The problem — Processed light-sheet, CLEM, and spatial-omics datasets routinely exceed single-machine RAM. Sharing them reproducibly across labs — or between acquisition and analysis teams — requires either a running web server or a file format that handles multi-scale pyramids and remote access natively. Neither condition is routinely met in biology labs.
What it is / how it works — MoBIE (Multimodal Big Image Explorer) is an open-source Fiji plugin built on BigDataViewer and ImgLib2. It reads chunked multiscale image data (OME-Zarr, N5, or HDF5) directly from local disk or cloud storage (S3, Google Cloud) and renders arbitrary pyramid levels interactively without loading the full volume. A JSON project specification encodes views — predefined combinations of image layers, segmentation overlays, table annotations, and display settings — that can be shared as a URL or file, making a complex data exploration fully reproducible. Paired tables (e.g. morphology measurements per segmented cell) are loaded alongside images, enabling click-to-select workflows where a cell in the image highlights its row in the table.
Where it fits — MoBIE is an exploration and QC tool at the boundary of the preprocess and segment stages in light-sheet and volumetric workflows. It is the inspection layer: after ingest and stitching produce a multiscale OME-Zarr store, MoBIE is how an analyst verifies alignment seams, reviews segmentation quality on representative z-planes, and communicates findings with collaborators — all without moving the data. It also serves developmental biology and CLEM datasets where correlated modalities must be overlaid precisely.
Where it breaks — MoBIE is a viewer; it does not run batch processing or export analysis. Very large segmentations or dense annotation tables can slow interactive rendering. Cloud access is only as fast as bucket permissions and network latency allow — poorly configured S3 ACLs are a common setup failure.