Gradient Biotech

Study workflow

Pathology follows a left-to-right research flow. Tile generation is required before browser viewing and segmentation on large slides.

Create study → Upload slides → Generate tiles → Tissue detect → Segment cells → Spatial quantification → Cohort / Interpret

Dashboard — area home

The pathology dashboard (/areas/pathology/dashboard) shows recent studies and links to create a new study.

Readiness signal: at least one study with one uploaded slide and a completed tile or analysis run.

Study home — project overview

The study page (/areas/pathology/studies/{id}) is your project hub:

  • Slide ingestion — upload form with modality and stain type selectors
  • Slides grid — thumbnail cards linking to slide detail pages
  • Run history — all pipeline runs for this study

Slide detail — analyze and inspect

Each slide opens at /areas/pathology/studies/{id}/slides/{slideId}:

  • Slide viewer — pan/zoom with overlay layers
  • Pipeline controls — tile, tissue detect, segment, spatial quantification, multiplex phenotyping, color deconvolution, object classifier training, OME-TIFF export, spatial alignment
  • QC panels — tissue fraction, cell count, spatial metrics
  • Cell table — detected objects with area and confidence
  • Region metrics table — compartment-level density and cell statistics

See Slide viewer.

Analyze — study-wide launcher

The Analyze page (/areas/pathology/studies/{id}/analyze) lists every slide in the study with quick-launch buttons for tiling, analysis jobs, and spatial alignment, plus the cohort workspace.

Cohort — group comparison

The Cohort page (/areas/pathology/studies/{id}/cohort) covers:

  • CSV sample import and per-slide sample/group/timepoint assignment
  • A batch pipeline launcher across selected slides
  • The group-comparison table (t-test/Mann-Whitney or ANOVA/Kruskal-Wallis, FDR-corrected)

Interpret — AI-assisted summaries

The Interpret page (/areas/pathology/studies/{id}/interpret) generates a grounded narrative from one or more completed runs, with cited metrics and a research-use disclaimer. It is not yet linked from the study navigation — open it directly at that URL.

Typical paths

Study typePath
Region image demoUpload PNG → tiles → tissue detect → segment → quantify
WSI reviewUpload SVS/OME-TIFF → tiles → tissue QC → segmentation at scale
TME quantificationSegment → spatial quantification → export region metrics CSV
Annotated regionsImport GeoJSON → tissue detect → quantify within labeled compartments

Integration with Computational Biology

Spatial transcriptomics ingestion, deconvolution, and generic spatial statistics are owned by the Computational Biology area. Pathology contributes the WSI tissue context layer — morphology-guided region extraction and overlay alignment — without duplicating compbio spatial pipelines.

Future workflow (planned)

  • Publication-ready report export (methods text, figure export, reproducibility manifest)
  • Cellpose/StarDist deep-learning segmentation models
  • Slide-level and region-level collaboration comments