This narrative review (PubMed/MEDLINE searched through August 2026) evaluated the evidence for AI, machine learning, and deep learning applications across the full myeloproliferative neoplasm (MPN) continuum — from automated bone marrow biopsy analysis and blood count-based diagnosis to risk stratification, therapy response prediction, drug discovery, and EHR mining.
AI has produced meaningful technical advances across multiple MPN domains, but most published models rely on internal or single-cohort validation only; prospective evaluation has been reported solely for blood count-based fibrosis prediction, and no tool has yet been shown to change clinical decision-making or improve patient outcomes.
As a narrative review, it is subject to selection bias in the literature appraised. The majority of underlying studies lack external validation, limiting generalizability. Conference abstracts were included, introducing risk of reporting immature or unvetted findings.
Clinicians should treat current AI tools in MPN as research-stage aids: predictive performance alone does not equal clinical utility, and no AI tool has yet been prospectively shown to improve patient outcomes. Prospective, externally validated studies are needed before these tools can be adopted in routine care.
Explore related topics