Fig. 4
From: Comprehensive Benchmark Dataset for Pathological Lymph Node Metastasis in Breast Cancer Sections
Radar chart analysis of MIL models across different encoders and dataset versions. Each subplot shows the performance of multiple MIL models on the Camelyon-17-Origin23 and Camelyon-17-Refine datasets using three feature encoders: (a,e) PLIP3, (b,f) UNI4, and (c,g) Gigapath5. The outermost yellow line represents the average of AUC and F1-score, green represents AUC, and blue represents F1-score. (d,h) summarize the total ranking across all encoders for each dataset. Top-3 ranked models are highlighted with dashed boxes.
