Skip to content

Latest commit

 

History

History
218 lines (162 loc) · 7.88 KB

File metadata and controls

218 lines (162 loc) · 7.88 KB

Imaging Analysis Architecture

Status: Production (openimagedata: 100% real, deepcell: 100% real) Last Updated: 2026-01-31


System Overview

graph TB
    subgraph Input["📁 Input Images"]
        HE[H&E TIFF<br/>RGB brightfield<br/>Morphology]
        IF1[IF_CD8 TIFF<br/>Grayscale fluorescence<br/>T cells]
        IF2[IF_Ki67 TIFF<br/>Grayscale fluorescence<br/>Proliferation]
        MX[MxIF TIFF<br/>3-channel RGB<br/>TP53/Ki67/DAPI]
    end

    subgraph OID["🔧 mcp-openimagedata<br/>(5 tools, 100% real)"]
        Load[Load Image]
        Comp[Generate<br/>Composite]
        Anno[Annotate<br/>H&E]
    end

    subgraph DC["🔬 mcp-deepcell<br/>(3 tools, 100% real)"]
        Seg[Segment<br/>Cells]
        Quant[Quantify<br/>Markers]
        Viz[Generate<br/>Overlay]
    end

    subgraph CC["🎯 mcp-cell-classify<br/>(3 tools, 100% real)"]
        Class[Classify<br/>Phenotypes]
        Multi[Multi-marker<br/>Phenotyping]
        PhenoViz[Phenotype<br/>Visualization]
    end

    subgraph Output["📊 Outputs"]
        Morph[H&E Morphology<br/>Necrosis, cellularity]
        Counts[Cell Counts<br/>CD8+, Ki67+]
        Pheno[Phenotypes<br/>TP53+/Ki67+ co-expression]
        Imgs[PNG Visualizations<br/>Composites, overlays]
    end

    HE --> Load
    Load --> Anno
    Anno --> Morph
    Anno --> Imgs

    IF1 --> Load
    IF2 --> Load
    MX --> Load

    Load --> Comp
    Comp --> Imgs

    Load --> Seg
    Seg --> Quant
    Quant --> Class
    Quant --> Multi
    Class --> Counts
    Multi --> Pheno
    Multi --> PhenoViz
    PhenoViz --> Imgs
    Viz --> Imgs

    classDef inputStyle fill:#e1f5ff,stroke:#0288d1,stroke-width:2px
    classDef oidStyle fill:#fff3e0,stroke:#f57c00,stroke-width:2px
    classDef dcStyle fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
    classDef ccStyle fill:#e8f5e9,stroke:#388e3c,stroke-width:2px
    classDef outputStyle fill:#f1f8e9,stroke:#689f38,stroke-width:2px

    class Input inputStyle
    class OID oidStyle
    class DC dcStyle
    class CC ccStyle
    class Output outputStyle
Loading

Quick Navigation

Servers


What This Is

Imaging analysis component for histology and multiplexed immunofluorescence (MxIF) in the Precision Medicine MCP system.

Current workflows:

  1. H&E: Brightfield morphology assessment (chromogenic stains, RGB TIFF)
  2. MxIF: Fluorescence cell segmentation (fluorescent antibodies, multi-channel TIFF)

Servers:

  • mcp-openimagedata (100% real - loading, registration, feature extraction, visualization)
  • mcp-deepcell (100% real - segmentation + per-cell marker quantification)
  • mcp-cell-classify (100% real - phenotype classification + visualization, lightweight)

H&E vs MxIF: Key Differences

Feature H&E MxIF
Microscopy Brightfield Fluorescence
Staining Chromogenic dyes Fluorescent antibodies
Format RGB TIFF Grayscale (single) or Multi-channel (multiplex)
Purpose Visual morphology Quantitative protein expression
Analysis Visual inspection Automated cell segmentation
Servers openimagedata ONLY openimagedata → deepcell → cell-classify
Output Annotated images Cell counts, phenotypes
Example Necrosis detection CD8+ T cell quantification

Key Point: H&E uses chromogenic stains (not fluorescence!) for visual assessment.


