Table 2.

Evaluation metrics, formulas, and interpretations for multi-class classification of pathological images.

Metric	Formula	Interpretation
Sensitivity (Recall)	Sensitivity = TP/(TP + FN)	Ability of the model to correctly identify positive samples
Specificity	Specificity = TN/(TN + FP)	Ability of the model to correctly identify negative samples
Positive Predictive Value (PPV/Precision)	Precision = TP/(TP + FP)	Proportion of predicted positives that are true positives
Negative Predictive Value (NPV)	NPV=TN/(FN+TN)	Proportion of predicted negatives that are true negatives
F1-Score	F1 = 2 × (Precision × Recall)/(Precision + Recall)	Harmonic mean of precision and recall, balancing accuracy and recall