Light-Sheet 3D/4D: a pipeline architecture for trustworthy imaging.
01 · Pipeline Engineering
Turn an experimental method into an executable system.
Microscopy analysis often begins as a working method understood by one scientist: instrument settings, processing decisions, and scripts developed around a particular experiment. Fovea Lab provides a structured environment for turning that method into a pipeline that can be run, examined, and developed over time.
A scientist can begin in the visual workspace, ask the agent to help develop a stage, or open the underlying code when direct control is needed. Each way of working modifies the same pipeline rather than creating a separate version of the analysis. The scientific purpose of the experiment remains connected to the computation, allowing the workflow to mature from early exploration into a system that another researcher can understand and use.
02 · Evidence Integrity
Make confidence part of the computation.
A final measurement does not reveal whether the source data were complete, whether a model behaved unexpectedly, or whether an important parameter changed between runs. Fovea Lab evaluates the analysis as it runs and preserves the evidence needed to understand the result afterward.
Checks are placed at the stages where failures can occur. Scientists can return from a result to the pipeline that produced it, inspect whether the run remained within accepted conditions, and trace unexpected behavior to the point where it entered the analysis. Verification becomes part of the workflow itself rather than a report reconstructed after the experiment is finished.
03 · Agentic Development
Let AI improve the workflow without taking control away from the scientist.
Fovea Lab gives agents access to the structured experimental context that governs the pipeline. Instead of producing isolated code without understanding the surrounding system, an agent can inspect a specific stage, propose a bounded revision, and evaluate its effect against criteria established by the scientist.
A proposed change remains separate from the approved workflow until it has been reviewed. Its effect can be compared with the current version, while the decision to adopt it remains with an authorized user. AI can accelerate pipeline development while the platform preserves accountability and a stable record of what was actually run.
Pipeline architectures, module by module.
Pipeline backbone — reshapes per modality. Select a module for its sub-steps, tools, and deliverable.
Light-Sheet 3D/4D
High confidenceIngest
Ingest terabyte-scale light-sheet volumes, multi-view/tile/angle acquisitions and 4D time series into chunked, streamable, multiscale data structures. The goal is to preserve acquisition geometry (view angles, tile transforms, voxel calibration, time-lapse index) intact while enabling distributed registration, deconvolution and downstream analysis without loading whole volumes into RAM.
Operations
- Dataset discoveryIndex views, angles, tiles, channels, z-planes and timepoints from raw microscope exports (CZI, IMS, HDF5/XML, TIFF series).
- Axis and scale harmonizationNormalize axes to TCZYX order; persist physical voxel scales (XYZ in micrometers) and time intervals.
- Deskewing (lattice light-sheet)optionalCorrect the oblique scan angle of lattice/diSPIM acquisitions to produce orthogonal volumes; applied when stage-scan geometry requires it.
- Chunk planningChoose Zarr/N5 chunk shapes matched to downstream workloads (registration, deconvolution, streaming to viewer).
- Streaming conversionWrite OME-Zarr or BDV/N5-compatible stores with multiscale pyramids without loading full volumes into memory.
- Checksum and integrity validationCompute per-chunk checksums and validate file completeness against the expected acquisition manifest.
Inputs
- raw multi-view/multi-tile 3D/4D light-sheet volumesCZI · IMS · TIFF series · HDF5/XML · OME-TIFF · Imaris HDF5 · KLB
Outputs
- streamed multiscale volume storeOME-Zarr · N5 · HDF5/XML (BDV)
- volume inventory with checksumsYAML · Parquet · JSON
QC checks
- Volume completeness — Verify all expected views, tiles, channels and timepoints are present; flag missing planes or truncated files.
- File integrity (checksums) — Validate per-chunk MD5/Blake2 checksums against acquisition manifest; detect partial writes or corruption.
- Voxel size consistency — Check that physical voxel sizes are consistent across views and match instrument metadata; flag >5% discrepancy.