Server Status

mcp-openimagedata

Status: ✅ 100% Real (deployed to GCP Cloud Run) URL: https://mcp-openimagedata-ondu7mwjpa-uc.a.run.app

Tools (5):

  • fetch_histology_image — PIL image loading with glob fallback for partial ID matching
  • register_image_to_spatial — Otsu tissue detection, bbox-based affine/rigid estimation, phase-cross-correlation refinement for deformable; Visium + generic x/y CSV support
  • extract_image_features — LBP + GLCM texture (25 features), connected-component morphology (15 features), intensity stats + entropy (10 features); per-ROI support
  • generate_multiplex_composite — RGB MxIF composites (1-7 channels)
  • generate_he_annotation — Annotate H&E morphology with region overlays

Use cases:

  • Load H&E and IF/MxIF images
  • Register histology to spatial transcriptomics spot coordinates
  • Extract texture, morphology, and intensity features per region
  • Generate multiplex RGB composites
  • Annotate H&E regions (necrosis, high cellularity)

mcp-deepcell

Status: ✅ 100% Real (deployed to GCP Cloud Run) URL: https://mcp-deepcell-ondu7mwjpa-uc.a.run.app

Tools (3):

  • segment_cells — DeepCell-TF nuclear/membrane segmentation, 16-bit TIFF mask output
  • quantify_markers — Per-cell mean/max/min intensity for multiple markers, CSV output
  • generate_segmentation_overlay — Cell boundary visualization overlaid on original image

Use cases:

  • Segment cells from DAPI nuclear stain or membrane markers
  • Quantify per-cell marker intensities for downstream classification
  • Validate segmentation quality with overlay visualizations

mcp-cell-classify

Status: ✅ 100% Real (lightweight, no TensorFlow dependency)

Tools (3):

  • classify_cell_states — Single-marker threshold classification (proliferating/quiescent/intermediate)
  • classify_multi_marker — Multi-marker phenotyping (e.g., Ki67+/TP53- assignments)
  • generate_phenotype_visualization — Color cells by marker expression (positive/negative)

Use cases:

  • Classify cell phenotypes from segmentation masks + marker images
  • Multi-marker co-expression analysis (Ki67+/TP53+ double-positive cells)
  • Generate publication-quality phenotype visualizations

Note: Split from mcp-deepcell for lighter dependencies (~200MB vs ~2GB Docker image). Users can swap in alternative classifiers (FlowSOM, Leiden, scikit-learn).


PatientOne Integration (TEST_4)

Test Files

File Type Microscopy Servers Purpose
PAT001_tumor_HE_20x.tiff H&E Brightfield openimagedata Morphology assessment
PAT001_tumor_IF_CD8.tiff IF (single) Fluorescence openimagedata + deepcell CD8+ T cell counts
PAT001_tumor_IF_KI67.tiff IF (single) Fluorescence openimagedata + deepcell Ki67+ proliferation index
PAT001_tumor_multiplex_IF_TP53_KI67_DAPI.tiff MxIF (3-ch) Fluorescence openimagedata + deepcell Multi-marker phenotyping

Expected Findings

H&E Morphology:

  • Necrosis: Present (pale regions)
  • Cellularity: 70-80% tumor cells
  • Architecture: HGSOC (papillary, high-grade)

MxIF Quantification:

  • CD8+ cells: ~12 (LOW, immune exclusion)
  • Ki67+ cells: ~112 (HIGH, 45% proliferation)
  • TP53+/Ki67+ double-positive: ~85 cells (35%, active growth with mutation)

Quick Start

For users: See server READMEs above for workflow details → Run PatientOne TEST_4_IMAGING

For developers: See mcp-openimagedata README and mcp-deepcell README for tool details


Related Workflows


See also: Main Architecture | PatientOne README