Laura Waller, University of California, Berkley

Computational imaging jointly designs hardware and algorithms to push beyond the classical limits of imaging, enabling measurement of new quantities (e.g. 3D, phase, and super-resolution) with simple, inexpensive hardware. In this talk, I show recent advances that push both spatial and temporal throughput for 2D and 3D fluorescence microscopy. First, I will describe a diffractive, multiplexed microscope that uses engineered point spread functions and a multi-sensor array to achieve gigapixel-scale imaging at video rates, enabling micrometer-resolution imaging over centimeter-scale fields of view for dynamic biological systems . Second, I will introduce a neural space–time model that jointly reconstructs images and motion from sequential measurements, eliminating motion artifacts while recovering sample dynamics without training data or priors. Together, these approaches illustrate a shift from static imaging toward high-throughput, dynamic measurement, opening new opportunities for observing complex biological processes across scales.

Sara McArdle, La Jolla Institute for Immunology

With the rise of highly multiplexed immunofluorescence, cell classification is becoming increasingly challenging. While QuPath provides great tools for training an object classifier, new strategies are needed for more complex phenotyping. This workshop will demonstrate some recommended techniques for organizing training data, managing file sizes for faster processing, and data and image export.

John Bogovic, HHMI Janelia

This workshop will demonstrate Fiji tools to read and write chunked file formats (HDF5, N5, and OME-Zarr). We will learn to import and export images with ImageJ’s user-interface, and how to visualize large with BigDataViewer. Beyond the basics, participants will learn common jargon (“chunk”, “block”, “shard”), best practices for chunk size and compression, how metadata standards enable interoperability across tools, and how to integrate these formats into imaging workflows.

VIRTUAL Workshop

Sean Du’Hare, Allen Institute for Cell Science

This workshop introduces participants to bioimage metadata – where it lives, how to access it, and why it matters for analysis and reproducibility. Using BioFile Finder (BFF), attendees will explore real datasets, uncover embedded metadata (e.g., imaging conditions, channels, and experimental context), and learn how to query and filter it effectively. The session emphasizes practical workflows for navigating large, heterogeneous imaging collections without needing to download files. By the end, participants will understand how to leverage metadata to drive discovery, validation, and downstream analysis.

Ed Evans, University of Wisconsin-Madison

Cellcast is a new project that aims to modernize and redistribute (i.e. recast) established cell segmentation models built on the Burn deep learning framework with a WebGPU backend. Models that are ported into cellcast do not depend on libraries like TensorFlow and PyTorch, allowing cellcast to deliver inference and training with just a NumPy dependency (for Python). In this workshop we will cover the architecture of cellcast and how to implement it in your existing Python workflows and Fiji. In the last part of the workshop we will cover how to compile cellcast locally with different backends (e.g. CUDA) and how to use cellcast directly with other languages like Java.

Caleb Hulbert, HHMI Janelia Research Campus

Paintera is an open-source tool for 3D annotation and proofreading of large volumetric image data. This workshop will demonstrate a segmentation workflow in Paintera, focusing on visualizing and annotating arbitrary rotations on image data using tools such as Shape Interpolation with Segment Anything and morphological operations on 3D labels. We will go over importing existing label source using N5 and ImgLib2, and how to use Paintera labels in other annotation and visualization tools.

VIRTUAL Workshop

Anwai Archit, Georg-August-Universität Göttingen

Our work builds on the Segment Anything Model (SAM), a deep learning model for universal image segmentation developed by Meta AI (formerly Facebook). SAM can segment any kind of object in images based on user input, using a large deep neural network that processes the image and points or boxes indicating the object of interest. While SAM can be applied to microscopy, it does not yield good results for difficult tasks, for example cells with complex shapes or complex organelles such as golgi or ER. Our work addresses this problem, by further training it on microscopy data. We build a software tool based on napari for interactive and automatic segmentation around our offered models. We have exciting updates to share (coming soon) about our software.

