Hematological malignancies, particularly multiple myeloma (MM) and lymphomas, pose major clinical challenges due to their biological complexity and inter-patient heterogeneity. Although diagnostic and therapeutic approaches have evolved, significant gaps remain in the integration of multi-omics data, risk stratification, and treatment response prediction.

This state-of-the-art review examines recent developments in artificial intelligence (AI) applied to MM and lymphomas. A systematic literature search was conducted in PubMed, Web of Science, Scopus, and specialized journals including the Journal of Hematology & Oncology, Leukemia, and Blood for studies published between 2018 and 2024. After screening and full-text assessment, 50 studies met the inclusion criteria following PRISMA guidelines (Figure 1). Studies were selected based on methodological rigor and clinical relevance.

AI models demonstrate robust capabilities across diagnostic, prognostic, and therapeutic applications. Single-center studies report outstanding metrics, including AUCs up to 0.99 for myeloma lesion classification, while multicenter validation yields more conservative yet robust metrics. In both diseases, multimodal approaches consistently outperform unimodal models across all clinical applications. Despite these advances, key challenges in data diversity, technical heterogeneity, and model interpretability remain under active investigation.

AI shows transformative potential for MM and lymphoma management, particularly through multimodal integration. Bridging the gap with clinical practice requires transparency, computational efficiency, and ethically grounded validation. In addition, close collaboration among clinicians, data scientists, and institutions is essential. These combined efforts are key to establishing AI as a reliable tool in everyday hematology.