Journal: PNAS Nexus
Article Title: AI system for diagnosing mucosa-associated lymphoid tissue lymphoma and diffuse large B cell lymphoma using ImageNet and hematoxylin and eosin–stained specimens
doi: 10.1093/pnasnexus/pgaf137
Figure Lengend Snippet: Process of providing training image data to the AI model. TMAs for DLBCL were stained using H&E, and for BCL6, MUM1, and CD10, IHC was used to acquire essential data for Hans’ classification. Based on the positivity or negativity of these stains, the samples were labeled as either GCB or non-GCB. Each image was captured at 400× magnification. The original images were divided into 16 patch images. From these, patches that lacked sufficient tissue content (A), fibrotic scar tissue (B), adipose tissue (C), or blood vessels (D) were excluded. The remaining images were resized and provided as teacher image data to the model. Scale bars represent 20 µm.
Article Snippet: This study analyzed data from 160 patients, comprising 25 normal lymph nodes (NL), 26 MALT lymphoma, 31 GCB, and 78 non-GCB cases purchased from Biomax tissue microarrays (TMAs) (Fig. and Table ).
Techniques: Staining, Labeling