Andrew McCall, University at Buffalo

In this workshop we’ll be discussing and using a novel colocalization/spatial correlation algorithm that does not require overlapping fluorescent signal, making it compatible with super-resolution microscopy techniques.

VIRTUAL Workshop

Eugene Katrukha, Utrecht University

In this workshop, you will learn how to explore and visualize in 3D multiple (and single) large (and small) volumetric microscopy datasets using the open-source software Fiji and the BigVolumeBrowser plugin. We will combine it with simultaneous visualization of meshes or point clouds (transcriptomics, SMLM data) and creation of high-quality animations/renders. We will be using walkthroughs and examples from the plugin’s documentation.

Bruno Pannuzio, Institut Pasteur de Montevideo

Fluorescence Lifetime Imaging (FLIM) and Hyperspectral Imaging are essential for resolving complex biological environments, yet extracting quantitative insights from these multidimensional datasets is often hindered by a fragmented software landscape. Despite the power of the phasor approach, there is a notable lack of open-source, interactive, and user-friendly tools that combine phasor analysis for both FLIM and hyperspectral data within a single interface. To address this, we present napari-phasors, a comprehensive plugin built upon the PhasorPy library. By integrating these capabilities into the napari multidimensional viewer, users can leverage the vast ecosystem of existing napari tools, such as advanced segmentation masks and community-driven plugins, creating a synergistic environment for sophisticated bioimage analysis.

The plugin provides a streamlined workflow starting from the direct import of diverse raw file formats such as .ptu, .sdt, and .lsm. We demonstrate its ability to perform simultaneous multi-layer analysis, enabling the management of complex experimental conditions in a single session. Beyond multi-harmonic calibration, filtering and pixel-wise mapping of lifetime, phase, and modulation, napari-phasors features a robust statistical framework. The results of the analysis distribution can be interactively visualized through integrated histograms and statistics tables, creating a complete workflow that spans from raw data to quantitative visualization plots. With specialized features like FRET and component analysis, napari-phasors provides a high-performance, accessible solution for the broader scientific community.

Caroline Malin-Mayor, HHMI Janelia Research Campus

The Motile Tracker is a GUI-based program for interactively generating, visualizing, and editing segmentation and tracking data. The workshop will go over installation, importing and exporting data, editing functionality, and how to use and configure incorporated machine learning methods for tracking. As the Motile Tracker is a community effort, we will also be seeking feedback from users on usability and desired features, and potential code contributions.

This workshop will be co-led with Anniek Stokkermans from Hubrecht Institute and Teun Heuijben from the Biohub, pending their ability to attend in-person.

Kevin Dean, UT Southwestern Medical Center

In this hands-on workshop, participants will install and launch navigate, explore the core interface in synthetic hardware mode, and build practical image-acquisition workflows without needing a live microscope. We will start with the basics of configuration and acquisition, then move into simulated examples that show how navigate supports smart microscopy, including selective imaging, multiposition workflows, and multiscale acquisition strategies. Attendees will leave with a practical starting point for using navigate on their own systems and a clear sense of how the platform can scale from basic image acquisition to more advanced automated workflows.

Esteban Miglietta, Broad Institute of MIT & Harvard

During this workshops, participants will first hear a ~30-40 intro to image analysis in general and to CellProfiler in particular. Then they will follow a written tutorial to build a simple segmentation + measurement pipeline for analysing sample data provided. This hands-on part typically takes ~60-90min, during this time, they will be assisted by me and (depending on availability and number of participants) some TA.

Eduardo Brito-Alarcon, UNAM

Achieving trustworthy super-resolution imaging requires navigating a complex landscape of parameters, sample preparation, and algorithmic processing. However, many critical hurdles, including logistical constraints and workflow inconsistencies, are rarely discussed in formal literature. We present a comprehensive overview of these pitfalls and offer actionable solutions to improve reproducibility and image integrity. This session concludes by introducing ‘The Imaging Orchestrator,’ a tool designed to automate and standardize these processes, which will be explored in detail during our accompanying workshop.

