Artificial intelligence in myeloproliferative neoplasms: From digital bone marrow pathology to clinical decision...

Myeloproliferative neoplasms (MPN) are clonal hematopoietic stem cell disorders whose diagnosis, prognostication, and management depend on integrating heterogeneous clinical, morphological, and molecular data. This narrative review (PubMed/MEDLINE searched through August 2026, supplemented by hand-searching and conference abstracts) appraises the evidence for AI applications across the MPN continuum. Artificial intelligence (AI), and its principal subfields, machine learning (ML) and deep learning (DL), learn predictive patterns directly from data rather than from explicitly programmed rules, and are increasingly suited to this multidimensional complexity. Across the MPN continuum, AI has produced substantial technical progress across several domains: automated bone marrow biopsy analysis and continuous quantification of fibrosis and megakaryocyte morphology; peripheral blood- and complete blood count-based diagnosis, including blood-based prediction of marrow fibrosis and emerging liquid-biopsy approaches; quantitative disease assessment such as spleen volume segmentation; risk stratification for survival, thrombosis, disease progression, and post-transplant outcomes; therapy response prediction; in silico drug discovery; and natural language processing for large-scale mining of electronic health records; maturity and level of validation differ markedly across these domains. External validation is increasingly reported in selected applications, though most published models to date remain confined to internal or single-cohort validation, and prospective evaluation has been reported only for blood count-based fibrosis prediction. Correspondingly, to our knowledge, no tool has yet been shown to change clinical decision-making or improve patient outcomes, and predictive performance should not be conflated with clinical utility. As multimodal data integration, foundation models, and generative approaches mature, AI has the potential to support a faster, more objective, and more personalized diagnostic and decision-making pathway in MPN.