Morphology-based cytogenetic risk prediction in multiple myeloma from bone marrow smears

Researchers developed a computational framework to predict cytogenetic risk in multiple myeloma directly from bone marrow smears. By combining a hematology-specific vision transformer with attention-based learning, the model identifies morphological features linked to high-risk genetic markers. This automated approach achieved AUCs up to 0.85, offering a rapid, cost-effective screening alternative to traditional genetic testing.