Deep Learning on Histopathological Images Enables Accurate Diagnosis of Common B-Cell and Hodgkin Lymphoma

Abstract

Simple Summary Lymphoma, a complex cancer of the lymphatic system, includes several subtypes that require different treatments. However, distinguishing these subtypes under a microscope is notoriously difficult, often requiring expensive and time-consuming additional tests. To address this challenge, we developed an artificial intelligence (AI) system named BH-LymphDS, which functions as a digital pathology assistant to analyze routine biopsy slides. By training our system on the largest known collection of lymphoma images (over 1900 slides), we found that it could identify the three most common lymphoma subtypes—Diffuse Large B-cell Lymphoma, Follicular Lymphoma, and Hodgkin Lymphoma—with an accuracy matching or even surpassing that of three experienced hematopathologists. Our findings suggest that this AI tool could help pathologists make faster, more accurate preliminary diagnoses, potentially reducing the need for certain supplementary tests and improving diagnostic efficiency, particularly in settings with limited specialist resources.