2. Subgroup analyses.
| Subgroups |
No. of studies (participants) |
Sensitivity (95% CI) |
P‐value for relative sensitivity |
Specificity (95% CI) |
P‐value for relative specificity | |
| Study design | Clinical series | 46 (38,788) | 0.987 (0.977, 0.993) |
Reference | 0.984 (0.975, 0.993) |
Reference |
| Registries | 8 (11,731) | 0.975 (0.919, 0.993) |
0.458 | 0.975 (0.936, 0.990) |
0.464 | |
| AI algorithm | Logistic regression | 8 (2,889) | 0.983 (0.957, 0.993) |
Reference | 0.992 (0.974, 0.997) |
Reference |
| Bayesian network | 3 (788) | 0.994 (0.972‐0.999) |
0.260 | 0.982 (0.834, 0.998) |
0.666 | |
| Convolutional neural network | 13 (13,452) | 0.979 (0.945‐0.991) |
0.734 | 0.978 (0.960, 0.988) |
0.110 | |
| Discriminant analysis | 3 (462) | 0.977 (0.945, 0.990) |
0.628 | 1.000 (0.814, 1.000) |
0.093 | |
| Decision tree | 5 (8,96) | 0.976 (0.895, 0.995) |
0.731 | 0.978 (0.935, 0.993) |
0.299 | |
| Neural network | 10 (16,296) | 0.973 (0.920, 0.991) |
0.561 | 0.968 (0.931, 0.986) |
0.093 | |
| Other | 6 (4,338) | 0.990 (0.892, 0.999) |
0.629 | 0.968 (0.931, 0.987) |
0.068 | |
| Random forest | 2 (3,487) | 1.000 (0, 1.000) | 0.038 | 0.997 (0.994, 0.999) |
0.270 | |
| SVM | 4 (7,911) | 0.994 (0.982, 0.998) |
0.203 | 0.993 (0.928, 0.999) |
0.916 | |
| Imaging technique | OCT | 6 (19,585) | 0.971 (0.941, 0.985) |
Reference | 0.984 (0.885, 0.998) |
Reference |
| Tomography | 26 (27,267) | 0.993 (0.985, 0.996) |
0.042 | 0.986 (0.976, 0.992) |
0.910 | |
| Topography | 21 (3,579) | 0.965 (0.931, 0.983) |
0.744 | 0.978 (0.958, 0.989) |
0.756 | |
| Data type | Images | 13 (27,532) | 0.980 (0.950, 0.992) |
Reference | 0.975 (0.947, 0.988) |
Reference |
| Parameters | 39 (22,792) | 0.987 (0.976, 0.947) |
0.461 | 0.984 (0.975, 0.990) |
0.342 | |
CI: confidence interval.