Report a VIPP analysis

Readers should be able to identify the software, reconstruct the graph, find the data, and understand which decisions or exclusions were manual.

Methods checklist

Report:

• napari-vipp version (and commit for nightly/development builds);
• Python and important dependency versions or an environment file;
• workflow JSON and, if used, generated Python;
• input dataset identifiers, acquisition details, series selection, and axes;
• channel mapping, physical scale, and any metadata corrections;
• source checksums/identities and the source revision accepted at calculation;
• for multi-input nodes: how physical-grid alignment, scale, units, and origin were established, plus any explicit registration/resampling performed;
• node sequence, decisive parameter values, and 2D/3D handling;
• for automatic thresholds: method, stack/slice scope, and the saved float histogram-bin count where applicable; for Minimum, also the maximum smoothing iterations;
• for Rescale Intensity or Clip: percentile/data-range/explicit-value mode and its saved cutoffs;
• how non-finite source values were detected, justified, and handled when they occurred;
• tuning set, held-out validation set, reference annotation procedure, and quantitative metrics;
• manual calculations, exclusions, retries, and QC criteria;
• batch source pairing, naming, output formats, overwrite behavior, workflow and config hashes, manifest/run id, item/output statuses, retries, and partial or skipped outputs;
• location and license of data, workflows, code, validation artifacts, and output tables.

Suggested concise wording

Images were processed with napari-vipp 0.12.0a3 using the archived workflow JSON [identifier]. The workflow was developed on [development set] and frozen before evaluation on [held-out set]. Intermediate masks and labels were reviewed using predefined criteria [reference], and [metrics] were calculated against [reference annotations/phantoms].

An appropriate threshold detail could read: “The complete float32 stack was segmented with Otsu's method using 1,024 histogram bins; all finite pixels were included and non-finite pixels were assigned to background.” Do not report the inspector chart's display bins as the algorithm setting.

Adapt the text to what was actually done. Do not state that VIPP guarantees reproducibility, metadata fidelity, or biological correctness.

Archive an analysis bundle

Prefer a DOI-backed repository for final artifacts. Include:

analysis/
  README.md
  environment.yml or requirements.txt
  workflow.json
  generated_pipeline.py        # if used
  vipp_batch_config.json        # if batch processing was used
  vipp_batch_manifest.json      # latest finalized run
  batch-manifest-archives/
  batch-item-sidecars/
  inputs-manifest.csv
  validation/
  qc/
  outputs/

Use checksums or persistent dataset identifiers when input images cannot be redistributed. Review paths and metadata for confidential information before deposit.

For a batch analysis, preserve the exact workflow/config pair and their hashes; do not archive only the thin runner. Retain partial, skipped, and failed status evidence as well as successful outputs. Sidecars are a recovery trail and should be interpreted with the finalized manifest after an interruption.

Cite VIPP

Use the current record in the application's CITATION.cff. Until a software paper or DOI is available, cite the exact software release and repository URL in addition to describing the workflow.