Fig. 9

Performance comparison with mean and standard deviation. The bar chart shows the mean performance and standard deviation (represented by error bars) of two methods:one trained using PathoGraph and the other using patch graph, evaluated across five random seeds. Performance metrics include accuracy, weighted one-versus-rest F1 score, and one850versus-rest F1 scores for each class (“Benign”, “Atypical”, and “Malignant”).