Within AI4Life, we ran three Open Calls and three Public Challenges (2023-2025), supporting 22 bioimage analysis projects from 151 applications and engaging 225 challenge participants, with the aim of applying FAIR deep learning in the life sciences. Our experience offers a view of the current state of bioimage analysis, the landscape of available tools, as well as the existing gaps between method developers, tool producers and potential users. It highlights that even after careful selection for AI-ready projects, most still require substantial effort to apply deep learning, and that the field still relies heavily on established, well-rounded methods to solve common problems. We come to the conclusion that for scientific AI in biology, the rate-limiting step is not methods and models but data, annotations, and shared infrastructure underneath them.

Abstract Cryogenic correlative light and electron microscopy (cryo-CLEM) enables visualization of biological specimens with molecular specificity while preserving near-native macromolecular structure. However, the severely limited resolution of conventional cryo-fluorescence microscopes restricts the accuracy of correlation with cryo-electron microscopy. Super-resolution cryogenic CLEM (SR-cryo-CLEM) offers a potential solution, but presents substantial technical challenges, including mechanical instability and ice contamination. Here, we introduce a modular cryogenic light microscope optimized for single-molecule localization microscopy (cryo-SMLM) that mitigates such limitations. The system is constructed primarily from off-the-shelf components, enabling straightforward and cost-effective assembly, and is operated using fully open-source Python software for flexible and customizable control. The mechanically and thermally stabilized architecture, combined with an axial focus-lock system, maintains sample positioning within a standard deviation of 40 nm. Ice contamination is minimized by imaging inside a purged enclosure, enabling prolonged acquisitions. Together, the platform provides robust localization precision, reproducible imaging performance, and an accessible solution for SR-cryo-CLEM.

Cnidarians are classically defined by a single oral opening, a hallmark of the "blind gut" model in early animal evolution. Here, we identify a pressure-sensitive aboral valve in Nematostella vectensis that operates independently of egestion. This valve dissipates elevated hydraulic pressure during morphogenesis, expelling fluid through transient epidermal ruptures triggered by muscular ring opening. This unexpected function was revealed by a comprehensive analysis of mesogleal basement membrane biogenesis. We show that the global dynamics of this extracellular matrix transduce muscular hydraulics to drive tissue rearrangement and stabilize shape, while localized FGFRb-dependent matrix remodeling establishes the aboral valve. By positioning the mesoglea as an integrator of biomechanics, tissue remodeling, and aboral valve function, these findings expand nonbilaterian openings beyond the digestive paradigm as a hydraulic regulator.

Cryo-volume electron microscopy (CVEM) enables three-dimensional imaging of biological ultrastructure in a near-native state but has been limited by low image contrast and charging artifacts that hinder data interpretation and complicate automation of data acquisition. Here we present an experimental and computational workflow that combines orthogonal cryo-SEM imaging, spot-geometry optimized O+ plasma-FIB milling, dedicated acquisition-control routines, and dedicated image alignment procedure. The workflow enables autonomous acquisition of volumetric datasets from vitrified cells and tissues at ~15-20 nm isotropic resolution. In addition, sub-volume averaging of 113 nuclear pore complexes extracted from CVEM dataset of Cos-7 cell yielded its reconstruction at 9.4 nm resolution. Together, these results establish CVEM as a robust platform for autonomous high-resolution volumetric imaging and structural analysis of vitrified biological specimens.

Coordinated collaboration is essential to realize the added value of and infrastructure requirements for global image data sharing in the life sciences. In this White Paper, we take a first step at presenting some of the most common use cases as well as critical/emerging use cases of (including the use of artificial intelligence for) biological and medical image data, which would benefit tremendously from better frameworks for sharing (including technical, resourcing, legal, and ethical aspects). In the second half of this paper, we paint an ideal world scenario for how global image data sharing could work and benefit all life sciences and beyond. As this is still a long way off, we conclude by suggesting several concrete measures directed toward our institutions, existing imaging communities and data initiatives, and national funders, as well as publishers. Our vision is that within the next ten years, most researchers in the world will be able to make their datasets openly available and use quality image data of interest to them for their research and benefit. This paper is published in parallel with a companion White Paper entitled Harmonizing the Generation and Pre-publication Stewardship of FAIR Image Data, which addresses challenges and opportunities related to producing well-documented and high-quality image data that is ready to be shared. The driving goal is to address remaining challenges and democratize access to everyday practices and tools for a spectrum of biomedical researchers, regardless of their expertise, access to resources, and geographical location.