Thierry Pécot, Rennes University

While the emergence of robust 3D interactive segmentation approaches has enabled to quickly create small high-quality 3D training datasets, microscopy image analysis still lacks a generalist 3D deep learning model capable of robust segmentation across diverse biological structures. We propose to take advantage of the increasing number of publicly available 3D microscopy datasets to train a semantic segmentation model to robustly segment object interiors and boundaries. Our approach integrates a post-processing pipeline based on the distance transform watershed algorithm to identify individual objects. We conduct a comprehensive benchmark against state-of-the-art pretrained models (Cellpose, StarDist, and EmbedSeg) on two challenging tasks: nuclei and whole-cell segmentation. Beyond standard comparisons, we evaluate all models under low-data regimes, systematically assessing performance when fine-tuned with varying amounts of training data (from 0% to 100% of available annotations).

Juno Chacko, University of Wisconsin-Madison

Modern microscopy produces increasingly complex, high-dimensional datasets, yet acquisition workflows remain largely static, manually driven, and rarely informed by the data they generate in real time. This is in part because meaningful analysis during acquisition has traditionally demanded coding expertise and effort that cannot be sustained in real time, a barrier that AI agents can solve in parallel. Here, we present Data-Reactive Acquisition and Microscope Steering for Agentic Microscopy (DREAMS). DREAMS is built around a dedicated MCP server for real-time microscope control, embedded within a continuous feedback loop of acquisition, analysis, and decision-making. It serves as the core MCP server for microscope control, with auxiliary servers extending its capabilities such as structured note-keeping, data management, enabling context-aware workflows. Integration with reactive notebook environments such as Marimo and other frontend interfaces is also demonstrated, providing flexible entry points for both interactive exploration and automated pipelines. Using fluorescence lifetime imaging as an example workflow, we show how AI-driven decision-making improves acquisition efficiency and data quality. Similarly, applications in collagen fiber analysis illustrate integration with existing tools and extensible, adaptive imaging pipelines. This modular, protocol-driven architecture is designed for open science, enabling community extensibility across diverse imaging systems. Compatibility with agent-to-agent protocols and agent-orchestration frameworks further suggests a path toward federated imaging networks, where multiple specialized agents collaborate in real time to drive complex, multi-modal experiments without human intervention.

VIRTUAL presentation

Mary Brown, University of Minnesota – Twin Cities

Use small images with ground-truth annotations to create label images of them, create augmentations of the label images by rotating them with FIJI software, segment test images with the Cellpose-SAM plugin for FIJI, and use the Mask Instant Comparator plugin to assess accuracy of the Cellpose-SAM segmentations against the ground-truth annotations.

Sara McArdle, La Jolla Institute for Immunology

Type 1 diabetes (T1D) is a progressive autoimmune condition that culminates in loss of insulin-producing beta cells. Here, we combined cyclical multiplexed immunostaining, whole-slide imaging, digital pathology, and semi-automated image analyses to interrogate pancreatic tail and head sections through T1D stages, including at-risk and at-onset cases. The dataset, comprising 9 markers across ~25,000 islets, is among the largest collections of human pancreatic histopathology images. Quantifying architectural features, endocrine cell composition, immune cell burden, and spatial relations between the islets reveals histopathological correlates that foreshadow distinctive T1D histopathology already at the preclinical stage. All image analysis work was performed in the open-source platform QuPath and the raw images, calculated metrics, and processing scripts are publicly available.

Mark Tsuchida, University of Wisconsin-Madison

Micro-Manager is a widely used open-source platform for controlling microscopes and automating image acquisition across hundreds of devices and vendors. This workshop introduces Micro-Manager from the ground up, starting with the basics of how computer control of microscope hardware works, then moving through hardware configuration and interactive use of the GUI (including Multi-Dimensional Acquisition), along with a sampling of advanced features. In the second half, we turn to Python and use pymmcore-plus to drive devices and run MDAs programmatically, closing with a worked example of data-driven (“smart”) acquisition in which on-the-fly image analysis steers subsequent acquisition. By the end, participants should be comfortable operating Micro-Manager for routine experiments and equipped to start writing their own Python-based acquisition scripts.

