Table 1.
Performance of methodological comparison among algorithms.
| Outcome | Accuracy | Sensitivity | Specificity | |
|---|---|---|---|---|
| Naïve Bayes | VAO | 0.966 (0.916–0.991) | 0.964 (0.875–0.996) | 0.969 (0.892–0.996) |
| High IOP | 0.975 (0.928–0.995) | 0.972 (0.902–0.997) | 0.979 (0.889–0.999) | |
| Random forest | VAO | 0.950 (0.894–0.981) | 0.946 (0.849–0.989) | 0.953 (0.869–0.990) |
| High IOP | 0.941 (0.883–0.976) | 0.944 (0.862–0.984) | 0.938 (0.828–0.987) | |
| Neural network | VAO | 0.950 (0.894–0.981) | 0.909 (0.801–0.970) | 0.984 (0.916–0.999) |
| High IOP | 0.933 (0.872–0.971) | 0.986 (0.924–0.999) | 0.854 (0.722–0.939) |
VAO visual axis opacification, IOP intraocular pressure, accuracy (TP + TN)/(TP + TN + FP + FN), sensitivity TP/(TP + FN), specificity TN/(TN + FP), TP true positive, TN true negative, FP false positive, FN false negative, CI confidence interval.
A methodological comparison among naive Bayes, random forest, and neural network was performed using the average performance of the 5-fold cross-validation by the training set of 594 patients.