Table 3
Results comparison in different axial length groups
| AL group | Number of cases (%) | Wilcoxon-test p value | Method | ±SD | MedAE |
| Short (AL≤23 mm) | 589 (17.5) | <0.01 | OKULIX | 0.394±0.333 | 0.313 |
| ML | 0.373±0.328 | 0.290 | |||
| % improvement | 5.3% | 7.3% | |||
| Medium (23 mm<AL≤26 mm) | 2429 (72.4) | 0.22 | OKULIX | 0.340±0.306 | 0.265 |
| ML | 0.343±0.307 | 0.270 | |||
| % improvement | −0.9% | −0.5% | |||
| Long (AL >26 mm) | 339 (10.1) | <0.01 | OKULIX | 0.409±0.361 | 0.322 |
| ML | 0.365±0.328 | 0.271 | |||
| % improvement | 10.8% | 15.8% |
This table shows the performance of OKULIX and ML-based approaches for patients in the short, medium and long axial length group. The per cent improvement was calculated as
. The Wilcoxon test was performed to compare the difference in the absolute errors between OKULIX and ML-based approaches in three different axial length groups.AL, axial length; MAE, mean absolute error; MedAE, median absolute error; ML, machine learning.