Brain Northan, True North Intelligent Algorithms

In this workshop, we will discuss strategies for setting up Jupyter notebooks and napari plugins for deep learning–based restoration and segmentation. The end goal is to make it easier to compare different segmentation and/or restoration approaches (e.g., Cellpose, Stardist, and classical image processing) within the same workflow (e.g., a single script, notebook, or napari plugin). Traditionally, this has been complicated by the fact that different deep learning libraries often require incompatible sets of dependencies. We will show how technologies such as appose, pixi, cellcast, and others can help address this challenge. We will also introduce a project called Napari-AI-Lab, which uses these technologies to provide both an API and a set of interactive widgets for multi-framework deep learning comparisons.

Mark Hiner-Weir, University of Wisconsin-Madison

The Fiji Chatbot is designed to provide users with an interface for using Large Language Models (LLMs) in a context aware of your actual installation, with an initial focus on common image analysis and scripting tasks. In this workshop we will discuss expectations and best practices for LLM use in image analysis, work through carefully guided tasks together with the chatbot, discuss your particular needs and use cases, and see how they align with the capabilities of the chatbot at that time. Bring your questions, your data (if you wish), and consider: how do you want this tool to work for you?

Mike Nelson, University of Wisconsin-Madison

QuPath is widely used for analysis of large 2D images, and its extension mechanism enables capabilities far beyond post-acquisition analysis. At LOCI, we have developed a suite of open-source QuPath extensions that add microscope hardware control and automated acquisition (QPSC), publication-quality image export with integrated QUAREP-LiMi guidance (QuIET), retrainable deep learning pixel classification, slide label OCR for automated metadata extraction, enhanced annotation tools, classifier validation with bootstrap statistics, and specialized imaging analysis including polychromatic polarization microscopy. In this workshop we will demonstrate these extensions through a mix of live microscope demos, interactive exercises, and hands-on exploration, showing how QuPath can serve as a unified environment from acquisition through publication. The first hour will be dedicated to introducing the tools, with the second hour being optional exploration of the extensions with provided or user-provided data sets.

Tiago Ferreira, HHMI – Janelia Research Campus

SNT is a complete, open-source framework for neuron tracing, analysis, and visualization of neuronal reconstructions, powered by an ever-growing user and developer community. Recent development has addressed the growing scale and diversity of neuroimaging data: native support for multi-terabyte datasets, automated and semi-automated tracing routines, real-time quality monitoring during reconstruction, expanded morphological analysis, and tighter integration with community resources and public data repositories. We demonstrate how these capabilities combine into practical workflows and discuss how SNT’s scriptable architecture allows users to adapt these workflows to their own imaging modalities and biological questions. No neuroscience background is assumed.

Curtis Rueden, University of Wisconsin-Madison

Innovation in biological image analysis increasingly happens in Python (Cellpose, StarDist, SAM, Trackastra), yet biologists rely on established platforms like Fiji/ImageJ for their day-to-day workflows. Fiji now offers multiple complementary ways to work with Python: A) Python mode runs real CPython in-process alongside Java (powered by PyImageJ and Jaunch); B) the new Appose-based Python scripting language lets you write `.py` scripts directly in Fiji’s Script Editor backed by isolated CPython environments; and C) using Appose a library from scripts enables multiple simultaneous Python environments with zero-copy shared memory for passing image data—allowing tools with incompatible dependencies to coexist. This workshop surveys the current Python landscape in Fiji, showcases real-world integrations (SAMJ, TrackMate, Mastodon), and concludes with a hands-on exercise in which participants use Fiji’s Script Editor to build their own Python-powered tool from scratch.