- Volume size sanity — Confirm reported volume dimensions (XYZ × T × C × views) match file byte-sizes within tolerance.
Failure modes it guards against
- Missing or truncated files — Verify file count and sizes against acquisition manifest before proceeding; fail closed.
- Incorrect voxel calibration — Cross-check OME-XML metadata with instrument log; require explicit unit confirmation.
- Incorrect axis order (XYZ vs ZYX) — Enforce TCZYX normalization at ingest; add axis-label unit tests.
- Memory blowup from non-chunked reads — Always use chunked/streaming reads (zarr, BDV); never load full volume into RAM.
- Deskew not applied to lattice acquisitions — Detect scan geometry from metadata; auto-apply deskew for LLS/diSPIM data.
Tools (11)
Read proprietary and OME microscopy files (CZI, IMS, ND2, LIF, TIFF); preserve vendor metadata as OME-XML.
HDF5/XML and N5 data model and streaming viewer for large multi-view light-sheet datasets.
Convert, validate and write OME-Zarr/OME-NGFF datasets; multiscale pyramid generation.
- zarrLibrary · MIT
Chunked N-dimensional array storage backend used by OME-Zarr; supports local, S3, GCS.
- tifffileLibrary · BSD-3-Clause
Read/write TIFF, BigTIFF, OME-TIFF, and pyramidal image variants; fast Python-native.
- bioioOSS · BSD-3-Clause
Plugin-based Python image reader supporting OME-TIFF, CZI, ND2, OME-Zarr; successor to AICSImageIO.
- bfioOSS · MIT
Python interface to Bio-Formats for reading/writing OME-TIFF and OME-Zarr without requiring full Java runtime.
- FractalOSS · BSD-3-Clause
Distributed bioimage workflow orchestration system centered on OME-Zarr; handles TB-scale plate ingestion.
- LLSpyOSS · BSD-3-Clause
Lattice light-sheet post-processing utility including deskewing, deconvolution and format conversion.
- Imaris File ConverterCommercial · commercial
Convert proprietary Imaris IMS format to OME-TIFF or other open formats.
- n5-ij (N5 ImageJ plugin)OSS · BSD-2-Clause
ImageJ/Fiji plugin for reading and writing N5 format volumes; used in BDV/BigStitcher ecosystem.
Metadata captured
| Field | Type | Example | Standard |
|---|---|---|---|
| pipeline_run_id | string | ls-ingest-2026-06-05-abc123 | — |
| software_versions | object | {'python': '3.12', 'bioio': '3.3.0', 'ome_zarr_py': '0.10.2', 'container': 'sha256:...'} | REMBI |
| input_dataset_uri | uri | s3://client-study/raw/embryo01.czi | OME-NGFF |
| output_dataset_uri | uri | s3://client-study/zarr/embryo01.zarr | OME-NGFF |
| view_count | int (count) | 4 | OME-XML |
| tile_grid_shape | array (count) | 3, 4 | OME-XML |
| voxel_size_xyz | array (micrometer) | 0.208, 0.208, 1.0 | OME-XML |
| z_step_micrometer | float (micrometer) | 1.0 | OME-XML |
| time_points | int (count) | 360 | OME-XML |
| time_interval | float (second) | 30.0 | OME-XML |
| channel_count | int (count) | 3 | OME-XML |
| chunk_shape | array (pixel) | 1, 1, 64, 512, 512 | OME-NGFF |
| deskew_applied | bool | False | — |
| acquisition_system | string | Zeiss Z.1 / Lightsheet 7 | REMBI |
You get
- Streamed OME-Zarr store · OME-Zarr
- Acquisition inventory · YAML/Parquet
References
- OME-NGFF specification spec
- REMBI: Recommended Metadata for Biological Images paper
- BigStitcher: reconstructing high-resolution image datasets of cleared and expanded samples paper
- Image processing tools for petabyte-scale light sheet microscopy data review
- Bio-Formats: a resource for interoperability of microscopy data docs