cp-extract-segmentation-artifacts

cp-extract-segmentation-artifacts produces the main segmentation output set.

It writes the labels, outlines, and object tables that downstream single-cell steps reuse.

Purpose

Use this skill when you want:

  • segmented nuclei

  • segmented whole cells

  • review images to inspect whether the boundaries look reasonable

  • reusable object definitions for later crop and feature-extraction steps

Main Outcome

This skill produces the standard segmentation artifact bundle:

  • nuclei and whole-cell label images

  • outline overlays for review

  • CellProfiler object tables

  • workflow files that document how the segmentation was run

Inputs

This skill reads:

  • a project config such as configs/project_config.demo.json

  • the segmentation input table written by cp-prepare-segmentation-inputs, or default segmentation inputs resolved from the config

  • the selected segmentation .cppipe

  • the raw Cell Painting images and illumination-correction files from the project config

  • an optional output directory

Outputs

The main user-facing results are:

  • label images for nuclei and whole cells

  • outline overlays for visual QC

  • CellProfiler tables such as Image.csv, Cells.csv, Nuclei.csv, and Cytoplasm.csv

  • workflow files such as the segmentation .cppipe, summary, and manifest

Agent Demo

This page is based on a real local OpenClaw inspection of a completed demo segmentation result:

  • session id: segartifact-inspect-v1

  • recorded on 2026-05-06

  • model: vibe/gpt-5-mini

Request

I already have a completed demo segmentation result. Please inspect the result under demo/workspace/outputs/agent_demo_segmentation/02_extract_artifacts_v6 and tell me what outputs were produced there and what they mean biologically when I review a Cell Painting segmentation result.

Structured Trace

user_input:
I already have a completed demo segmentation result. Please inspect the result under demo/workspace/outputs/agent_demo_segmentation/02_extract_artifacts_v6 and tell me what outputs were produced there and what they mean biologically when I review a Cell Painting segmentation result.

recorded_turn:
- the saved OpenClaw turn answered this request directly
- the reply identified the segmentation artifact step that produced the directory
- the reply listed the main workflow files, CellProfiler tables, label images, and outline overlays
- the reply explained what nuclei labels, cell labels, outlines, and object tables mean biologically

final_answer_snapshot:
The completed result contains the standard segmentation artifact set: labels, outlines, object tables, and workflow files for two demo fields.

Representative Figure

Representative segmentation output for one real Cell Painting image

Caption: Representative output from cp-extract-segmentation-artifacts, showing the original field, a selected crop, the nucleus boundaries, and the whole-cell boundaries.

Observed Result

The inspected completed result under 02_extract_artifacts_v6 contained:

  • workflow files such as:

    • load_data_for_segmentation.csv

    • segmentation_workflow_config.json

    • CPJUMP1_analysis_mask_export.cppipe

    • pipeline_skill_manifest.json

    • segmentation_summary.json

  • CellProfiler tables:

    • Image.csv

    • Cells.csv

    • Nuclei.csv

    • Cytoplasm.csv

    • Experiment.csv

  • review images:

    • labels/*cell_labels.tiff

    • labels/*nuclei_labels.tiff

    • outlines/*cell_outlines.png

    • outlines/*nuclei_outlines.png

The completed run covered:

  • 2 image fields

  • wells A01 and A02

  • 4 cell objects total

  • 4 nuclei objects total

  • 4 cytoplasm objects total

Recorded Reply

I inspected the existing completed segmentation result under:

- `demo/workspace/outputs/agent_demo_segmentation/02_extract_artifacts_v6`

This is a completed run from the stable segmentation skill:

- **`cp-extract-segmentation-artifacts`**

Next Skills