Ben Schmid, FAU Competence Centre Optical Imaging Centre Erlangen

Efficient control of experimental and image-analysis pipelines often requires scripting, which remains inaccessible to many non-programmers. Here, we introduce nlScript, a comprehensive, developer-oriented toolbox for implementing domain-specific natural language interfaces combining the determinism of formal languages with the intuitiveness of English syntax. nlScript replaces complex graphical user interfaces with a unified natural language interface. It offers developers a concise way to define grammar rules, provides a built-in editor for user input that infers auto-completion rules automatically from the defined grammar, handles input parsing and executes the resulting instructions. On top, nlScript supports developers in creating a large language model for free user input. nlScript can automatically generate valid input script samples using developer-defined grammar rules. Rephrasing these samples with an existing general-purpose LLM results in a dataset consisting of mappings from free-form sentences to valid input scripts, readily usable to fine-tune a highly specialized large language model for the application by hand. The entire process for creating the LLM, including training data generation and fine-tuning, can thereby be completed within 3-4 hours. Models created in this way translate free user input into a syntactically valid natural language script. Full control remains in the users’ hands, who can validate the output of the model before executing the script. We demonstrate nlScript’s value in several applications, including a natural-language based software to configure complex acquisition pipelines on modern microscopes.

Esteban Miglietta, Broad Institute of MIT & Harvard

This workshop is an extension of the “Intro to pipeline building in CellProfiler” workshop, as it requires basic familiarity with Cellprofiler. This workshop can be fused to the previous one in one block (ideally with some break in the middle) if needed.

In this workshop, we will briefly explain the use of the Ilastik and Cellpose sofwares (what they can do, what they are useful for, how to use them) and how to incorporate their capabilities into a CellProfiler pipeline via the use of plugins and software containers.

Like the previous workshop, this one will consist of an introductory lecture, followed by a hands-on exercise following a written tutorial.

VIRTUAL presentation

Christopher Schmied, Leibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP Berlin)

Microscopy figures are the primary vehicle for communicating imaging results, yet many published figures contain avoidable issues: colorblind-inaccessible color combinations, missing scale bars, undocumented brightness/contrast adjustments, and lossy compression that degrades quantitative data. This workshop walks participants through the QUAREP-LiMi WG12 Checklists for Image Publishing using a practical, hands-on workflow from raw multichannel microscopy data to a complete, publication-ready figure. Using Fiji and Inkscape, attendees will process channels with accessible color assignments, add calibrated annotations, assemble multi-panel figures in vector format, and document their methods for reproducibility. All materials — including sample data, templates, and the interactive checklists — are freely available for reuse in participants’ own teaching and research

Esteban Miglietta, Broad Institute of MIT & Harvard

During this workshop, participants will first receive a lecture (~45min) on the basics of colocalization analysis and then engage in group discussions by playing the ColoC game . (https://biop.github.io/coLoc/). We will discuss the uses and limitations of colocalization analysis, explore some of the most commonly used metrics and talk about proper acquisition, all focusing on the biological question at hand. During the game section, participants will be given a sample biological question and experimental setup and they will have to design the proper acquisition and analysis strategy to successfully answer the biological question.

Alexandre Cunha, California Institute of Technology

The Cell Tracking Challenge encompasses three interconnected valuable activities in the contemporary biological image analysis landscape. The Challenge provides curated and annotated 2D and 3D time lapse videos from many different microscopy modalities, all freely available to the public. This is part of our ongoing effort to continue providing quality data and annotations

for an ever-growing demand from supervised machine learning methods. Using such annotated datasets, we administer three correlated benchmarks, namely cell tracking, cell segmentation, and most recently cell linking, which allow us to continuously compare and rank methods submitted thus far by sixty groups from all continents but Africa, and published on web accessible leaderboards. Lastly, we strive to serve as a catalyst for engaging researchers, developers, end users, and the bioimage analysis community at large so together we contribute

rich data and FAIR methods towards our end goal of serving science to benefit humankind. This is what motivate us to present our thirteen years of

accomplishments and roadmap to attendees of the workshop where we hope hearing directly from the community how we could better serve the unmet needs within the scope of our effort and make a call for participation and contributions.

