and molecular content. Various fluorophores or antibodies can be employed to selectively stain specific biological structures within samples and enhance their contrast under microscopy [3]. After staining and tissue clearing, LSM provides rapid 3D imaging of intricate biological structures with high spatial resolution, offering valuable insights into various biomedical fields, such as neuroscience [4], immunology [5], oncology [6] and cardiology [7].

Automated image analysis approaches enable scientists to extract structural and functional, cellular and subcellular information from LSM images of various biosamples at an accelerated pace. To analyze LSM images, segmentation plays a pivotal and essential role in identifying and distinguishing different biological structures [8]. For large LSM images, such as those of whole organs or organisms, manual segmentation is time-intensive, with individual images containing up to 10000^3 voxels. As a result, there is a growing demand for automatic segmentation methods. Recent strides in deep learning-based segmentation models offer promising solutions for automating the segmentation of LSM images [9-10]. While these models achieve performance comparable to expert human annotators, their success largely relies on supervised learning, which requires extensive, high-quality manual annotations. These models are usually task-specific, designed for particular structures, with limited generalizability across different applications [11]. Therefore, the widespread adaptation of deep learning-based segmentation models is constrained, as the annotation for every specific LSM  image segmentation task requires experts with domain knowledge, making the process impractical for many scenarios. It is crucial to develop generalizable models capable of serving multiple LSM image segmentation tasks.

Self-supervised learning offers significant advantages in this regard, as it allows deep learning models to pretrain on large-scale, unannotated datasets, thereby learning useful and generalizable representations of LSM image data. Subsequently, the model can be fine-tuned on a smaller labeled dataset for specific segmentation tasks [12].  Notably, self-supervised learning has not been extensively explored within the LSM field, despite the presence of vast sets of LSM data of different biological structures. Some properties of LSM images e.g. the high signal-to-noise ratio, makes them particularly well-suited for self-supervised learning.

The SELMA3D 2024 challenge represents a significant advancement in self-supervised learning research within the field of 3D LSM images. To the best of our knowledge, it is the first attempt to benchmark self-supervised learning for LSM image segmentation tasks. Self-supervised learning offers potential benefits for models across a range of downstream tasks. The SELMA3D challenge focuses on segmentation, a crucial task in LSM image analysis. In the 2024 edition, we categorized biological structures frequently studied in LSM images into two main types based on morphology: tree-like structures, including vessels and microglia, and spot-like structures, including cell nuclei, c-Fos+ cells and amyloid-beta plaques. Participants were asked to develop a universal self-supervised learning approach for 3D LSM semantic segmentation, one that can benefit segmentation of both types of structures. The top performing teams achieved Dice scores exceeding 70% for both structure types. However, the results obtained by different participants suggested a single self-supervised learning strategy struggles to consistently enhance feature learning for both types simultaneously.

Based on the results and experience from 2024, for the second edition of the SELMA3D challenge hold in 2025, we propose to expand and improve the challenge in the following aspects:

Firstly, we redefine the classification of biological structures into two categories: isolated structures and contiguous structures. Isolated structures refer to distinct, spatially separate components without physical connections, for example cell nuclei, c-Fos+ cell, amyloid-beta plaques and microglia. In contrast, contiguous structures highlight the physical continuity between parts of a structure that are connected without interruption, for example vessels and nerves. Based on this classification, the SELMA3D 2025 challenge will be divided into two tasks: 1) self-supervised segmentation of isolated structures in 3D light-sheet microscopy images, 2) self-supervised segmentation of contiguous structures in 3D light-sheet microscopy images. This division reduces the time required for participants to download and preprocess training datasets for each task. Moreover, it allows participants to develop self-supervised learning strategies tailored to the morphological characteristics of specific structures. Participants can select a task based on their research interests and available time, providing greater flexibility and focus.

