Lymphoma Cell Nuclei Classification using Color and Morphology Features

Abstract Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of Non-Hodgkin’s Lymphoma, presenting a great challenge for treatment due to its highly heterogeneous nature. DLBCL diagnosed based on microscopy images patient tissue samples. To help gain better understanding DLBCL, we developed an automated computer vision method analyze morphological and color-based information within biopsies. We analyzed dataset whole slide by segmenting individual cells representing cell morphologies through set engineered features. The features were evaluated using variety visualization machine learning (ML) classification techniques. Current state-of-the-art deep methods use as input in tasks achieving high performance but lacking interpretability. A big lies finding out what pixel-based utilize prediction. Here, present technique that not only yields prediction accuracy also provides insights into which are key show have highest importance classification, allowing accurate identification various types with 84% multi-class 91% binary classification. Our results provide valuable exploring image datasets in-depth view tumor microenvironment.