Table 3 Evaluation results using the PCA principal axis method. The results include four evaluation metrics: precision, sensitivity, F1 score, and specificity. The ‘Support’ column indicates the sample size for each classification, while the max (L1/L2, R1/R2) and max (L3/L4, R3/R4) represent two measurement methods of reference labels
From: A cascading learning method with SegFormer for radiographic measurement of periodontal bone loss
Precision | Sensitivity | F1-score | Specificity | Support | |
|---|---|---|---|---|---|
Stage I (max (L1/L2, R1/R2)) | 0.7651 | 0.6196 | 0.6847 | 0.9542 | 368 |
Stage I (max (L3/L4, R3/R4)) | 0.7685 | 0.6451 | 0.7014 | 0.9552 | 355 |
Stage II (max (L1/L2, R1/R2)) | 0.7424 | 0.7885 | 0.7648 | 0.7337 | 936 |
Stage II (max (L3/L4, R3/R4)) | 0.7334 | 0.8046 | 0.7674 | 0.7326 | 906 |
Stage III (max (L1/L2, R1/R2)) | 0.7802 | 0.7960 | 0.7880 | 0.8980 | 593 |
Stage III (max (L3/L4, R3/R4)) | 0.8149 | 0.7752 | 0.7945 | 0.9112 | 636 |
Macro avg (max (L1/L2, R1/R2)) | 0.7625 | 0.7347 | 0.7458 | 1897 | |
Macro avg (max (L3/L4, R3/R4)) | 0.7722 | 0.7416 | 0.7444 | 1897 |