Artificial Intelligence in Hematologic Malignancies: From Digital Morphology to Clinical Decision Support

Hematologic malignancies generate complex, multi-source data: blood and bone marrow morphology, flow cytometry, cytogenetics, next-generation sequencing, imaging, and longitudinal clinical outcomes. Clinicians must integrate these signals into time-sensitive decisions across diagnosis, risk stratification, and treatment planning. Artificial intelligence (AI), including machine learning, deep learning, and more recent foundation models and large language models (LLMs), can help by extracting reproducible patterns from high-dimensional datasets and by improving consistency in tasks such as cell classification, MRD interpretation, relapse prediction, and therapy response estimation. Recent work — generalizable bone marrow models such as DeepHeme, hematopathology-specific foundation models, and the first large international validation cohorts in leukemia — has shifted the field from single-center proofs of concept toward multi-site evidence. Yet generalizability, prospective validation, and safe integration into clinical workflows remain unresolved, and translation into routine practice is still uneven.This Research Topic aims to showcase clinically meaningful, validated AI for hematologic malignancies, with a focus on tools that address real decision points in hematology and hematopathology. The collection will emphasize AI approaches that improve diagnostic precision, refine prognostic classification, support treatment selection, and optimize monitoring strategies. We encourage submissions that connect model development to actionable endpoints — classification aligned with WHO/ICC categories, prediction of relapse or transformation, response and toxicity prediction, MRD-informed modeling, and AI-augmented prognostic scoring in AML, MDS, lymphoma, and myeloma — and that demonstrate strong evaluation through external validation, prospective testing, or well-justified clinical validation strategies. We also welcome work on emerging modalities such as foundation models for digital hematopathology, LLM- and retrieval-augmented decision support, and multimodal architectures that combine pathology, omics, flow cytometry, imaging, and EHR data. Our goal is to help define best practices for AI development, reporting, and governance in hematologic oncology, and to highlight studies that move beyond proof-of-concept toward measurable real-world clinical impact.We welcome Original Research, Review, Mini Review and Perspective articles on themes including, but not limited to:- AI for digital hematopathology and morphology, including peripheral blood/bone marrow smear analysis, foundation models, and interpretable deep learning- AI in flow cytometry and MRD, including automated gating, classification, and relapse-risk modeling- Risk stratification and outcome prediction in leukemia, lymphoma, and myeloma using clinicogenomic and real-world datasets, including AI-augmented prognostic scoring- Multimodal models (omics + pathology + flow/EHR/imaging) and clinically useful explainability and uncertainty estimation- LLMs and retrieval-augmented generation for clinical decision support, pathology report summarization, and structured data extraction in hematology- External validation and generalizability, including multi-center, multi-platform, and international testing- Implementation and governance: workflow integration, reporting standards (e.g., CLAIM, TRIPOD+AI), model monitoring, bias auditing, and prospective evaluationPlease note: manuscripts consisting solely of bioinformatics or computational analysis of sole public databases which are not accompanied by validation (independent cohort or biological validation in vitro or in vivo) are out of scope for this section and will not be accepted as part of this Research Topic. Topic editors: Mohamed A Yassin, Salem Alshemmari Source: https://www.frontiersin.org/research-topics/81098/artificial-intelligence-in-hematologic-malignancies-from-digital-morphology-to-clinical-decision-support