{
  "abstract": "Background Digital pathology has enabled high-throughput, remote analysis of stained tissue slides for research and diagnostics. However, AI-based analysis tools have lagged, especially for complex staining protocols like multiplex immunofluorescence (mIF). Key challenges include heterogeneous antibody panels and difficulty generalizing algorithms across different chemistries. To address this, we present a novel end-to-end workflow combining ZEISS Axioscan 7 spatial biology, SlideStream scan automation software, and Mindpeak’s PhenoScout AI system. At the core is PhenoScout, a library of foundation AI models designed for tissue analytics across wide variety of mIF and IHC stainings, requiring no prior image analysis experience.Methods Using a representative antibody panel (FOXP3, CD3, Ki67, CD8, CD68, PanCK) designed to interrogate tumor microenvironment phenotypes, Concept Life Sciences stained a cohort of tissue sections using mIF protocols as well as separate IHC DAB stainings. Slides were digitized using the ZEISS Axioscan 7 spatial biology system with SlideStream workflow automation software, ensuring consistent, automated, high-resolution image capture. Images were processed through PhenoScout’s foundation models, which have NOT not been trained on this image set.PhenoScout includes pre-trained AI modules for:Region detection: e.g., tumor vs. stroma using cytokeratin markersCell detection: DAPI-based cell center detectionMarker positivity classification: compartment-specific intensity-based classificationPhenotype analysis: integration of multi-channel biomarker signals to assign mIF-based phenotypesThese models were trained on millions of annotated cells covering common membranous and nuclear biomarker patterns. Validation involved comparing IHC and mIF versions of markers across serial sections.Results This seamless workflow enabled robust plug-and-play imaging, automated image upload on Image Management System, and browser-based visualization and analysis with minimal user intervention. Initial results across different panels showed high concordance between mIF- and IHC-based phenotyping for key immune markers (FOXP3, CD3, CD8, CD68), epithelial marker (PanCK), and proliferation marker (Ki67). The PhenoScout algorithms performed consistently across tissue types and staining variations, enabling accurate spatial quantification of phenotypic distributions.Conclusions We demonstrate a fully integrated pipeline for AI-powered tissue analysis across multiplexed and chromogenic stains, combining ZEISS SlideStream automation with flexibility of Mindpeak PhenoScout. The foundation AI models enable first-pass interpretation and discovery of spatial biomarker patterns, accelerating translational research. Future work will focus on developing precision-trained models for advanced oncology, immunology, and drug development applications.",
  "authors": [
    {
      "affiliations": [
        "Mindpeak GmbH, Hamburg, Germany"
      ],
      "name": "Fabian Schneider"
    },
    {
      "affiliations": [
        "Mindpeak GmbH, Hamburg, Germany"
      ],
      "name": "Heiko Schmitt"
    },
    {
      "affiliations": [
        "Mindpeak GmbH, Hamburg, Germany"
      ],
      "name": "Ralf Banisch"
    },
    {
      "affiliations": [
        "Concept Life Sciences, Edinburgh, UK"
      ],
      "name": "Christopher Mills"
    },
    {
      "affiliations": [
        "Concept Life Sciences, Edinburgh, UK"
      ],
      "name": "Lydia Blackburn"
    },
    {
      "affiliations": [
        "ZEISS Microscopy GmbH, Munich, Germany"
      ],
      "name": "Sherry Derakhshani"
    },
    {
      "affiliations": [
        "ZEISS Microscopy GmbH, Munich, Germany"
      ],
      "name": "Moritz Widmaier"
    }
  ],
  "title": "71 PhenoScout: foundation AI models for end-to-end tissue analysis of mIF and IHC slides",
  "uid": "7828d924-03b7-561a-9ccd-21412ae1653e"
}