AI-driven morphological clustering for lymphoma stratification

Diffuse large B-cell lymphoma (DLBCL) is the most common lymphoid neoplasm in Western countries and is an aggressive, particularly heterogeneous tumor that differs from non-Hodgkin's lymphoma in symptoms, genetic profile, and therapeutic response. The use of artificial intelligence approaches for integrating large datasets is critical to understanding the complexity of the pathology and tumor environment to improve survival predictions in cancer studies. Stratification of cancer patients is a crucial step in personalizing therapies and optimizing clinical approaches by allowing separation of patients based on clinical characteristics, biomarkers, and histological data. The aim of this study is the clustering of patients with DLBCL, based on morphological features, using artificial intelligence to obtain a stratification of subjects reflecting their survival. To obtain significant homogeneous groups, histopathological images from lymphoma patients were used using unsupervised partitional algorithms. Integration of these machine learning methods into clinical practice can lead to optimization of treatment strategies and effective disease management, increasing the likelihood of therapeutic success and improving the patients' own quality of life.