{
  "abstract": "Artificial intelligence (AI) has reached an inflexion point in diagnostic pathology. Convolutional neural networks and foundation models trained on millions of histopathology images have the potential to place AI alongside traditional diagnostic tools. These models detect cancer, distinguish histological subtypes, predict patient outcomes and impute actionable biomarkers. Among their most promising capabilities is the ability to directly impute molecular and immunophenotypic features from H&E slides, potentially eliminating the need for additional immunohistochemistry or access to a molecular laboratory. Consensus forecasting by an international Delphi panel of 24 computational pathology experts anticipates routine, impactful use of AI across pathology workflows by 2030, while also underscoring today’s translation gap and the regulatory/ethical work still required for safe clinical implementation.1",
  "authors": [
    {
      "affiliations": [
        "Pathology, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA",
        "Harvard Medical School, Boston, Massachusetts, USA"
      ],
      "name": "Vikram Deshpande"
    },
    {
      "affiliations": [
        "Harvard Medical School, Boston, Massachusetts, USA",
        "Department of Pathology, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA"
      ],
      "name": "Monika Vyas"
    },
    {
      "affiliations": [
        "Dana-Farber Cancer Institute, Boston, Massachusetts, USA"
      ],
      "name": "William Lotter"
    }
  ],
  "title": "Towards generalisable and equitable artificial intelligence in pathology",
  "uid": "fb89714d-bca2-5be4-8646-782419fc9bf0"
}