During gastrulation, mouse epiblast cells form the three germ layers that establish the body plan and initiate organogenesis. While single-cell atlases have advanced our understanding of lineage diversification, spatial aspects of differentiation remain poorly defined. Here, we applied spatial transcriptomics to mouse embryos at embryonic (E) E7.25 and E7.5 days and integrated these data with existing E8.5 spatial and E6.5-E9.5 single-cell RNA-seq atlases. This resulted in a spatiotemporal atlas of over 150,000 cells with 82 refined cell-type annotations. The resource enables exploration of gene expression dynamics across anterior-posterior and dorsal-ventral axes, uncovering spatial logic guiding mesodermal fate decisions in the primitive streak. We also developed a computational pipeline to project additional single-cell datasets into this framework for comparative analysis. Freely accessible through an interactive web portal, this atlas offers a valuable tool for the developmental and stem cell biology communities to investigate mouse embryogenesis in a spatial and temporal context.

Segmentation of very large images is a common problem in microscopy, medical imaging or remote sensing. The problem is usually addressed by sliding window inference, which can theoretically lead to seamlessly stitched predictions. However, in practice many of the popular pipelines still suffer from tiling artifacts. We investigate the root cause of these issues and show that they stem from the normalization layers within the neural networks. We propose indicators to detect normalization issues and further explore the trade-offs between artifact-free and high-quality predictions, using three diverse microscopy datasets as examples. Finally, we propose to use BatchRenorm as the most suitable normalization strategy, which effectively removes tiling artifacts and enhances transfer performance, thereby improving the reusability of trained networks for new datasets.

Artificial Intelligence methods are powerful tools for biological image analysis and processing. High-quality annotated images are key to training and developing new methods, but access to such data is often hindered by the lack of standards for sharing datasets. We brought together community experts in a workshop to develop guidelines to improve the reuse of bioimages and annotations for AI applications. These include standards on data formats, metadata, data presentation and sharing, and incentives to generate new datasets. We are positive that the MIFA (Metadata, Incentives, Formats, and Accessibility) recommendations will accelerate the development of AI tools for bioimage analysis by facilitating access to high quality training data.

Differential gene expression establishes the distinct physiology and morphology of cell types in an animal body. Single-cell sequencing and volume EM represent milestones toward the characterization of cell types, yet are difficult to combine for a comprehensive view on the cellular genotype-phenotype link. Here, we map a whole-body single-cell transcriptome into the PlatyBrowser, a multimodal cellular atlas for the marine annelid Platynereis dumerilii , and establish this combination uniquely for an entire animal. We learn that, in the 6-days-old worm, the majority of genes are tightly co-regulated to jointly implement one of eight major cellular morphotypes representing epidermis, gut, vasculature, myofibres, glia, motile cilia, glands, or neurons. Focusing on neurons, we uncover 14 families that by transcription factor identity, axonal projection, or sensory-secretory apparatus resemble conserved neuron types found in vertebrates, insects, or nematodes. We hypothesize that these existed in urbilaterian ancestors and represent the ancient core of nervous system centralization. ### Competing Interest Statement The authors have declared no competing interest.

