Table 5.
Confusion matrix and performance of the model used for pathological validation of perineural invasion (whole slide images).
First pathologist (LYL)* | |||
Actual: positive (n) | Actual: negative (n) | ||
Predicted: positive | TP (12) | FP (7) | |
Predicted: negative | FN (4) | TN (384) | |
Sensitivity: 75.00% | Specificity: 98.20% | Accuracy: 97.29% | |
Second pathologist (CHY)* | |||
Actual: positive (n) | Actual: negative (n) | ||
Predicted: positive | TP (11) | FP (8) | |
Predicted: negative | FN (1) | TN (385) | |
Sensitivity: 91.66% | Specificity: 97.96% | Accuracy: 97.77% |
TP, true positive; FP, false positive; FN, false negative; TN, true negative.
*Accuracy for recognition of nerve structures: First pathologist, 87.90% (407/463); Second pathologist, 87.09% (405/460).