Secondly, we aim to vastly enhance the dataset by expanding both the number and diversity of samples. In SELMA3D 2024, our dataset comprised 35 whole-brain or brain-subregion 3D images, with 9 images of contiguous structures and 26 of isolated structures. This year, through collaborations with other laboratories and research SELMA3D 2025: Self-supervised learning for 3D light-sheet microscopy institutes, e.g. Alain Chédotal’s lab [13-15], we increase the total to 58 3D images, including 28 images of contiguous structures and 30 of isolated structures. Additionally, we have expanded the quantity of annotated patches, providing a richer foundation for model training. Furthermore, while last year’s images were limited to brain regions, this year we have incorporated images from various organs and regions across the body to enhance data diversity. We have also broadened the range of biological structures included. The SELMA3D 2024 dataset contained c-Fos+ cells, cell nuclei, amyloid-beta plaques, blood vessels, and microglia. This year, we have expanded it to include neurons, lymphatic cells, and fluorescent proteins for isolated structure segmentation, as well as peripheral nerves, gut nerves, and lymphatic vessels for contiguous structure segmentation. This expansion enhances the foundation for model training by diversifying the dataset and providing participants with a broader range of examples to develop and evaluate their approaches. Ultimately, the increased dataset size and diversity will enable a more in-depth exploration of how self-supervised learning can improve performance and generalize across various tasks.

Thirdly, we plan to expand the quantitative evaluations for the segmentation tasks. Building on the insights gained from SELMA3D 2024, we will incorporate a more comprehensive set of evaluation metrics tailored to the morphological characteristics of the targeted structures. These enhanced segmentation evaluations are designed  to provide deeper insights into model performance, addressing both accuracy and morphological consistency. By offering detailed and structure-specific evaluations, we aim to better guide the development of advanced self-supervised learning strategies and improve the interpretability of results. Given the above aspects, we aim to optimize the challenge setting, establishing a more comprehensive benchmark for self-supervised learning in 3D LSM image segmentation. We look forward to organizing the second edition of the SELMA3D challenge and welcoming submissions.

References:
[1] E.H.K. Stelzer, F. Strobl, B. Chang, et al. Light sheet fluorescence microscopy. Nature Reviews Methods Primers 1(1): 73, 2021 Nov.
[2] H.R. Ueda, A. Ertürk, K. Chung, et al. Tissue clearing and its applications in neuroscience. Nature Reviews Neuroscience 21(2): 61-79, 2020, Jan.
[3] P.K. Poola, M.I. Afzal, Y. Yoo, et al. Light sheet microscopy for histopathology applications. Biomedical engineering letters 9: 279-291, 2019 July.
[4] H.R. Ueda, H.U. Dodt, P. Osten, et al. Whole-brain profiling of cells and circuits in mammals by tissue clearing and light-sheet microscopy. Neuron, 106(3): 369-387, 2020 May.
[5] D. Zhang, A.H. Cleveland, E. Krimitza, et al. Spatial analysis of tissue immunity and vascularity by light sheet fluorescence microscopy. Nature Protocols: 1-30, 2024 Jan.
[6] J. Almagro, H.A. Messal, M.Z. Thin, et al. Tissue clearing to examine tumour complexity in three dimensions. Nature Reviews Cancer, 21(11): 718-730, 2021 July.
[7] P. Fei, J. Lee, R.R.S.Packard, et al. Cardiac light-sheet fluorescent microscopy for multi-scale and rapid imaging of architecture and function. Scientific Reports 6: 22489, 2016 Mar.
[8] F. Amat, B. Höckendorf, Y. Wan, et al. Efficient processing and analysis of large-scale light-sheet microscopy data. Nature protocols 10: 2015: 1679-1696, 2015 Oct.
[9] N. Kumar, P. Hrobar, M. Vagenknecht, et al. A Light sheet fluorescence microscopy and machine learning-based approach to investigate drug and biomarker distribution in whole organs and tumors. bioRxiv 2023.09.16.558068.
[10] M.I. Todorov, J.C. Paetzold, O. Schoppe, et al. Machine learning analysis of whole mouse brain vasculature. Nature Methods 17: 442-449, 2020 Mar.
[11] Y. Zhou, M.A. Chia, S.K. Wagner, et al. A foundation model for generalizable disease detection from retinal images. Nature 622: 156–163, 2023 Sept.
[12] R. Krishnan, P. Rajpurkar, E.J. Topol. Self-supervised learning in medicine and healthcare. Nature Biomedical Engineering 6: 1346-1352, 2022 Aug.
[13] M. Belle, D. Godefroy, C. Dominici, et al. A simple method for 3D analysis of immunolabeled axonal tracts in a transparent nervous system. Cell reports 9(4): 1191-1201, 2014 Nov.
[14] M. Belle, D. Godefroy, G. Couly, et al. Tridimensional visualization and analysis of early human development. Cell 169(1): 161-173, 2017 Mar.
[15] R. Blain, G. Couly, E. Shotar, et al. A tridimensional atlas of the developing human head. Cell 186(26): 5910-5924, 2023 Dec.