Fig. 7
From: Deep learning for histopathological segmentation of smooth muscle in the urinary bladder
Comparison of patch-based (blue) and pixel-based (orange) models with whole image-based inference in Mean Jaccard Index (A), Mean Dice Coefficient (B), Pixelwise Accuracy (C), Precision (D), Recall (E), Specificity (F), and F1 Score (G). The models were evaluated by ninefold cross-validation and the seven TUR images in each fold were used to calculate the evaluation metrics. For patch-based or pixel-based models respectively, the group means are indicated by black dashed lines