Tim Monko, napari

In this hands-on workshop you will explore napari, a powerful open-source multi-dimensional image viewer built for scientific data analysis in Python. Whether you are a biologist, microscopist, or image analyst, this workshop will guide you through the fundamentals of visualizing and analyzing complex image data with napari’s visual interface and rich plugin ecosystem. Participants will gain practical experience loading and exploring multi-dimensional datasets, customizing visualizations, and leveraging napari’s growing library of community-developed plugins. Previous workshop materials can be viewed at https://napari.org/workshops/

VIRTUAL presentation

Timon Stegmaier, Institute of Functional Epigenetics, Helmholtz Munich

In this workshop, I will showcase Cell-ACDC, an all-in-one GUI for analysing microscopy data of almost any dimensionality. As a demonstration, I will analyse a 3D z-stack image of Chlamydomonas algae. I also encourage participants to bring their own data to try it out themselves and get quick feedback and answers.

VIRTUAL presentation

Suganya Sivagurunatha, Uppsala University

The workshop will touch on the basics of image registration, register data from metabolomics/spatial transcriptomics and light microscopy and visualize the integrated data using TissUUmaps 4.

Participants will learn basics of image registration and the know-how of visualization, integration and interactive exploration of the multi-modal data. The workshop will introduce TissUUmaps, a free and open source browser-based tool for fast visualization and exploration of spatial data.

Song Hu, Washington University in St. Louis

Brain energy metabolism is sustained by dense microvascular networks that dynamically deliver oxygen to active neural tissue. Photoacoustic microscopy (PAM), based on the optical absorption contrast of hemoglobin, enables label-free imaging of cerebrovascular structure, blood flow, oxygenation, and oxygen metabolism. However, conventional PAM remains limited in resolving 3D microvascular function with isotropic precision and in linking hemodynamic measurements to cellular activity and tissue metabolism. In this talk, Dr. Hu will first introduce super-resolution functional PAM, which uses label-free red blood cell tracking to reconstruct 3D microvascular structure, flow, and oxygenation at cellular resolution. He will then present their new work on the integration of PAM with two-photon microscopy for single-cell neurometabolic imaging, enabling simultaneous measurements of single-neuron activity and single red-blood-cell oxygen release in the awake mouse brain. Finally, he will discuss their ongoing work on deep learning-powered super-resolution PAM, in which co-registered TPM-PAM vascular datasets are used to train a physics-informed neural network that overcomes the acoustic bandwidth-limited axial resolution of PAM. Together, these developments highlight the synergy of light- and sound-based microscopy as a quantitative platform for studying neurovascular function and brain energy metabolism in vivo.

NOTE: this session will not be recorded and will only be available live

Kevin Dean, Lyda Hill Department of Bioinformatics, UT Southwestern Medical Center

Understanding how molecular state is translated into cellular architecture requires imaging approaches that can preserve three-dimensional context, resolve nanoscale structure, and scale to quantitative analysis across many cells and tissues. Here, I will discuss two complementary strategies from our group that connect highly multiplexed fluorescence imaging with computational analysis of cellular and subcellular organization. Here, I will describe Cyclically Multiplexed Expansion Microscopy, or Cy-ExM, which combines cryo-preservation, expansion microscopy, iterative immunolabeling, volumetric oblique plane microscopy, and image registration to map many molecular targets within the same expanded cell. This workflow enables high-fidelity, three-dimensional visualization of subcellular architecture across repeated labeling cycles, supporting nanoscale molecular atlases of organelles, cytoskeletal systems, nuclear structures, and membrane-associated compartments. I will also show how expansion and light-sheet imaging workflows can be extended from single-cell ultrastructure to intact tissue. By integrating iterative expansion microscopy, oblique plane microscopy, axially swept light-sheet microscopy, three-dimensional segmentation, surface analysis, and mitochondrial morphometry, we quantified how oncogenic drivers remodel liver architecture across tissue, cellular, membrane, and organelle scales. In mosaic models of hepatocellular carcinoma initiation, NRAS activation primarily altered cell shape and membrane curvature, whereas CTNNB1 activation produced mitochondrial remodeling toward a pericentral-like structural state. Integrating these features improved discrimination of oncogenic state, demonstrating that molecular perturbations can be encoded as distributed, multiscale architectural phenotypes.

