Multimodal ML framework for integrative analysis in hematology. The integration of heterogeneous data types (clinical variables, radiologic imaging, histopathology, and high-throughput sequencing) using a multimodal ML pipeline is depicted. After preprocessing and data fusion, ML models are trained to capture relationships across modalities. These models support key clinical applications including the classification of hematologic malignancies (eg, lymphoma subtyping), prediction of clinical outcomes (eg, survival or progression after CAR T-cell therapy), discovery of prognostic and predictive biomarkers, and development of digital twins to simulate individualized treatment responses. Icons in this figure were generated using ChatGPT (OpenAI).