The characterization of phenotypes in cells or organisms from microscopy data largely depends on differences in the spatial distribution of image intensity. Multiple methods exist for quantifying the intensity distribution - or image texture - across objects in natural images. However, many of these texture extraction methods do not directly adapt to 3D microscopy data. Here, we present Spherical Texture extraction, which measures the variance in intensity per angular wavelength by calculating the Spherical Harmonics or Fourier power spectrum of a spherical or circular projection of the angular mean intensity of the object. This method provides a 20-value characterization that quantifies the scale of features in the spherical projection of the intensity distribution, giving a different signal if the intensity is, for example, clustered in parts of the volume or spread across the entire volume. We apply this method to different systems and demonstrate its ability to describe various biological problems through feature extraction. The Spherical Texture extraction characterizes biologically defined gene expression patterns in Drosophila melanogaster embryos, giving a quantitative read-out for pattern formation. Our method can also quantify morphological differences in Caenorhabditis elegans germline nuclei, which lack a predefined pattern. We show that the classification of germline nuclei using their Spherical Texture outperforms a convolutional neural net when training data is limited. Additionally, we use a similar pipeline on 2D cell migration data to extract the polarization direction and quantify the alignment of fluorescent markers to the migration direction. We implemented the Spherical Texture method as a plugin in ilastik to provide a parameter-free and data-agnostic application to any segmented 3D or 2D dataset. Additionally, this technique can also be applied through a Python package to provide extra feature extraction for any object classification pipeline or downstream analysis.

Identification of spot-like structures in large, noisy microscopy images is a crucial step for many life-science applications. Imaging-based spatial transcriptomics (iST), in particular, relies on the precise detection of millions of transcripts in low signal-to-noise images. Despite recent advances in computer vision, most of the currently used spot detection techniques are still based on classical signal processing and require tedious manual tuning per dataset. Here we introduce Spotiflow, a deep learning method for subpixel-accurate spot detection that formulates spot detection as a multiscale heatmap and stereographic flow regression problem. Spotiflow supports 2D and 3D images, generalizes across different imaging conditions and is more time and memory efficient than existing methods. We show the efficacy of Spotiflow by extensive quantitative experiments on diverse datasets and demonstrate that its increased accuracy leads to meaningful improvements in biological insights obtained from iST and live imaging experiments. Spotiflow is available as an easy-to-use Python library as well as a napari plugin at https://github.com/weigertlab/spotiflow .

Cells are essential to understanding health and disease, yet traditional models fall short of modeling and simulating their function and behavior. Advances in AI and omics offer groundbreaking opportunities to create an AI virtual cell (AIVC), a multi-scale, multi-modal large-neural-network-based model that can represent and simulate the behavior of molecules, cells, and tissues across diverse states. This Perspective provides a vision on their design and how collaborative efforts to build AIVCs will transform biological research by allowing high-fidelity simulations, accelerating discoveries, and guiding experimental studies, offering new opportunities for understanding cellular functions and fostering interdisciplinary collaborations in open science.

SUMMARYAt the onset of murine gastrulation, pluripotent epiblast cells migrate through the primitive streak, generating mesodermal and endodermal precursors, while the ectoderm arises from the remaining epiblast. Together, these germ layers establish the body plan, defining major body axes and initiating organogenesis. Although comprehensive single cell transcriptional atlases of dissociated mouse embryos across embryonic stages have provided valuable insights during gastrulation, the spatial context for cell differentiation and tissue patterning remain underexplored. In this study, we employed spatial transcriptomics to measure gene expression in mouse embryos at E6.5 and E7.5 and integrated these datasets with previously published E8.5 spatial transcriptomics1and a scRNA-seq2atlas spanning E6.5 to E9.5. This approach resulted in a comprehensive spatiotemporal atlas, comprising over 150,000 cells with 88 refined cell type annotations as well as genome-wide transcriptional imputation during mouse gastrulation and early organogenesis. The atlas facilitates exploration of gene expression dynamics along anterior-posterior and dorsal-ventral axes at cell type, tissue, and organismal scales, revealing insights into mesodermal fate decisions within the primitive streak. Moreover, we developed a bioinformatics pipeline to project additional scRNA-seq datasets into a spatiotemporal framework and demonstrate its utility by analysing cardiovascular models of gastrulation3. To maximise impact, the atlas is publicly accessible via a user-friendly web portal empowering the wider developmental and stem cell biology communities to explore mechanisms of early mouse development in a spatiotemporal context.