Sreenivas Bhattiprolu, Carl Zeiss Microscopy, LLC

Despite rapid advances in imaging technology, a persistent gap exists between data acquisition and meaningful analysis. Through field visits to over 40 institutions, community workshops, and years of creating educational content, a consistent pattern has emerged: researchers lack foundational training in image analysis, rely on manual methods, and underestimate the reproducibility costs of doing so.

This talk traces that observation across a decade, from early efforts to bring computational literacy to biology, through the rise of deep learning and foundation models, to where the field is heading next. As workflows evolve from 2D to 3D, biological complexity increases but so does the analytical challenge. Meeting that challenge requires not just better tools, but researchers who understand how to use them.

Part 2 of workshop

Tim Monko, napari

In this hands-on workshop you will explore napari, a powerful open-source multi-dimensional image viewer built for scientific data analysis in Python. Whether you are a biologist, microscopist, or image analyst, this workshop will guide you through the fundamentals of visualizing and analyzing complex image data with napari’s visual interface and rich plugin ecosystem. Participants will gain practical experience loading and exploring multi-dimensional datasets, customizing visualizations, and leveraging napari’s growing library of community-developed plugins. Previous workshop materials can be viewed at https://napari.org/workshops/

VIRTUAL presentation

Joran Deschamps, Human Technopole

CAREamics is a deep-learning image restoration library that allows running staple algorithms such as Noise2Void for image denoising. The new version of CAREamics now include advanced features, such as background patch filtering or compatibility with large datasets (Zarr), but also more recent advances with the example of MicroSplit, an channel unmixing method. In this workshop, we will give an overview of CAREamics, and walk participants through various ways of running its algorithms in their pipelines, including python scripts and the CAREamics UI.

Anita Mahadevan-Jansen

Wenxuan Zhao, Morgridge Institute for Research

This workshop is a step-by-step demonstration of FLIM Playground, walking participants through every section and module of the GUI. In the Data Extraction section, we will go through Data Extraction Configuration, metadata organization, calibration, lifetime extractors (fitting, phasor), intensity-based extractors (morphology, texture), and categorical feature extraction to produce a single-cell feature table from raw FLIM data. In the Data Analysis section, we will then demonstrate each analysis module in turn — Feature Comparison, Feature Histogram, Field of View Comparison, Feature Distribution, Phasor Analysis, Dimension Reduction, and Classification — using the extracted dataset. We will end by exporting the analysis as a standalone Python script and publication-ready SVG figures, so attendees leave with a fully reproducible workflow.

VIRTUAL presentation

Francesco Padovani, Institute of Functional Epigenetics, Helmholtz Munich

In this workshop we will see how to easily leverage state-of-the-art segmentation and tracking models to annotate multi-dimensional microscopy data. We will then leverage the graphical user interface of Cell-ACDC to correct and annotate single cells. Finally, we will see how to compute several numerical measurements from an arbitrary number of fluorescence channels.

Toni Tang, University of Virginia

The function of biomineralizing tissues such as bone, cartilage, and tendon depends not on composition alone but on how mineral, collagen, water, and cells are arranged in three dimensions (3D) across length scales. Many of the features that matter most are nanometer in scale and sit within the mineralized matrix, beyond the reach of common imaging methods. Laboratory micro-CT resolves large volumes but not nanoscale structure, while conventional electron microscopy resolves ultrastructure but only in two dimensions (2D) or in very small volumes. As a result, much of the architecture has gone unidentified.

This talk centers on a correlative, multiscale imaging approach for resolving these features. I will first describe a correlative workflow that links the mesoscale to the nanoscale. Large-volume imaging locates deeply buried features within their whole-tissue context, and site-specific milling then isolates those same regions for high-resolution 3D imaging. I will then turn to what this workflow reveals at the smallest scales, focusing on an extensive nanochannel network in fully mineralized bone that is roughly an order of magnitude finer than osteocyte canaliculi yet occupies a far larger volume fraction. Resolving and segmenting these dense 3D networks uncovered an inverse relationship between nanochannel volume and local calcium content, pointing to a role in ion and small-molecule transport. Throughout, I will discuss the practical data challenges of acquiring, reconstructing, and segmenting nanoscale networks within mineralized tissue, and how multiscale 3D imaging is turning previously inaccessible architecture into quantitative, testable links between nanostructure and tissue function.

