Pipelines

All pathology pipelines run in backend-analysis and are dispatched asynchronously. List live definitions and parameter schemas:

curl http://localhost:8001/pathology/pipelines

Ingestion and viewing

Pipeline	Input	Outputs
tile_slide	Uploaded slide image	Tile manifest, thumbnail, pyramid levels

Key parameters: tile_size (default 256 px).

Tissue analysis

Pipeline	Input	Outputs
tissue_detect	Slide with storage path	Tissue mask PNG, region bounding boxes, focus QC summary
segment_cells	Tiled slide	Cell overlay PNG, cells JSON, cell count, confidence summary

Segmentation parameters: model (default threshold_baseline; Cellpose/StarDist planned).

Spatial quantification

Pipeline	Input	Outputs
spatial_quantification	Completed segmentation run	Density heatmap, nearest-neighbor stats, region metrics, infiltration phenotype, CSV/JSON exports

Requires segment_cells to complete first.

Cohort (planned)

Pipeline	Input	Outputs
cohort_compare	Multiple slides with sample group labels	Group comparison of pathology biomarkers

API endpoint exists; full UI workflow is in Phase 4 — see products/pathology/plan.md.

Job lifecycle

queued → running → complete | failed

Poll status: GET /pathology/jobs/{run_id}

Job dispatch endpoints:

POST /pathology/jobs/tile-slide
POST /pathology/jobs/tissue-detect
POST /pathology/jobs/segment-cells
POST /pathology/jobs/spatial-quantification
POST /pathology/jobs/cohort-compare

Versioning

Every run records pipeline_version from the analysis service configuration. Include run IDs and versions in methods sections for reproducibility.

Implementation sources

Tissue detection and baseline segmentation use scikit-image and OpenCV kernels. Future phases integrate Cellpose, StarDist, and HistomicsTK-style color deconvolution. See products/pathology/plan.md for the full implementation map.