Validation metrics are key for the reliable tracking of scientific progress and for bridging the current chasm between artificial intelligence (AI) research and its translation into practice. However, increasing evidence shows that particularly in image analysis, metrics are often chosen inadequately in relation to the underlying research problem. This could be attributed to a lack of accessibility of metric-related knowledge: While taking into account the individual strengths, weaknesses, and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multi-stage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides the first reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Focusing on biomedical image analysis but with the potential of transfer to other fields, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. To facilitate comprehension, illustrations and specific examples accompany each pitfall. As a structured body of information accessible to researchers of all levels of expertise, this work enhances global comprehension of a key topic in image analysis validation.

We present a new set of computational tools that enable accurate and widely applicable 3D segmentation of nuclei in various 3D digital organs. We have developed an approach for ground truth generation and iterative training of 3D nuclear segmentation models, which we applied to popular CellPose, PlantSeg and StarDist algorithms. We provide two high-quality models trained on plant nuclei that enable 3D segmentation of nuclei in datasets obtained from fixed or live samples, acquired from different plant and animal tissues, and stained with various nuclear stains or fluorescent protein-based nuclear reporters. We also share a diverse high-quality training dataset of about 10,000 nuclei. Furthermore, we advanced the MorphoGraphX analysis and visualization software by, among other things, providing a method for linking 3D segmented nuclei to their surrounding cells in 3D digital organs. We found that the nuclear-to-cell volume ratio varies between different ovule tissues and during the development of a tissue. Finally, we extended the PlantSeg 3D segmentation pipeline with a proofreading tool that uses 3D segmented nuclei as seeds to correct cell segmentation errors in difficult-to-segment tissues.

The emergence of novel pathogens, exemplified recently by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), highlights the need for rapidly deployable and adaptable diagnostic assays to assess their impact on human health and guide public health responses in future pandemics. In this study, we developed an automated multiplex microscopy assay coupled with machine learning-based analysis for antibody detection. To achieve multiplexing and simultaneous detection of multiple viral antigens, we devised a barcoding strategy utilizing a panel of HeLa-based cell lines. Each cell line expressed a distinct viral antigen, along with a fluorescent protein exhibiting a unique subcellular localization pattern for cell classification. Our robust, cell segmentation and classification algorithm, combined with automated image acquisition, ensured compatibility with a high-throughput approach. As a proof of concept, we successfully applied this approach for quantitation of immunoreactivity against different variants of SARS-CoV-2 spike and nucleocapsid proteins in sera of patients or vaccinees, as well as for the study of selective reactivity of monoclonal antibodies. Importantly, our system can be rapidly adapted to accommodate other SARS-CoV-2 variants as well as any antigen of a newly emerging pathogen, thereby representing an important resource in the context of pandemic preparedness.

Intravital 2P-microscopy enables the longitudinal study of brain tumor biology in superficial mouse cortex layers. Intravital microscopy of the white matter, an important route of glioblastoma invasion and recurrence, has not been feasible, due to low signal-to-noise ratios and insufficient spatiotemporal resolution. Here, we present an intravital microscopy and artificial intelligence-based analysis workflow (Deep3P) that enables longitudinal deep imaging of glioblastoma up to a depth of 1.2mm. We find that perivascular invasion is the preferred invasion route into the corpus callosum and uncover two vascular mechanisms of glioblastoma migration in the white matter. Furthermore, we observe morphological changes after white matter infiltration, a potential basis of an imaging biomarker during early glioblastoma colonization. Taken together, Deep3P allows for a non-invasive intravital investigation of brain tumor biology and its tumor microenvironment at subcortical depths explored, opening up opportunities for studying the neuroscience of brain tumors and other model systems.