Sabine Mai

VIRTUAL presentation

Mariana De Niz, Northwestern University

Intravital microscopy (IVM) combines the surgical implantation of windows with advanced microscopy methods for in situ imaging of complex and dynamic biological processes. However, IVM generates large and complex multi-dimensional datasets which are difficult manually or even through simple automation. In our work, we have explored and optimized AI-based methods for image pre-processing, analysis, and predictive modeling. These tools have allowed us to investigate cell migration and vascular blood flow in situ, which we hope to use to better understand vascular pathology in the context of blood flow disturbances.

Michelle Digman

Juan Caicedo

Erik Jorgensen

Kedar Narayan, Frederick National Laboratory, National Cancer Institute, NIH

Electron microscopy (EM) has long been the gold standard for high resolution imaging of cellular ultrastructure. Recent advances have extended EM to the third dimension, allowing “volume EM” to emerge as an exciting avenue to explore previously inaccessible questions in biology, especially when combined with correlative methods. In spite of these developments, volume EM has remained mostly descriptive, as segmentation of cellular features from ultrastructural image volumes is still – unlike for much of optical microscopy – a major hurdle. Deep Learning based approaches show promise when applied to 2D and 3D EM data, but these solutions must work in resource- and expertise-limited settings. Here I discuss empanada, our napari plugin for point-and-click segmentation and proofreading of large numbers of organelles such as mitochondria and nuclei from 2D and 3D cellular EM images. Widening the segmentation bottleneck enables quantitative and statistically robust ultrastructural studies in cell biology, opening the door to new discoveries.

Kirk Czymmek, Donald Danforth Plant Science Center

Volume electron microscopy and multiplex correlative imaging can reveal structure-function relationships and the subcellular distribution of targeted molecules. Volume electron microscopy and multiplex correlative imaging can reveal structure-function relationships and the subcellular distribution of targeted molecules in plants, but plant tissues are notoriously hard to image: waxy cuticles, cell walls, and air spaces impede fixation, introduce optical aberrations, and yield poorly conductive samples. Optimizing multiscale and correlative microscopy offers a powerful way to identify, relocate, and image target structures while preserving their context within bulk tissue.

We have developed plant specific cryo-vEM workflows and heavy-metal staining resin-embedded protocols to address key questions in plant research. These preparations suited correlative workflows pairing x-ray microscopy (XRM) with serial block-face SEM (SBF-SEM), focused ion beam SEM (FIB-SEM) with cryo-fluorescence and cryo-electron tomography: We also adapted multiplex microscopy—well established in mammalian tissue but rarely used in plants—to map biomolecular probes across heterogeneous tissue, localizing nucleic acids via hybridization chain reaction (HCR) alongside proteins, carbohydrates, and ultrastructure in the same sample. Robust, broadly shared protocols of this kind will benefit the plant microscopy community and adjacent disciplines alike.

John Bogovic, HHMI Janelia

Complementary information from multiple imaging modalities has yielded great insights in bioimaging, for example in cell biology and connectomics. Unique challenges arise during the analysis of such multi-modal and multi-scale imaging datasets, but new storage formats and metadata standards have made data scalable and shareable, and software interoperable. We developed the N5 API and libraries to bring chunked file formats and OME-Zarr to open-source software in the ImageJ/Fiji ecosystem. Now, software such as Imglib2, BigDataViewer, BigWarp, Paintera, BigStitcher, and others can efficiently read and write arbitrarily large datasets from local storage or cloud storage. Furthermore, the N5 Fiji plugins enable researchers to use these formats from a familiar user interface. This talk will demonstrate practical applications of these tools in service of Janelia’s unique datasets.