Runs and reproducibility

Every analysis step is a versioned pipeline run suitable for methods sections and audit trails.

Finding runs

Location	What you see
Study home — Run history	All pipeline runs for the study
Slide detail — pipeline buttons	Latest completed run per pipeline type
Slide detail — QC panels	Summary statistics from latest runs

Run record

Each run stores:

• Pipeline type (tile_slide, tissue_detect, segment_cells, spatial_quantification, …)
• Status, timestamps, and run_id
• Parameter JSON and pipeline version
• Result summary (cell count, tissue fraction, density, etc.)
• Artifact paths for masks, overlays, and export files

Provenance chains

Common chains:

spatial_quantification → segment_cells → tile_slide → slide upload
segment_cells → tile_slide → slide upload
tissue_detect → tile_slide → slide upload

Segmentation requires tiles. Spatial quantification requires completed segmentation.

Artifacts

Pipeline outputs live under:

data/pathology/tiles/{slide_id}/     tile pyramid and thumbnail
data/pathology/artifacts/{run_id}/   masks, overlays, JSON, CSV

Access via slide detail export links or the metadata API:

curl "http://localhost:8000/pathology/artifacts/file?path=..."

Failed runs

Failed jobs retain error messages in the job poll response. Common causes:

• Tiles not generated before segmentation
• Unsupported or corrupt image format
• Segmentation not complete before spatial quantification
• Missing OpenSlide library for certain WSI formats

Fix the underlying issue and re-run from the slide detail page.

Versioning

Every run records pipeline_version (currently 0.1.0 for pathology pipelines). Include run IDs and versions in methods sections for reproducibility.

Re-running pipelines

Launch the same pipeline again from the slide detail page. A new run is created; the UI displays results from the latest completed run per pipeline type.