Table 3 Confusion matrix (actual versus predicted classes), accuracy, sensitivity, specificity, AUROC of the support vector machine classification on each of 11 models.
From: Machine learning prediction of pathologic myopia using tomographic elevation of the posterior sclera
Model | Variables | Predicted | Actual classes | Accuracy (%) | Sensitivity (%) | Specificity (%) | AUROC (%) | |
|---|---|---|---|---|---|---|---|---|
Healthy | PM | |||||||
1 | TEPS fovea→disc and TEPS fovea→DPE | Healthy | 189 | 9 | 85.27 | 77.50 | 86.70 | 82.10 |
PM | 29 | 31 | ||||||
2 | TEPS distance and TEPS fovea→DPE | Healthy | 177 | 9 | 80.62 | 77.50 | 81.19 | 79.3 |
PM | 41 | 31 | ||||||
3 | TEPS disc→DPE and TEPS fovea→DPE | Healthy | 190 | 9 | 85.6 | 77.50 | 87.16 | 82.33 |
PM | 28 | 31 | ||||||
4 | TEPS distance and TEPS fovea→disc | Healthy | 190 | 14 | 83.72 | 65.50 | 87.1 | 76.08 |
PM | 28 | 26 | ||||||
5 | TEPS disc→DPE and TEPS fovea→disc | Healthy | 189 | 9 | 85.27 | 77.50 | 86.70 | 82.10 |
PM | 29 | 31 | ||||||
6 | TEPS distance and TEPS disc→DPE | Healthy | 201 | 21 | 85.27 | 47.50 | 92.20 | 69.85 |
PM | 17 | 19 | ||||||
7 | 4 TEPS variables | Healthy | 192 | 9 | 86.43 | 77.50 | 88.07 | 82.79 |
PM | 26 | 31 | ||||||
8 | AL | Healthy | 174 | 14 | 77.52 | 65.00 | 79.82 | 72.41 |
PM | 44 | 26 | ||||||
9 | CT | Healthy | 165 | 15 | 73.64 | 62.50 | 75.69 | 69.09 |
PM | 53 | 25 | ||||||
10 | AL & CT | Healthy | 167 | 10 | 76.36 | 75.00 | 76.61 | 75.80 |
PM | 51 | 30 | ||||||
11 | All variables | Healthy | 204 | 8 | 91.47 | 80.00 | 93.58 | 86.47 |
PM | 14 | 32 | ||||||
