Skip to main content

Table 2 Performance of the regression models for post-operative vault prediction

From: Predicting post-operative vault and optimal implantable collamer lens size using machine learning based on various ophthalmic device combinations

Device (s)

Top 3 algorithms

Validation set

Test set

R2 (95% CI)

MAE (95% CI)

R2 (95% CI)

MAE (95% CI)

UBM

et

0.398 (0.380 to 0.416)

145.056 (142.161 to 147.961)

0.400 (0.382 to 0.418)

144.856 (141.936 to 147.709)

rf

cbt

Pentacam

et

0.369 (0.352 to 0.387)

149.026 (146.302 to 151.788)

0.363 (0.345 to 0.380)

149.174 (146.292 to 152.051)

cbt

rf

Sirius

et

0.410 (0.392 to 0.427)

143.577 (140.580 to 146.573)

0.404 (0.386 to 0.422)

144.017 (140.997 to 147.032)

cbt

rf

UBM and Pentacam

et

0.450 (0.426 to 0.475)

137.316 (134.338 to 140.143)

0.452 (0.427 to 0.477)

137.046 (134.078 to 140.092)

cbt

lgb

UBM and Sirius

cbt

0.467 (0.439 to 0.494)

132.698 (129.679 to 135.779)

0.468 (0.442 to 0.494)

132.985 (128.982 to 136.821)

et

lgb

Pentacam and Sirius

cbt

0.418 (0.402 to 0.435)

141.930 (138.869 to 145.154)

0.417 (0.400 to 0.433)

141.890 (138.807 to 144.869)

et

lgb

UBM and Pentacam and Sirius

cbt

0.504 (0.480 to 0.527)

129.893 (127.758 to 132.046)

0.499 (0.470 to 0.528)

130.655 (128.949 to 132.111)

et

lgb

Only other devices

et

0.208 (0.175 to 0.242)

175.31 (169.631 to 181.099)

0.210 (0.181 to 0.240)

172.132 (167.290 to 176.934)

rf

cbt

  1. cbt CatBoost regressor; et Extra trees regreessor; rf Random forest regressor; gbr Gradient boosting regressor; lgb Light gradient boosting machine; MAE mean absolute error; CI Confidence interval